Pricing the Immersion Option - Data-Center Liquid Cooling Market, the OCP Standards Map, and a Pure-Immersion Acquisition Window

1. Liquid Cooling Market Overview: TAM, CAGR, and Data-Center Share

Data-center liquid cooling is being told as two starkly different stories at once: "rapid expansion" and "hard to quantify." The former reflects industry consensus—AI accelerators have pushed rack power density onto an unprecedented step-change, turning liquid cooling from an alternative into a functional necessity. The latter exposes an analytical difficulty: under the same label "data-center liquid cooling market," the 2028–2032 size forecasts range, depending on the firm, from roughly $7 billion to roughly $28 billion—a gap of nearly 3×. This chapter first clarifies the definitional divergence in §1.1, then gives market size, technology segmentation, and regional evolution under a single anchor definition in §1.2; §1.3 then examines liquid cooling's true share through two different denominators—the entire data center vs. the frontier AI rack—and §1.4 finally unpacks the growth drivers and the inflection-point friction. All market figures are tagged with firm, definition, and year; figures from different firms cannot be directly summed or compared side by side.

1.1 Definitions and Methodology: Why the Figures Differ by Nearly 3×

Most leading firms' CAGR judgments for the 2028–2032 liquid cooling market fall within a similar 20%–33% band—the disagreement is not over "how fast it grows" but over "how large it is measured to be."

1.1.1 Three Categories of Reasons for Definitional Divergence

The first layer of difference is "which revenue tier is counted." Dell'Oro confines the liquid cooling market to vendor revenue for liquid-cooling thermal-management hardware (cold plate, immersion, rear-door heat exchange): about $3 billion in 2025 and close to $7 billion in 2029 [S-001]. MarketsandMarkets [S-064] and The Business Research Company [S-065] layer solutions and services—installation, operations and maintenance, integration—on top of the hardware, raising the size by 1.5× to 3×.

The second layer is the rolling upward revision of the time window and AI assumptions. A single firm will rewrite the same metric within two years: MarketsandMarkets' 2023 release of "$7.8 billion in 2028" was revised up in 2025 to "$21.1 billion in 2032," and the current definition pushes it further to "$27.7 billion in 2033" [S-064]. Any cross-firm comparison must lock to the same version and the same cut-off year; otherwise it is comparing definitions, not comparing markets.

1.1.2 Cross-Firm Bridge Table

Placing the same-period figures, the definitions, and the relative position to the anchor in one table makes the source of the divergence immediately clear.

Table 1.1: Cross-Firm Liquid Cooling Market Definition Bridge

Setting aside the definitional noise, three points can be pinned down: the mainstream CAGR is 20%–33%; the narrow definition for 2028–2030 lands at about $6–8B; the broad definition adds another 1.5× to 3× for services/integration. The Omdia figure includes air cooling and cannot be placed alongside the others.

1.2 Market Size and Growth Rate

1.2.1 Dell'Oro Anchor Definition and Annual Trajectory

Under the Dell'Oro definition, the liquid cooling market starts at about $1.5 billion in 2024, doubles to about $3 billion in 2025, and approaches $7 billion in 2029; within it, direct liquid cooling (DLC)—the largest segment, disclosed separately—is $1.1 billion in 2024 and $5.8 billion in 2029 [S-001][S-060]. In July 2024 Dell'Oro also disclosed a 5-year cumulative figure (2024–2028 >$15 billion), consistent with the endpoint trajectory above [S-140], but the three single-year figures for 2026/2027/2028 are not publicly released (they sit inside the paid Advanced Research report). Those years in the table below are interpolated from the endpoints and the growth rate, and are explicitly marked "estimated."

Table 1.2: Dell'Oro Liquid Cooling Market Trajectory 2024–2029 ($ billions)

The doubling in 2025 is a direct result of accelerating AI deployment; growth thereafter reverts to the mid-20% range. DLC's share of LC rises from about 73% in 2024 to about 83% in 2029—cold plate continues to absorb the vast majority of the incremental growth. Dell'Oro characterizes this phase as liquid cooling "becoming a functional necessity for large-scale AI deployment" [S-001].

1.2.2 Technology Segmentation and Coolants

Liquid cooling's internal shares are only coarse-grained in public figures. Mordor estimates that in 2025 direct-to-chip (DTC) accounts for about 42.85% of the liquid cooling market [S-142]; Grand View's immersion-segment report puts single-phase immersion at 70.6% of the entire immersion market (2024) [S-143]; Dell'Oro qualitatively confirms that "single-phase DLC dominates" [S-001]. The four fine-grained shares (DTC / single-phase immersion / two-phase immersion / RDHx) are not disclosed on public pages; this section uses the qualitative reading "DTC dominant, single-phase immersion far ahead of two-phase, and two-phase shrinking under tightening PFAS rules."

Table 1.3: Liquid Cooling Technology Segmentation (2024–2025 public figures)

Coolants must be kept separate from the system definitions above. The contemporaneous standalone "immersion systems" figure (Grand View): $286.8M (2024) → $1,006.6M (2030), CAGR 23.6% [S-143]. Other figures (Mordor and others) give a higher absolute value for "immersion coolants" (the $2–4B range), but they consolidate multiple classes of fluorinated and hydrocarbon dielectric fluids across industrial uses—direct DC immersion is only one part of that.

Table 1.4: Immersion Systems vs. Immersion Coolants — Definition Comparison

Do not treat "immersion coolants $2–4B" as a share of the DC liquid cooling market—the latter is overwhelmingly cold-plate hardware, while coolants correspond to the immersion subset plus shared industrial cooling. This report treats coolants as a separate input to Ch2 §2.5 and Ch3 (M&A), and does not commingle them with the liquid-cooling-system market definition.

1.2.3 Regional Distribution: United States, China, and the Rest

Liquid cooling adoption is led by the United States and China, with the rest of Europe/Asia following; the two countries' driving logics differ.

The United States is the largest liquid cooling market today, driven by hyperscaler capital expenditure and AI compute deployment. It has no unified national PUE ceiling (there are voluntary requirements and sustainability disclosures at the state/utility level); instead it is pressured in the opposite direction by water scarcity and community power-carrying capacity. IMARC's current public page gives the U.S. data-center liquid cooling market reaching about $3.45B by 2034, CAGR 15.76% (2026–2034) [S-144]; an earlier version once disclosed year-by-year point estimates of $759.4M for 2024 and $1.30B for 2025, but these are inconsistent with IMARC's current figures (possibly from an earlier paid snapshot), so this report does not adopt the early point estimates and follows the current CAGR and endpoint. On regional concentration, Mordor estimates that hyperscalers account for 47.02% of U.S. DC cooling (2025) [S-145]—far below the commonly heard "76% hyperscale" rumor (for which no corresponding table was found on public pages), so this report uses 47% as the basis. At the state/cluster level, Virginia (Northern Virginia, Loudoun County), Texas (Dallas/Austin), Oregon (Portland), and Arizona are hyperscaler concentration areas; states with severe water scarcity (Arizona in particular) are becoming priority adoption zones for DLC + dry cooling towers.

China follows a policy-driven path. Credence Research gives China's DC liquid cooling market at $497.62M in 2024 and $3,349.75M in 2032 (CAGR 26.92%) [S-133]; Mordor simultaneously shows that air cooling still accounted for 63.7% of China's DC cooling in 2024, with liquid cooling CAGR at 18.3% [S-134]. On policy: MIIT's 2021 three-year action plan first set a PUE <1.3 red line for large/hyperscale DCs (<1.25 in cold regions) [S-157]; the 2022 "Eastern Data, Western Computing" program launched and allocated PUE ceilings to hub clusters on a tiered basis—western hubs ≤1.2, eastern hubs ≤1.25 [S-159]; the 2024 joint action plan by NDRC/MIIT/the National Energy Administration/the National Data Administration tightened further: by the end of 2025, new-build and expansion/retrofit large and hyperscale DCs PUE ≤1.25, national hub nodes ≤1.2 [S-132][S-156]. This policy curve is about 14% tighter than the 1.46 measured average of China's hyperscale DCs in 2019 [S-158], and the room left for air cooling is being continuously compressed.

Other regions (Europe, Asia-Pacific excluding China, Latin America, etc.) are mostly back-derived as a residual from the global total minus US+CN in public figures—there is no clean first-hand regional breakdown, so this report tags them as "residual/estimated."

Table 1.5: Liquid Cooling Regional Evolution 2024 → 2032/34

The U.S. and Chinese liquid cooling markets each account for roughly 1/3 of the global total (rough calculation), with the remaining ~1/3 scattered across the rest of Europe/Asia. China's CAGR is markedly above the global average (26.92% vs. ~25%), reflecting the dual overlay of policy and new infrastructure build-out; the U.S. CAGR is slightly lower (15.76%) but on a large base, with the highest absolute increment.

1.3 Share Within the Data Center

How "important" liquid cooling is cannot be answered by a single number—the same facts read as entirely different stories under different denominators. §1.3.1 looks at the dollar share of total data-center physical infrastructure (DCPI), §1.3.2 looks at the cost share within the most frontier AI racks, and §1.3.3 uses penetration and rack density to connect the two.

1.3.1 Share of DCPI and Total Data-Center Capex

Dell'Oro raised its 2029 total for DCPI (including UPS / cooling / power distribution / racks / software and services) to about $63.1B, 5-year CAGR 15% [S-060][S-141]; it also set total data-center capex (including IT equipment) at $455B in 2024 (+51% YoY), surpassing $1T in 2029 [S-155]. Against this base, DLC at about $5.8 billion in 2029 is about 9% of DCPI; even on the broader liquid cooling definition (about $7 billion, including immersion/RDHx) it is only about 11%.

Table 1.6: DCPI and Total DC Capex Base 2024–2029 (partly interpolated)

(2024 DCPI is back-calculated from the endpoint + CAGR; the 2025–2028 single-year figures are interpolated; Dell'Oro's public PR does not disclose the intermediate years separately—confirmation requires the paid Advanced Research report.)

Under the narrower "cooling capex" denominator the share would be higher (cooling is only one part of DCPI), but Dell'Oro does not publish that figure; it is logged as a known data gap in §Conclusions.

1.3.2 Share of Frontier AI Rack Cost

Placing the rack BOM relayed by Morgan Stanley via Tom's Hardware alongside various analysts' estimates of the full-rack ASP makes liquid cooling's "content ratio" clear.

Table 1.7: Liquid Cooling BOM Share in Frontier AI Racks (verified)

The liquid-cooling cost breaks down as 18 compute trays × $2,260 = $40,680 + 9 NVSwitch trays × $1,020 = $9,180 = $49,860 (GB300) [S-130]. The full-rack ASP comes from analyst channels (the Morgan Stanley report is not public; NVIDIA has not confirmed official pricing) [S-130][S-131].

There are three layers of meaning. First, liquid-cooling hardware accounts for only about 1% of the frontier AI rack, and the ratio actually declines as rack specifications are upgraded (GB200 ~1.4% → GB300 ~0.8% → Rubin NVL144 ~0.7%)—the GPU pulls the full-rack ASP up far faster than liquid-cooling cost. Second, this means liquid cooling is not the rack's cost center but an "enabling bottleneck": it is only about 1%, but without it the roughly $7M of compute in the full rack cannot be deployed. Third, the pricing implication for acquisition: acquiring an immersion/cold-plate vendor is not buying "X% of the rack cost" but buying "the supplier of a necessary component for AI compute deployment, and a standards foothold"—a judgment that runs through §2.4 and Ch3's "price as an option, not a cash cow."

Scope note: the "liquid-cooling share" in this section refers to the in-rack liquid-cooling hardware BOM—cold plate + manifold + QD + part of the CDU—excluding facility-level cooling (outdoor heat source, FWS pumps, water treatment, cooling tower / dry cooler), which is counted separately under the "cooling" line within DCPI in §1.3.1.

1.3.3 Penetration and Rack Density

Penetration figures under the facility definition and the AI definition appear to conflict, but the denominators differ. Uptime's 2024 cooling-systems survey: across all facilities, 22% already use direct liquid cooling and 61% say they would consider it, with the adoption rate rising with facility size (<5 MW about 13%, 20 MW+ about 38%) [S-095]; the typical average rack power is still only about 8 kW [S-102]. When TrendForce switches the denominator to AI data centers, liquid cooling penetration rises from 14% in 2024 to 33% in 2025 [S-004]. The two sets of figures do not contradict each other—they show that liquid cooling penetration spreads not by facility count but by pinpoint insertion into high-density AI racks.

Table 1.8: DLC / Liquid Cooling Penetration Time Series

The historical curve shows Uptime's all-facility adoption rate creeping from ~17% in 2022 to 22% in 2024, while AI-DC penetration jumped from 14% to 33% over the same period—the inflection has already occurred in the AI subset, but the full facility base has not yet caught up. This confirms that the two seemingly very low figures—"~9% of DCPI by dollars" in §1.3.1 and "~1% of the rack BOM" in §1.3.2—can hold simultaneously with "33% penetration in the AI subset."

On the density threshold, Uptime reports that traditional perimeter cooling tops out at about 20–25 kW/rack; above that line, DLC is "better" on both economics and energy efficiency [S-139]. A consultant model estimates that DLC crosses over air on 10-year TCO at ~30 kW/rack and $0.12/kWh (consultant/advisory figure, non-standard, illustrative only) [S-138]. Compared with the 60–140 kW level of AI racks, this is well within liquid cooling's application range—the physical driver that §1.4 unpacks.

1.4 Growth Drivers and the Inflection Point

Demand for liquid cooling is pushed by two forces—the step-change jump in rack power density (a supply-side physical constraint) and the exponential growth of hyperscaler capital expenditure (demand-side capital volume)—while being slowed by one friction: the retrofit economics of the installed facility base. These three forces make up §1.4.1–§1.4.3.

1.4.1 The Step-Change in Rack Power Density

Over the past 6 years, rack power density has jumped from the industry-typical 6–8 kW to the 120–600 kW of AI full-rack solutions; air cooling has already hit its practical limit in the 20–40 kW range. Placing Uptime's "typical rack" trajectory alongside NVIDIA's full-rack solutions makes the step-change immediately clear.

Table 1.9: Rack Power Density Evolution 2019–2027

From 8 kW to 600 kW, frontier AI rack power density scaled up about 75× in 5 years, while the industry-typical rack barely moved—precisely why liquid cooling "inserts at pinpoint into high-density AI" rather than "spreading by facility count." Air cooling's practical 20–40 kW ceiling is left far behind [S-139], and 600 kW is entirely within the range that only liquid cooling, or denser forms (single-/two-phase immersion, double-sided cold plate), can deploy. This is the physical bedrock of liquid cooling shifting from "optional" to "necessary."

1.4.2 Demand Side: Hyperscaler Capex and NVIDIA Shipments

Hyperscaler capital expenditure has taken an "exponential + accelerating" shape over the past 4 years, directly driving incremental demand for liquid cooling. Using SEC 10-K / earnings-report first-hand data, here is each of the top five hyperscalers' year-by-year capex.

Table 1.10: Top Five Hyperscaler Capex 2022–2026 ($ billions)

Note: CreditSights / Introl estimate the top-5 total at ~$256B for 2024, ~$443B for 2025, and ~$602B for 2026, which differ markedly from the direct sum of each company's SEC filings (2024 ~$219B, 2025 ~$376B)—the former's basis may include investments/acquisitions or unlisted operating-lease items. This report anchors on the SEC first-hand sum, with the CreditSights figures as a secondary cross-reference [S-160].

Hyperscaler capex jumped from about $144B in 2023 to $376B in 2025 (2.6×, in under 2 years)—directly tied to Dell'Oro's doubling of the 2025 liquid cooling market [S-001]. Considering further that CreditSights estimates about 75% of 2026 hyperscaler capex (~$450B) flows into AI infrastructure [S-160], and that liquid cooling is a hard-constraint component of AI compute, the demand-side pull is structural.

NVIDIA's data-center GPU shipments are the other face of this demand. Starting from about 100K H100 in 2022, to about 420K GB200 in the second half of 2024, and an expected 1.5–2M units of Blackwell in total in 2025; analysts estimate that a ~60K-rack scenario in 2026 would correspond to about 4.3M GPUs—a roughly 40× scale-up in 4 years. To be clear: these figures come mainly from SemiAnalysis and supply-chain estimates; NVIDIA does not publicly disclose annual shipments by SKU, so this report tags them as "analyst/supply-chain figures"; claims such as "3.6M units of backlog sold out through mid-2026" were not found in public primary sources and are not adopted by this report.

Table 1.11: NVIDIA Data-Center GPU Shipment Estimates (analyst/supply-chain figures)

1.4.3 Inflection-Point Friction: PUE Stagnation and Retrofit Economics

Liquid cooling is already a physical necessity, but the transition of the entire data-center base is not frictionless. Retrofitting an existing air-cooled facility for liquid cooling requires adding secondary-side piping, manifolds, QDs, CDUs, and primary-side heat rejection (dry cooler / cooling tower); in most cases the capex is no lower than a new build, plus the switch in the operations and maintenance regime—this concentrates liquid cooling penetration in AI incremental volume and new-build/large facilities rather than across the entire data-center base.

Indirect corroboration is that the industry-average PUE has stagnated for years. The global average published by Uptime hovered within the narrow 1.55–1.59 band from 2020 to 2024.

Table 1.12: Global Data-Center PUE Evolution 2007–2024 (Uptime)

Data source: Uptime 2024 Global Data Center Survey [S-102].

PUE fell from 2.5 to 1.58 over the decade 2007–2018, but stalled after 2020: the low-hanging optimization fruit (hot/cold aisle containment, raising supply-air temperature, variable-frequency water pumps) is essentially exhausted, while liquid cooling can push PUE lower but requires larger retrofit investment to take effect. This also explains why China takes the policy-mandate path: MIIT in 2021 required new-build large DCs to have PUE <1.3 [S-157]; in 2022 Eastern Data, Western Computing brought hubs down to ≤1.2 / 1.25 [S-159]; in 2024 the NDRC joint action plan tightened further to ≤1.25 for large facilities nationwide and ≤1.2 for hubs [S-132][S-156]. Against a backdrop lacking intrinsic market incentive to break the PUE stagnation, administrative means push liquid cooling from "economically optional" to "compliance-mandatory."

A final supporting factor is the standardization follow-through from ASHRAE TC 9.9: on top of the existing W-class (facility-water temperature classes, long established), it added the S-class (secondary-side supply-liquid temperature classes for the technology cooling system) [S-135], providing a temperature-class baseline for standardized liquid-cooling adoption under different facility conditions. This lowers the design and procurement friction of both retrofit and new build, but does not by itself solve economics—the economics are solved jointly by the hyperscaler expansion curve (§1.4.2) and China's policy mandate (this section).

2. Liquid Cooling Approaches Compared: Cold-Plate vs. Immersion Architectures

Data-center liquid cooling is not one technology but a continuous spectrum from cold plate to immersion. Once rack power density jumps from the 10 kW class to GB300's 132–140 kW and Rubin Ultra's 600 kW, "which path to choose" begins to determine the data center's physical form—whether to deliver coolant to the chip surface, or to submerge an entire piece of IT equipment in liquid.

2.1 Technology Lineage and This Chapter's Definitions

2.1.1 Cooling Spectrum Matrix

Data-center cooling is a continuous spectrum: air cooling → rear-door heat exchanger (RDHx) → cold-plate direct liquid cooling (DLC) → single-phase immersion → two-phase immersion. Along this spectrum, the cooling medium gets closer and closer to the heat source and the bearable rack power density gets higher and higher, but deployment and ecosystem maturity decline in the opposite direction.

Table 2.1: Liquid Cooling Technology Spectrum Matrix

Cold plate covers the current AI mainstream density (120–140 kW) and has been verified to support the next-generation Kyber rack's 600 kW (paired with denser forms); immersion—two-phase in particular—has higher theoretical density, but commercial deployment above 100 kW is rare, and two-phase is constrained by coolant supply (see §2.5).

2.1.2 OCP Terminology Reference

A liquid cooling system generally consists of two loops—the TCS (Technology Cooling System, the secondary/technology side) directly contacts the IT equipment, and the FWS (Facility Water System, the primary/facility side) connects to outdoor heat rejection; the two are isolated by a CDU (Coolant Distribution Unit) via a plate heat exchanger [S-090].

2.2 Cold-Plate (Direct-to-Chip) Architecture

2.2.1 Component List

A cold-plate system has a clear component chain running outward from the chip. By commercialization status, each link has a relatively concentrated set of suppliers.

Table 2.2: Cold-Plate System Component List

2.2.2 Working Principle and Thermal Path

Thermal path: chip/package → TIM/IHS → cold plate → secondary-side TCS coolant → in-rack manifold + hose + QD → CDU plate heat exchanger → primary-side FWS → dry cooler / cooling tower / chiller / outdoor heat rejection [S-090].

The cold plate removes heat only from high-heat-flux components such as the chip; the remaining parts—memory, NICs, power supplies, etc.—still need some air cooling retained, so a cold-plate facility is usually a "liquid + air" hybrid rather than pure liquid cooling [S-101]. The GB200/GB300 NVL72 full-rack solution is exactly this paradigm—the cold plate handles the GPU/CPU, fans handle the rest.

2.2.3 CDU Form-Factor Comparison

The CDU is not a single product form. By deployment location and capacity, leading vendors offer four classes: in-rack, in-row, sidecar (wall-mounted/side cabinet), and facility (facility-level L2L/L2A). Capacity ranges all the way from ~70 kW to 2.5 MW, corresponding to different deployment scales and cadences.

Table 2.3: CDU Form-Factor Comparison

Form-factor choice follows the deployment path "in-rack pilot first → in-row cluster → facility consolidation"; newly built AI facilities usually go straight to the facility level from the start.

2.2.4 Interface Standardization: UQD/UQDB/Stäubli QDs and ORv3 Version Evolution

Standardization in the cold-plate space concentrates on two lines—the OCP UQD/UQDB universal quick-disconnect standard and the ORv3 rack-level liquid cooling interface specification. The version evolution of both lines roughly aligns with the 2019–2026 mainstreaming cadence of liquid cooling.

Table 2.4: Liquid Cooling Interface Standards and Version Evolution

Interface standardization has matured to the point of "connecting IT equipment with no exposed valves" (blind-mate), pushing the operational complexity of liquid-cooling deployment toward the level of air cooling. This infrastructure-layer standardization is the invisible support that lets cold plate deploy at scale across the AI mainstream.

2.3 Immersion Architecture

2.3.1 Component List (cross-referenced to the architecture diagram)

Table 2.5: Immersion System Component List

2.3.2 Single-Phase Immersion Working Principle

The dielectric fluid never boils; it carries heat away from the components via forced convection (pumped circulation) or natural convection; the liquid is cooled by the CDU/heat exchanger and returns to the tank [S-070][S-096]. Thermal path: components → dielectric fluid → in-tank circulation / CDU heat exchange → primary-side FWS → outdoors. Commercial deployment makes up the absolute majority of immersion—Grand View data show single-phase at 70.6% of the entire immersion market (2024) [S-143], because the coolants (PAO/synthetic ester/mineral oil) have stable supply and fewer chemical-compatibility issues.

2.3.3 Two-Phase Immersion Working Principle

The dielectric fluid boils at the surface of the hot components, the vapor rises to the condenser coil at the top of the tank and condenses, and the condensate returns to the tank by gravity [S-096][S-097]. In theory the tank can be pump-free—circulation driven solely by phase change and gravity. Thermal path: components → boiling dielectric fluid → vapor headspace → condenser coil → gravity return of condensate → primary side. But actual systems usually still need auxiliary pumps and fans to control flow direction and condensation efficiency—"completely pump-free" is a mechanistic statement, not the reality of all commercial systems.

2.3.4 Single-/Two-Phase Quantitative Comparison

Most publicly available two-phase vs. single-phase comparison figures come from vendor blogs or engineering reviews; standardized measurement data are very limited. The table below presents them faithfully and tags their basis.

Table 2.6: Single-Phase vs. Two-Phase Immersion Quantitative Comparison (mostly vendor-stated)

Two-phase has a physical advantage, but commercialization is in retreat—coolant supply is choked by PFAS regulation (see §2.5), and the fact that a large share of public figures are vendor-stated rather than independently tested means it has contracted to a niche in actual 2024–2026 deployments. This report's Ch3 immersion-acquisition discussion takes single-phase as the default backdrop.

2.4 Multi-Dimensional Comparison

2.4.1 ASHRAE Temperature Classes: W1–W5 and the New W40

ASHRAE established a class system for liquid-cooling facility-water temperature; the 2011 first edition was W1–W5, and the 5th edition (recent) renumbered them and added W40. Each class corresponds to different free-cooling feasibility and energy efficiency.

Table 2.7: ASHRAE Liquid-Cooling Facility-Water Temperature Classes

Data source: ASHRAE Thermal Guidelines 5th Edition reference card + liquid-cooling white paper [S-099][S-104]. W40 was added specifically for the next-generation liquid-cooling designs with 40°C inlet—directly tied to cold-plate/immersion systems' ability to still operate at higher water temperatures, and a substantive advance in liquid-cooling standardization.

2.4.2 Performance and Energy-Efficiency Comparison

Uptime's 2024 industry-average PUE is 1.56, with new builds reaching ~1.3 [S-102]; liquid cooling can indeed lower PUE, but Uptime cautions that "liquid cooling's efficiency expectations need a reality check"—the gain depends heavily on the metering boundary, climate, load, and whether IT fan power is counted [S-094]. The commonly cited "two-phase immersion 1.03–1.08 / cold plate 1.08–1.15" is vendor-stated, illustrative only [S-098]. 22% of operators already use DLC and 61% say they would consider it (Uptime 2024 survey [S-095]), with adoption rising with facility size (<5 MW ~13%, 20 MW+ ~38%); current DLC installations are dominated by water cold plate, and immersion has not yet been widely deployed in AI clusters [S-103][S-101].

2.4.3 Retrofit Economics (vendor-stated, illustrative only)

Retrofitting an existing air-cooled facility for liquid cooling requires adding secondary-side piping, manifolds, QDs, CDUs, and primary-side heat rejection (dry cooler / cooling tower); in most cases the capex is no lower than a new build, plus the switch in the operations and maintenance regime. The figures in the table below come mainly from a single vendor blog and are for directional industry reference only, not as inputs to an investment model.

Table 2.8: Retrofit vs. New-Build Economics (vendor-stated, illustrative only)

2.4.4 Reliability and Maintainability

The maintainability difference between cold plate and immersion comes mainly from three dimensions: leak detection, hot-swap difficulty, and the blast radius of a single point of failure.

Asetek's RackCDU makes CDU monitoring (temperature, flow, pressure, leak sensing) a software component [S-176][S-177]; the research community already has IoT/ML approaches claiming 96.5% accuracy for leak detection (validated on a synthetic dataset, not actual measurement) [S-175]. MTBF, hot-swap time, and field failure rate are rarely disclosed in leading vendors' first-hand specs, so this report does not force specific figures. Asetek claims cumulative sales of 6M+ cooling systems [S-176]—the basis includes consumer PC cooling, not all of it data center, and this report flags that explicitly when citing it.

On maintainability, the cold-plate architecture is close to "air cooling++": when a single rack or single server is pulled out for maintenance, the QD self-seals and does not affect neighbors; immersion is the opposite—lifting equipment out of the fluid for maintenance is a routine operation requiring dedicated tools and procedures (hoisting, fluid recovery/filtering), one of the operational reasons it remains niche.

2.4.5 Regional Adoption Preferences (US / Japan / China / HPC)

Liquid-cooling adoption is not uniform across regions. Credence Research estimates North America about 35%, Asia-Pacific about 30% (the fastest growth) [S-182]. The three major regions show clear differences in form-factor preference:

• United States: hyperscaler-led, with cold plate as the absolute mainstream, paired with a small amount of single-phase immersion piloting (Microsoft, Meta); water-stressed states (Arizona, Utah) push dry-cooler + DLC.
• Japan: HPC prefers cold plate—the Fugaku supercomputer uses Fujikura microchannel-fin cold plates [S-183]—not immersion; the follow-on FugakuNEXT (RIKEN) continues the liquid-cooling path and incorporates NVIDIA Blackwell.
• China: single-phase immersion has the longest deployment history—Alibaba's "soaking server" has been deployed since 2015, claiming a ~70% reduction in cooling energy consumption [S-184]; meanwhile cold plate is penetrating new-build AI facilities rapidly, layered on top of the NDRC policy mandate of PUE ≤1.25/1.2 (see §1.2.3 / §1.4.3).
• HPC laboratories (US / Europe / Japan): historically both cold plate and immersion (the Cray T90 used it as early as 1995); modern HPC clusters have essentially returned to the cold-plate path.

Globally, cold plate = the mainstream path, while immersion holds an important position in China and some HPC scenarios.

2.5 Coolant and Material Constraints: Why Immersion Has Remained Niche

Immersion's Achilles' heel—two-phase in particular—is the coolant. On 2022-12-20, 3M announced it would exit PFAS (including fluorinated fluids) manufacturing by the end of 2025 [S-012b]; the supply of the fluorinated dielectric fluids that two-phase relies on (the Novec/Fluorinert series, Galden PFPE) is therefore directly hit. Specific claims such as "discontinuing Novec/Fluorinert" and "last order 2025-03-31" are distributor-stated, not explicitly stated by 3M (see ledger C4).

2.5.1 Coolant Family Matrix

DC dielectric coolants fall broadly into five families; the single-phase mainstays are PAO / synthetic ester / GTL / synthetic hydrocarbon / mineral oil, while two-phase relies on fluorinated fluids.

Table 2.9: Data-Center Dielectric Coolant Families

The OCP/ACS Omega paper gives the general requirements for immersion coolants: volume resistivity ≥ 2.0 GΩ·m, dielectric breakdown ≥ 35 kV—all the families above can meet these when the grade is chosen appropriately [S-174].

2.5.2 PFAS and the 3M Exit Timeline

Table 2.10: PFAS Regulation and Coolant Supply Timeline

2.5.3 Substitute Supply Chains and the Vendor Map

After PFAS tightening, the single-phase immersion supply chain has actually expanded—large petrochemical companies have rolled out data-center-specific PAO/synthetic-hydrocarbon products one after another. This leaves the immersion space with an interesting configuration: the coolant supply base is diverse and does not constitute a single-point risk (for single-phase), but two-phase is locked to one or two substitute suppliers.

The vendor map of the entire liquid-cooling industry, distributed by category, is as follows (by global public figures, top 3–5 per category):

Table 2.11: Leading Liquid-Cooling Vendors by Category (top 3–5 per category)

GMI (Global Market Insights, global data-center liquid-cooling market) lists Schneider Electric, Vertiv, Rittal, Stulz, and Boyd as star players (Supermicro is not included) [S-185]; this report takes that as the benchmark. This map feeds directly into Ch3's acquisition discussion—the acquirable independent immersion players (GRC, Submer, Iceotope, Asperitas/Promersion) sit in the "immersion systems/tanks" category, in a different pool from the cold-plate/CDU giants.

Ch1 said that "AI compute turns liquid cooling from optional into necessary," and this chapter says that "of the two paths, cold plate is the current mainstream while immersion is still constrained by coolant"; Ch3 then covers who holds the pen on the specifications in the OCP immersion ecosystem; Ch3 screens acquisition targets within the immersion ecosystem—beginning with the next chapter.

3. The OCP Immersion Liquid Cooling Influence Map: From Governance Architecture to Company Ranking

Say over the OCP immersion space is allocated along three axes: the "PSC + 4 TC + 17 workstream" governance architecture established by Charter Rev 1.0 (2024-12-19); the 5 contribution processes (Specifications 9 steps / Design Files 7 steps / Software 11 steps / Documents 9 steps / Product Recognition 2 steps); and the SC dual channel. This chapter deconstructs it in the following order: ① the OCP governance coordinates—immersion's position under the Cooling Environments umbrella + the 2024-12 4-TC restructuring + the dual channel + the procedural/content two-dimensionality (§3.1); ② the actual distribution of the 5 contribution processes across the immersion space + each of the 4 TCs' representative specs + the typical path of Requirements Rev 2.10 + the dual-channel trigger mechanism at step 7 (§3.2); ③ the complete author rosters of the 7+ core specs (§3.3); ④ the complete roster of the 4 TC + 17 workstream chairs (§3.4); ⑤ the precise key-individual ↔ company mapping (§3.5); ⑥ the company-dimension 4 TC × influence matrix (§3.6); ⑦ the final influence ranking of 28+ companies (§3.7); ⑧ the five-tier company layering (§3.8); ⑨ the 2019–2026 influence evolution over time (§3.9); ⑩ the implications for the Ch4 acquisition framework (§3.10).

All personnel affiliations, spec authors, and workstream chairs are based on 2026-05-28 cross-referencing of multiple first-hand sources—OCP PDFs + wiki + Foundation pages + LinkedIn; second-hand or inferred points are flagged.

3.1 The OCP Governance Coordinates of Cooling Environments / Immersion

3.1.1 Immersion's Position Under the Cooling Environments Umbrella

Within OCP, immersion is not a standalone top-level Project but one of the 6 Subprojects under the Cooling Environments Top-level Project Community—alongside CDU, Cold Plate, Door HX, Heat Reuse, and HMLC (Hardware Management Liquid Cooling) [S-187]. The Cooling Environments parent project itself has two leads: Sean Sivapalan (NVIDIA) + John Fernandes (Meta) [S-220]—NVIDIA's presence in this parent-project lead seat is NVIDIA's "indirect but stable" entry into immersion-space governance.

At the spec-document level there is an easily overlooked detail: current specs still carry the early "ACS / Advanced Cooling Solutions" prefix (e.g., the S-070 filename ocp-acs-immersion-requirements-rev-2-1-pdf), but the first-hand OCP directory [S-187] and the current wiki [S-221] have standardized on "Cooling Environments" as the Top-level Project name. This report retains the ACS prefix when citing spec documents and uses Cooling Environments when referring to the project structure.

3.1.2 The Immersion Subproject Governance Profile and the 2024-12 Governance Restructuring

The Immersion subproject's current leads are co-leads Rolf Brink (Promersion) and Jessica Gullbrand (Intel) [S-220][S-221]—and Gullbrand is also the OCP SC Co-Chair for the 2026-2028 term. Brink is the lead author of every historical version of the OCP Immersion Requirements (the 2019 original under the Asperitas byline; 2022 Rev 2.0 and 2023-08 Rev 2.10 both under the Promersion byline) [S-070][S-071][S-228]; Gullbrand is an Intel Principal Engineer, at Intel continuously since 2006, and the engineering spokesperson in the immersion field [S-073].

But inside the Immersion subproject, a refinement of governance granularity took place on 2024-12-19—the OCP Immersion Project Charter Rev 1.0 [S-212] formally established the 2025-onward governance architecture of a PSC (Project Steering Committee) + 4 Technical Committees (Fluids / Solutions / Reliability / ITE). OCP's first-hand pages still call immersion as a whole a Sub-Project, but the governance granularity has been refined from the old two-tier "single sub-project + scattered workstreams" into a four-tier nesting of "Sub-Project → PSC → 4 TC → workstream." Under the new architecture, each TC governs several workstreams, and TC chairs are appointed with PSC approval and report uniformly to the PSC.

Table 3.1: Immersion Subproject 2024-12 Governance Restructuring Comparison

The significance of this restructuring is that the Immersion subproject was upgraded from an "open workflow" into a "small mini-program"—the 4 TC chairs now hold the say that used to be scattered at the workstream layer, and all 4 chairs come from specific companies: the Fluids TC is chaired by the petrochemical major BP (Fletcher), the Solutions TC by the hyperscale edge-computing operator Denvr Dataworks (Short), the Reliability TC by Intel (Yates), and the ITE TC by the IT-equipment integrator Unicom Engineering (Hipes). This chair distribution reveals the immersion space's current four-way interest structure: petrochemicals lead fluids, Intel leads reliability, operators lead systems, and ODMs/integrators lead IT adaptation.

3.1.3 The Dual Channel: Two Governance Pathways via the Cooling Environments ICR + the SC Co-Chair

The ICR for the Cooling Environments top level (the 6 Subprojects including immersion) in the OCP SC 2026-2028 term is Steve Mills (Meta) [S-206]—Mills also concurrently serves as the Heat Reuse Sub-project Lead under Cooling Environments. This means that when an immersion spec, after passing PSC review, is sent up to the SC for a vote, Meta (through Mills) is the default endorser.

But the immersion space has a second SC channel—Intel's Gullbrand is simultaneously the Immersion sub-project co-lead and an SC Co-Chair (one of two Co-Chairs, alongside Meta's Dharmesh Jani). This forms a unique "dual channel" structure:

Table 3.2: The SC Dual Channel for Immersion Specs

The path by which an immersion spec gets through the SC is mainly guarded jointly by Meta + Intel. Meta, through Mills's dual role of Cooling Environments ICR and Heat Reuse Lead, controls the overall endorsement of Cooling Environments; Intel, through Gullbrand's dual role of Co-Chair and Immersion co-lead, controls the agenda entry. When any key revision of an immersion spec, or a new spec, reaches the SC voting stage, it is almost impossible to pass without the tacit consent of these two companies. This governance structure is externally entrenched, and is the fundamental reason the immersion space is "niche yet stably ordered"—the "governance-channel continuity" risk that an acquirer must assess in due diligence.

3.1.4 Procedural Dimension × Content Dimension: The Orthogonal Relationship Between the 5 Contribution Processes and the 4 TCs

The Charter Rev 1.0 PSC + 4 TC architecture describes the hierarchical position of content review—which layer and which TC handles a spec's content. But OCP's contribution system has another independent axis: the 5 contribution processes (Specifications 9 steps / Design Files 7 steps / Software 11 steps / Documents 9 steps / Product Recognition 2 steps) describe the procedural steps. The two are orthogonal, and any immersion contribution operates on both dimensions at once.

Table 3.3: The Two-Dimensional Structure of an Immersion Contribution

Source: step data [S-190]; TC architecture [S-212][S-221]; dual channel [S-206]

This table aligns the §3.1.2 PSC + 4 TC architecture and the §3.1.3 dual channel on the same timeline. Key findings:

• Steps 1–6 are completed entirely within the Immersion sub-project (workstream → TC → PSC); the say of the 4 TCs is concentrated in these 6 steps.
• Step 7 is the only gate by which an immersion spec rises from inside the sub-project to OCP-wide—this is where the dual channel triggers.
• Steps 8–9 are handled procedurally by the Foundation, with no substantive involvement of the TCs or the dual channel.

In other words, the 4 TCs and the dual channel do not conflict; they respectively govern "internal content consistency" (the 4 TCs in steps 1–6) and "external global release" (the dual channel at step 7). This division of labor is the fundamental reason the immersion space's governance is "multi-party in interest yet stable in output."

3.2 How the Immersion Contribution Process Operates Under the New Governance Architecture

§3.1 established the two-dimensional framework of "4-TC content review + dual-channel SC escalation." This section uses that framework to walk through how immersion contributions actually operate: §3.2.1 first looks at the actual distribution of the 5 processes across the immersion space; §3.2.2 then looks at what each of the 4 TCs governs and which representative specs each has produced; §3.2.3 uses Immersion Requirements Rev 2.10—the landmark "jointly lead-authored across 4 TCs" spec—as a typical-path case; and finally §3.2.4 unpacks in detail the dual-channel trigger mechanism at the step-7 SC review.

3.2.1 The Actual Distribution of the 5 Contribution Processes Across the Immersion Space

The actual distribution of OCP's 5 contribution processes (Specifications 9 steps / Design Files 7 steps / Software 11 steps / Documents 9 steps / Product Recognition 2 steps) [S-190] across the immersion space is highly uneven—the vast majority of immersion contributions go through the Specifications 9-step process. The 11 immersion-related contributions from 2019–2026 (including the Charter) confirmed by the R3 research are classified by process as follows.

Table 3.4: Immersion Contributions Classified by Contribution Process

Key observations: ① of the 11 contributions, 9 are Specifications (82%), with only 1 Document (the Charter Rev 1.0 governance document); ② there are no Software-process contributions—the immersion space has no open-source software contributions, and every contribution is a spec document or reference design; ③ the Design Files 7-step process is also not used on its own (design files are usually submitted as attachments to the Specifications 9-step process); ④ Charter Rev 1.0 is not itself a Spec, but it defines the internal 4-TC review path for all subsequent Specs—it is a process-structural document rather than a content contribution.

This means that of the 5 contribution processes, the immersion space in practice uses only the single Specifications 9-step main process. Subsequent discussion in this chapter assumes by default that all immersion specs go through the 9-step process and no longer distinguishes between them.

3.2.2 The Workstreams Each of the 4 TCs Governs, and Their Representative Specs

Charter Rev 1.0 [S-212] consolidated the say inside the immersion sub-project into 4 Technical Committees by content theme. The R3 research [S-221] confirms each of the 4 TCs' chair / workstreams / representative specs as follows.

Table 3.5: Each of the 4 TCs' Chair + Workstreams + Representative Specs

Key observations: ① the Fluids TC + Reliability TC jointly review Material Compatibility—the phenomenon of a single spec spanning multiple TCs is extremely common in the immersion space, as detailed in the §3.2.3 Requirements Rev 2.10 case; ② the Solutions TC's 3 chairs/co-chairs all come from different business types (operator Denvr / power Murata / OEM Asperitas), making it the most camp-dispersed of the 4 TCs; ③ in OCP wiki public data the ITE TC has only a chair and no co-chair—a lone case of a "single-point chair" in the governance structure, a potential bus-factor risk; ④ the cross-Subproject Industry Liaison + Community workstreams are not folded into the 4 TCs, but Mullick's say in the immersion space via the Industry Liaison chair is independent of the 4 TCs.

3.2.3 The Full Path of a Typical Spec: Immersion Requirements Rev 2.10

Immersion Requirements Rev 2.10 (2023-08-18, sole Author = Rolf Brink / Promersion) [S-070][S-228] is the most typical "jointly lead-authored across 4 TCs" spec in the immersion space—it covers all 4 content themes: fluid + solution + reliability + ITE. Unfolding its 9-step process shows simultaneously how the content dimension (the 4 TCs) and the procedural dimension (the 9 steps) converge on one concrete spec.

Table 3.6: The Full Two-Dimensional Path of Immersion Requirements Rev 2.10

Source: Rev 2.10 author roster [S-070][S-228]; 4 TC chairs [S-221]; SC dual channel [S-206]; process steps [S-190]

Key observations: ① Rev 2.10's cross-4-TC content nature mobilized a joint SME review across all 4 TCs at steps 4–5—the largest cross-TC collaboration sample in the immersion space to date; ② the PSC collective approval at step 6 is the "highest internal gate"—the PSC consists of Brink + Gullbrand + the 4 TC chairs, 6 people in total, and all 6 must agree before it can be sent up; ③ the SC review at step 7 is the "highest external gate"—where the dual channel triggers (detailed in §3.2.4); ④ the 23 co-authors across 13 companies is the highest record of co-signing breadth in OCP immersion—and the reason the PSC's 6-person gate passes stably (the author companies already bring built-in buy-in).

3.2.4 The Dual-Channel Trigger Mechanism at the Step-7 SC Review

§3.1.3 already introduced the dual channel's fundamentals: Channel A = Mills (Cooling Env ICR), Channel B = Gullbrand (SC Co-Chair) [S-206]. This section places it back in the context of the Spec 9-step process to make the trigger timing and mechanism explicit.

The specific roles of the dual channel at step 7:

• Agenda-setting (Channel B / Gullbrand): the SC meets for 2 hours on the first Thursday of each month—the first hour a members-only private strategy session, the second hour an open review [S-211]. As SC Co-Chair, Gullbrand, together with Meta's Dharmesh Jani, decides whether the immersion spec is put on the agenda for that meeting—she is the gatekeeper of the agenda entry. She is also the Immersion sub-project co-lead and most familiar with the spec's internal status—so she directly carries out the "information handoff" of an immersion spec from PSC passage to SC review, with the highest efficiency.
• Voting (Channel A / Mills): SC voting is by majority across 14 seats (12 ICR + 2 Co-Chair) [S-206]. As Cooling Environments ICR, Mills votes at the SC on behalf of all of Cooling Environments (including the 6 Subprojects: Immersion + Cold Plate + Door HX + Heat Reuse + CDU + HMLC)—his vote is the collective will of Cooling Environments, so an immersion spec by default obtains the full endorsement of Cooling Environments.
• Coordination (where Channels A + B meet): there is no direct reporting relationship between Mills (Meta) and Gullbrand (Intel), but they sit in the same SC meeting room. The coordination from agenda to vote is the actual function of the dual channel—Gullbrand puts it on the agenda and Mills casts the vote, and the immersion spec passes the SC.

The counterproof that neither channel is dispensable: suppose Gullbrand does not add a given immersion spec to the SC agenda—the spec cannot be sent up for review; suppose Mills votes against at the SC (or a dispute arises among the 6 Subprojects within Cooling Environments)—the spec cannot pass. These two veto powers are simultaneously in the hands of two companies (Intel + Meta)—the "double lock" on the immersion spec's escalation path.

Implication for Ch4: no immersion-related acquisition affects this dual-channel structure—Intel will not pull Gullbrand out of the SC Co-Chair because an acquirer steps in; Meta will not pull Mills out of the Cooling Environments ICR because the immersion landscape shifts. The dual channel is the most stable anchor of immersion-spec order, but the acquirer also cannot transfer this escalation channel to itself through acquisition—this is the part of the immersion space's governance assets that "cannot be acquired."

3.3 Immersion Core-Spec Contribution Dossier

From the first Requirements in 2019 to the Power Distribution Base Spec released at the 2026-04 EMEA event, OCP immersion specs have accumulated 7 core specs + 5 incremental 2024-2026 contributions (Charter Rev 1.0 + Heat Sink + Power Distribution + 2 iAOPs). Unpacking each spec's complete author list, cross-company migration trajectory, engineering coverage, and version evolution in chronological order is the most direct evidentiary basis for the §3.5 individual↔company mapping and the §3.7 company ranking. This section unfolds by 7 core specs + 1 segment on 2024-2026 incremental developments.

3.3.1 Immersion Requirements v1.0 (2019-05-20, foundational)

Immersion Requirements v1.0 is OCP immersion's first formal spec—a foundational document that for the first time defined the OCP compliance framework for immersion cooling, established the three requirement types SR/OR/CR, and introduced the TCS/FWS operating-condition terminology, the single-phase vs. two-phase distinction, and the three system structures Enclosed Chassis / Open Bath / Hybrid. The author list is flat, with no tiered contributor/reviewer sections—a trace of OCP's not-yet-standardized template in 2019 (the standardized template only landed after the 2022 Global Summit, i.e., the 2022 OCP Global Summit process reform).

Table 3.17: Requirements v1.0 Author Roster (2019-05-20)

Source: [S-071]

The 5 authors represent the immersion ecosystem's initial configuration in 2019: Asperitas (OEM leader) + Intel (chip + thermal design) + Iceotope (chassis immersion) + Schneider Electric (facility-level electrical) + Flex (system integration). Three of them underwent cross-company migration during 2020–2024: Brink → Promersion (2021), Gore → Vertiv (2020), Bean → GRC (2021); Gullbrand and Payne stayed at their original companies.

3.3.2 Open Cassette Specification v1.0 (2020-02-27, an Asperitas solo contribution)

The Open Cassette Specification is an entirely solo Asperitas contribution—the PDF lists only a single author, Merijn Koster (Asperitas), with no co-author / contributor / reviewer / acknowledgement sections, and the license section is signed only "by: Asperitas." The spec defines the universal mechanical interface for the immersion chassis (cassette), with the goal of supporting reuse of the same hardware across multiple immersion solutions. It is an open-interface contribution for Asperitas's own product, the AIC24.

Table 3.18: Open Cassette Specification v1.0 Author

Source: [S-072]

Koster left the immersion space after 2020 (now COO of The Digital Club, "Creating freshly brewed apps"), but the Open Cassette spec remains on OCP under the Asperitas byline—Asperitas continues to derive a sense of ownership and brand asset from this spec, even though Koster is no longer inside Asperitas. This is the standard case of "the individual fully exits the space, the spec stays with the original company," in contrast with Brink's "100% travels with the person" in §3.5.

3.3.3 Design Guidelines for Immersion-Cooled IT Equipment Rev 1.01 (2020-12-07, 9 authors from multiple parties)

The Design Guidelines is one of the immersion space's rare "operational-level" specs—covering material compatibility (incl. the Soxhlet extraction test), single-phase/two-phase immersion thermal design, fluid-flow-rate differences, boiler-plate assembly, mechanical design (IT chassis dimensions, power supply, storage, high-speed optical-network cabling), and electronic design (signal integrity, temperature adjustability, fan control, IPMI, firmware immersion support). Its distinctive feature is that it explicitly separates the three sections Authors / Reviewers / Acknowledgments.

Table 3.19: Design Guidelines Rev 1.01 Author / Reviewer / Acknowledgement Roster

Source: [S-076]

This spec is the last collective byline in 2020 of the "old author team" of Brink + Gullbrand + Gore + Payne and others (by Rev 2, Asperitas had switched to Andy Young to succeed Brink; Schneider had fully withdrawn; 2CRSI no longer appears). Four of the 9 authors subsequently underwent cross-company migration (Shah 3M→TMGcore→MARA; Brink Asperitas→Promersion; Gore Iceotope→Vertiv; Bean Schneider→GRC). Dell (Curtis) and LiquidCool (Zien) made their only OCP immersion byline through the Reviewer role on this spec—and did not continue afterward.

3.3.4 Material Compatibility in Immersion Cooling v1.0 (2022-11-28, co-signed by 16 people across 13 companies)

Material Compatibility v1.0 is OCP immersion's first spec to explicitly be "multi-company co-signed," reflecting the landing of the "multi-party joint contribution" mode after the 2022 Global Summit template reform (i.e., the 2022 OCP Global Summit process reform). Its content covers the material–coolant compatibility matrix; single-phase vs. two-phase dielectric-fluid test methods; pass thresholds for 8 key physical properties—volume change, mass change, Shore hardness, fluid color, breakdown voltage, dielectric dissipation factor (DDF), acid number, material color (<10% acceptable, 10–20% case-by-case, >20% unacceptable); and the single-phase accelerated-aging test method.

Table 3.20: Material Compatibility v1.0 Contributor Roster (the PDF labels the column "Authors")

Source: [S-074][S-229]

Of the 13 companies, Shell with 3 authors (Shivaprasad / de Azevedo / Null) is the most from a single company; after integrating M&I Materials in 2024, the effective Shell-camp author count rises to 4 (including Sengupta). Chain Enterprises is Margerison's consulting vehicle, and his dual byline at TMGcore + Chain Enterprises reflects his retained ability to act independently—this is the root of why Modine's 2024-01 acquisition of TMGcore IP did not secure a commitment to his personal OCP influence.

3.3.5 Base Specification for Immersion Fluids v1.0 (2022-12-01, 11 Intel exclusive authors + 30 cross-camp acknowledgees)

The Fluid Base Spec is the "base language" of OCP immersion coolants—defining core measurement parameters such as dielectric strength (ASTM D 1816 / IEC 60156), flash point (D 92 / ISO 2592), fire point, autoignition point, pour point, heat capacity, thermal conductivity, density, volumetric expansion, and signal integrity (dielectric constant @ 20/40 GHz, @ 20/70°C; loss tangent). The 41-page spec is exclusively bylined by Intel's 11-person author team—the only single-company-led spec among OCP immersion specs.

Table 3.21: Fluid Base Spec — 11 Intel Authors + 30 Cross-Camp Acknowledgees

Source: [S-073][S-230]

The Fluid Base Spec's distinctiveness lies in the "Intel-exclusive authors + 30 cross-camp acknowledgees" mode—the most concentrated participation of hyperscalers (Microsoft 4 / Alibaba 2 / ByteDance 5) in an OCP immersion spec. Microsoft, participating through Mark Shaw (GM Advanced Hardware) + 3 engineers, carries the most weight; Alibaba and ByteDance, under China's "internal tech stack" path to immersion adoption, made a concentrated "international visibility" appearance through OCP acknowledgement, but did not continue after 2022. Amy Short appears in the Fluid Base Spec under the "Lubrizol" byline, and after 2024 moved to Denvr Dataworks as "Senior Officer - Immersion" and became the Solutions TC chair—a rare immersion-space trajectory of "an engineer's role migrating with the individual to a hyperscaler-class operator."

3.3.6 Immersion Requirements Rev 2.10 (2023-08-18, 23 authors + 11 acknowledgees)

Immersion Requirements Rev 2.10 is the current version—"Rev 2.10" is OCP's internal "two-point-ten" (first amended publication), equivalent to the wiki shorthand "v2.1" and the URL slug "rev-2-1" (see the OCP historical milestones + S-228). This version is solely bylined as Author by Promersion's Rolf Brink, plus 22 co-authors (covering Intel / Schneider Electric / Iceotope / Flex / TMGcore / Cargill / GRC / Asperitas / Vertiv / Shell / M&I Materials / Submer / UL, 13 companies in total) + 11 acknowledgees.

Table 3.22: Requirements Rev 2.10 — Full Author and Acknowledgee Roster

Source: [S-070][S-228]

Rev 2.10 covers 12 chapters of immersion compliance: terminology; Quality & Safety (UL/FCC/CE compliance + electrical safety + liquid management); Immersion Fluids minimum requirements (SR 4.x, single-phase limited to synthetic ester/natural ester/fluorinated fluid); and Feature classifications (the three-stack grading High Availability / High Safety / Thermal Optimized). It is the basis for OCP Logo Accepted / Inspired certification. The 23 co-authors across 13 companies is the highest record of co-signing breadth among OCP immersion specs—this spec's "high-density multi-party co-signing" mode is the most direct product of the post-2022-process-reform "multi-company joint contribution."

3.3.7 Warranty Guidelines for Immersion-Cooled Technology Components, Equipment and Systems v1.0 (2024-04-26, 2 Editors + 19 companies)

The Warranty Guidelines is OCP immersion compliance's standardization of the warranty boundary for equipment—for the first time providing a template for dividing warranty responsibility among the three parties of immersion coolant + hardware OEM + integrator, covering the three layers of component, equipment, and system. Its distinctive feature: the PDF lists only companies, not specific individuals (except the two Editors).

Table 3.23: Warranty Guidelines v1.0 — All Editors + 19 Contributing Members

Source: [S-075]

The 19 Contributing Members include facility-engineering parties (Airedale / DCX / Deerns) + operators (Telehouse / Denvr Dataworks) + fluid vendors (Engineered Fluids / ExxonMobil / FUCHS / Honeywell / Soltex / Modine) + OEM (Supermicro) + storage/memory (Samsung / Micron / Seagate via Cordle) + immersion originals (LiquidStack / LiquidCool Solutions / Submer / Promersion / Modine via TMGcore). This is the broadest value-chain coverage in a single collective byline among OCP immersion specs—but all 19 are listed only by company name, with no individuals, in sharp contrast with Rev 2.10's "23 named co-authors." The reason is that the Warranty is a commercial-responsibility spec, not a technical spec—the byline granularity at the commercial layer naturally lands at the company rather than the individual.

3.3.8 2024-2026 Incremental Contributions: the Charter + 4 New Specs

After the 2024-04 Warranty, the immersion space saw 5 new contributions within 18 months, reflecting the rapid output of the 4-TC architecture after the Charter Rev 1.0 restructuring.

Table 3.24: 2024-2026 Immersion Incremental Contributions

The shared significance of these 5 increments is the radiating expansion of immersion specs from "fluids/materials/reliability" toward the dimensions of "systems + power + optical networking + cleaning chemistry." The Power Distribution Base Spec, jointly produced for the first time by Murata (a power specialist) + Submer (a single-phase immersion OEM), is the Solutions TC's first major output after Charter Rev 1.0; Formerica's release of two iAOP V0.1 specs (LR4 + DR1) within 2 months is the representative case of the "specialized vertical rise" of immersion optical networking.

3.4 The Full Roster of the 4 TCs + 17 Workstream Chairs

After the 2024-12 Charter Rev 1.0 restructuring of the OCP Immersion Subproject, the governance structure consists of 4 Technical Committees (Fluids / Solutions / Reliability / ITE) + 17 workstreams. Each TC chair is appointed with PSC approval and reports to the PSC; workstream chairs are appointed by the corresponding TC and report to that TC. This two-tier "committee → workstream" nesting is the finest-grained evidence for the §3.6 company influence matrix. The table below unfolds in the order of the 4 TCs, plus the three cross-subproject items Industry Liaison + Community Outreach + Enablement.

Table 3.25: Immersion 4 TC + 17 Workstream Chairs — Full Roster (2026-05-28)

Source: [S-217][S-221]

This roster reveals several key facts. First, all 4 TC chairs are held by specific companies: BP chairs Fluids, Denvr Dataworks chairs Solutions, Intel chairs Reliability, and Unicom Engineering chairs ITE—the four-way interest structure lands directly on the chair seats. Second, the top 5 "hub companies" chairing multiple workstreams: ① Intel (4 seats: Reliability TC chair + Thermochemical Reliability + ITE-SI + the implicit Gullbrand SC Co-Chair); ② BP (3 seats: Fluids TC + Specifications + Lifecycle history); ③ FUCHS / CPC / ExxonMobil / Asperitas / Denvr / Murata / Shell / Unicom (2 seats each). Third, petrochemical majors hold 100% of the say over fluids: the chair seats of the 4 fluid-class workstreams under the Fluids Committee (Specs / Lifecycle / Cleaning / Oxidation) are 100% held by BP / CPC / ExxonMobil / Shell—a core structural constraint for the Ch4 acquisition discussion. Fourth, Intel's Yates holds two seats simultaneously, Reliability TC chair + Thermochemical Reliability WS chair, and combined with Gullbrand's SC Co-Chair + Immersion co-lead dual channel, Intel's "both reviews and writes" dual identity of say in OCP immersion is more pronounced than any other company's.

The top 8 "hub individuals" chairing multiple workstreams (two or more seats): ① Kris Fletcher (BP, 3 seats); ② Samantha Yates (Intel, 2 seats); ③ Spencer Kerns (CPC, 2 seats); ④ Amy Short (Denvr, 2 seats); ⑤ Andy Young (Asperitas, 2 seats); ⑥ Kevin Gero (Murata, 2 seats); ⑦ Brian Kinkade (FUCHS, 2 seats); ⑧ Austin Hipes (Unicom, 2 seats). These 8 "two-seat" individuals form the core circle of OCP immersion governance, and chair roles are bound more tightly to specific individuals' careers than to company affiliation—confirming OCP governance's "individual-centric concentration" trait: OCP governance's "individual-centric concentration" is far more pronounced than "company-centric concentration." Kelley Mullick's retention of the Industry Liaison chair across Intel → Avayla → Iceotope is the most direct evidence of this phenomenon.

3.5 Summary of the Key-Individual ↔ Company Mapping

§3.2 and §3.3 / §3.4 each show, from different dimensions, OCP immersion's "individual-centric concentration" trait—an organization's influence depends heavily on a handful of individuals, and these individuals underwent significant cross-company migration over the past 5 years. This section deals specifically with the precise "key-individual ↔ company" mapping, an intermediate step that must be completed before merging "individual-dimension influence" up into the "company dimension." The mapping is in four tiers: ① core hubs (≥3 OCP positions); ② multiple OCP roles within a single company; ③ a key byline on a single spec; ④ the governance apex (SC / Sub-project Lead class).

3.5.1 Tier 1: Core Hub Individuals (≥3 OCP positions)

Table 3.26: Core Hub Individuals + Cross-Company Migration Trajectory

Source: [S-070][S-073][S-074][S-076][S-218][S-219][S-222][S-228][S-229][S-230]

Among the 5 Tier 1 hub individuals, Brink's "100% travels with the person" mode is the strongest evidence of OCP immersion's "individual-centric concentration" trait—Asperitas still exists as a corporate entity, but the Requirements lead-author identity has moved entirely to Promersion, a 2-10-person consulting firm. Bean's "partly travels with the person + reinforced both ways" is another archetype—through Bean, GRC was upgraded in one stroke from a vendor to a Co-Chair (the governance layer), while Schneider, after acquiring Motivair in 2024, regained say in liquid cooling from the cold-plate side (see §2.4), with both companies reinforced. Gore's "mostly travels with the person" caused Iceotope's author roster in Rev 2.10 to shrink sharply. Shah's trajectory (3M → TMGcore → MARA) reflects the rise and fall of the two-phase immersion value chain—3M (the coolant source, exited PFAS in 2025) → TMGcore (a two-phase full-system OEM, asset-acquired in 2024) → MARA (a Bitcoin miner, using two-phase immersion in-house).

3.5.2 Tier 2: Multiple OCP Roles Within a Single Company

Table 3.27: Individuals with Multiple OCP Roles at a Single Company

Source: [S-070][S-073][S-074][S-079][S-217][S-221][S-228]

UL's Sakamoto + Zhou dual signing is a hallmark of the immersion-spec layer's "dual byline + irreplaceability of certification"—any immersion equipment entering OCP Accepted certification must pass through UL. Asperitas's Andy Young is the actual successor after Brink's departure, simultaneously holding seats across the three lines Solutions TC + TCO + SI, and is Asperitas's current carrier in the OCP immersion system. The trajectories of Roytman and Mullick both confirm "chair roles migrate with the individual"—Roytman brought the TCO Co-Chair from Intel to Microsoft, and Mullick retains the Industry Liaison chair across Intel → Avayla → Iceotope.

3.5.3 Tier 3: Key Byline on a Single Spec

Table 3.28: Individuals with a Key Byline on a Single Spec

Source: [S-072][S-075]

Through the single Warranty Editor role, Lappenbusch carries all of Supermicro's byline weight in OCP immersion—Supermicro has no byline in the author rosters of the other 6 core specs; Lappenbusch alone carries it. Cordle's Editor Emeritus role is a potential exit signal for Seagate in immersion specs; Koster has changed careers into application development, and there is no individual successor for Asperitas's Open Cassette design asset.

3.5.4 Tier 4: Governance Apex (SC / Sub-project Lead Class)

Table 3.29: Governance-Apex Individuals

Source: [S-206]

Jani and Mills together form Meta's "dual signature" at the apex of OCP immersion governance—Jani controls the agenda through the SC Co-Chair, and Mills controls the SC vote of the 6 Subprojects (including Immersion) through the Cooling Environments ICR. The two of them, plus Networking ICR Xu Wang, give Meta 3 of the 14 SC seats (21%)—the highest single-company share within OCP.

3.5.5 What the Cross-Company "Influence Flow" Means

A summary analysis of the "influence flow" of the 5 Tier 1 cross-company-migrating individuals reveals several patterns directly useful to the Ch4 acquisition discussion. First, in the immersion space "author identity" mainly travels with the individual rather than staying with the company—Brink took the Requirements lead-author identity entirely to Promersion; Bean took the Material Compat author and Co-Chair identities to GRC; Gore took the Rev 2.10 co-author and Design Guidelines author identities to Vertiv; Roytman took the TCO Co-Chair identity to Microsoft. Second, the sole counterexample is Koster (Asperitas)—he left the immersion space but the Open Cassette spec stays with the original company under its byline, reflecting that "reference-design"-class specs are attributed more to the company while "requirements/spec"-class specs are attributed more to the individual. Third, the influence increment of the original company essentially stops—Asperitas, Schneider Electric, and Iceotope all failed to replenish their author lineups after the cross-company migrations (Asperitas partly maintains it through Young, Schneider fully withdrew, and Iceotope retains only Kadhim). The implication for Ch4 is clear: acquiring a "recipient"-type company such as Promersion / GRC / Vertiv obtains the author assets accumulated through cross-company migration; acquiring an "original-owner"-type company such as Asperitas / Schneider / Iceotope requires separately assessing the remaining team's ability to carry the legacy forward.

3.6 The Company-Dimension × 4 TC Influence Matrix

Aggregating the §3.4 roster of the 4 TC + 17 workstream chairs + the §3.3 author rosters of the 7 specs + the §3.5 key-individual mapping by company entity yields an influence matrix of "company × 4 TC + governance seat + SC channel + Marketplace." This is the core operation of lifting the §3.5 "individual × role" data up to the "company × governance structure" level—and the direct scoring basis for the §3.7 final company ranking.

Column definitions (8 columns covering OCP immersion governance comprehensively): ① PSC / governance apex—Brink + Gullbrand co-lead + the Mills/Gullbrand dual channel + Board seats; ② Fluids TC—chair/co-chair + workstream chair + spec author/contributor within that TC; ③ Solutions TC—same; ④ Reliability TC—same; ⑤ ITE TC—same; ⑥ cross-Subproject WS—Industry Liaison / Community Outreach / Enablement; ⑦ SC channel—whether it reaches SC Channel A (Cooling Env ICR Mills) or B (SC Co-Chair Gullbrand); ⑧ Marketplace—product / service listing.

Ratings: S = strongest (sole position at the governance apex); A+ = strong (TC chair / co-chair / spec sole author); A = moderately strong (spec co-author / WS chair); B+ = moderate (contributor / acknowledgement); B = moderately weak (industry participation); C+ = weak (single-point byline); C = peripheral (historical byline); — = not involved.

Symbols: ★ = TC or WS chair; ◇ = spec author; ● = ack / contributor; ⚐ = Marketplace listing.

Table 3.30: 30-Company × 4 TC + Governance + SC Channel + Marketplace Influence Matrix

Source summary: [S-070]–[S-077] + [S-191] + [S-212] + [S-216] + [S-217] + [S-220] + [S-221] + [S-228]–[S-230]

Matrix legend: see the column definitions and symbol notes (at the start of this section).

Key observations: ① Intel is the only company covering the PSC governance apex (co-lead) + author/chair in 3 of the 4 TCs (Fluids ack + Solutions co-author + Reliability chair + ITE chair) + full SC Channel B—the broadest influence distribution among all 30; ② Promersion appears only at the PSC + a single point in the Solutions TC (Requirements sole Author)—highly concentrated influence but from a single source; ③ Asperitas runs on dual lines across the Solutions TC + ITE TC (Young is simultaneously Solutions TC co-chair + SI co-lead)—the broadest cross-TC presence in the OEM camp; ④ the 4 petrochemical majors Shell + BP + CPC + ExxonMobil collectively cover the Fluids TC + Reliability TC (material compatibility), and each appears only in these two TCs—precisely the matrix evidence for the Ch4 §4.7 conclusion that "say over fluids is 100% locked to the petrochemical majors"; ⑤ Meta appears only at the PSC and SC Channel A (Mills + Jani)—Meta does not write specs, only guards the channel, playing the "reviewer / gatekeeper" role in the governance structure.

3.7 Final Company Influence Ranking

Summing the §3.6 matrix by "coverage breadth across the 8 columns (PSC + 4 TC + cross-Sub-WS + SC channel + Marketplace) × the average weight of each column × Qualification Panel voting eligibility" yields the final company influence ranking. Scoring logic: S = 4 points, A+ = 3 points, A = 2 points, B+ = 1.5 points, B = 1 point, C+ = 0.5 points, C = 0 points; bonuses: ① a PSC / governance-apex seat (co-lead or SC dual channel) +2 points; ② a post-Charter Rev 1.0 4 TC chair / co-chair +2 points; ③ a Marketplace listing +1 point; ④ Qualification Panel voting eligibility (automatically granted to Requirements Rev 2.10 co-author companies) +1 point. The resulting top 28 are as follows.

Table 3.31: Final Influence Ranking of 28 Companies in the Immersion Ecosystem

Additional Tier C+/C companies (Schneider Electric / Flex / 3M / LiquidStack / LiquidCool Solutions / Neste / Honeywell / Koura Global / Alibaba / ByteDance / Dell / Telehouse / Modine / Chemours / Formerica / Rittal / Engineered Fluids, etc.) are included in the §3.6 matrix but do not enter the top 28—they have only a single-point byline or peripheral participation in the immersion space and do not constitute sustained influence entities.

Source summary: [S-070]–[S-077] + [S-191] + [S-212]–[S-230]

The core structure revealed by the ranking: Intel + Promersion + Asperitas + Shell + Meta, 5 companies, form the "absolute top" of OCP immersion influence—the top 5 each establish their advantage along one of five independent dimensions (governance apex, spec lead author, design files, implicit coolant winner, and SC dual signature), and these 5 are complementary and non-overlapping on the influence dimension. The second tier—UL + BP + Submer + GRC + Vertiv, 5 companies—covers the four concrete dimensions of certification, fluids, product-commercialization, and cross-company recipiency, and is the level genuinely reachable by the Ch4 acquisition discussion (Promersion / Asperitas / Submer / GRC are all within the independent-immersion-player range; see Ch4). From the third tier down it is essentially a combination of fluid-chemistry petrochemical majors (CPC / ExxonMobil / FUCHS / Cargill) + the operator camp (Denvr / Microsoft) + single-domain chairs (Lubrizol / Unicom / Murata / ENEOS / Oleon / Magnalytix) + cross-company recipiency vehicles (Avayla / Iceotope).

3.8 Five-Tier Company Layering

Re-classifying the §3.7 final ranking along the three axes "standards-dominance score × commercial-scale score × acquirability" yields a five-tier company layering. This is the view that feeds most directly into the Ch4 acquisition discussion—stating in one place "which companies are acquirable, which are not, and what OCP assets each can bring."

Table 3.32: Five-Tier Company Layering of the OCP Immersion Ecosystem

Source summary: [S-040]–[S-056] (M&A) + [S-070]–[S-077] (specs) + [S-212]–[S-230] (new R3 data)

The five-tier layering reveals the fundamental constraint of the Ch4 acquisition framework: within Tier S and Tier A+, only 4 genuinely reachable independent immersion players exist—Asperitas (Brink legacy + Young CTO), GRC (Bean), Submer (product + spec loop), Iceotope (chassis IP + historical talent)—plus 2 individual-centric entities: Promersion (acqui-hire) + Avayla (acqui-hire Mullick). All of Tier A's fluid-chemistry majors and the operator camp are unreachable, which is the structural root of Ch3's "why say over immersion coolants is 100% locked to unacquirable petrochemical majors." This layering aligns closely with the Ch4 §4.3 candidate-target pool (GRC / Iceotope / Asperitas + Promersion + Engineered Fluids + LiquidCool tuck-in), but §3.8 adds one more layer—the OCP influence dimension—namely, that the acquirer must look not only at company scale and valuation but also at the actual carry-over of OCP governance assets.

3.9 Influence Evolution Over Time: 2019–2026

Placing the key OCP events of the immersion space over the past 7 years on a timeline makes clear how influence flows between companies. This timeline has a clear implication for Ch4's "timing": after the 2024-12 Charter Rev 1.0, the immersion space entered a period of intensive "4 TC + 17 workstream" governance, and all Tier 1 companies are repositioning.

Table 3.33: 2019–2026 Immersion Ecosystem Influence Evolution

Source summary: [S-070]–[S-077] + [S-191] + [S-206] + [S-212]–[S-230]

The timeline reveals several core patterns. First, immersion specs evolved from "5 equal companies" to "13 co-signers + 19 collective signatories"—the 2019 Requirements v1.0 had only 5 authors signing as equals, the 2023 Rev 2.10 already had 23 authors across 13 companies, and the 2024 Warranty was a 19-company collective byline. The exponential expansion of co-signing breadth reflects the immersion space's migration from "niche experiment" to "mainstream awareness." Second, the three cross-company migrations (Bean 2021, Brink 2021-2024, Gore 2020) cluster in 2020-2024—the key period of "secondary concentration" of immersion-spec author identity, transferring the spec lead-author identity from the original companies to three new companies—Promersion / GRC / Vertiv—stopping the former's increment and igniting the latter's. Third, the two acquisitions—Shell's acquisition of M&I Materials in 2023-12 + Modine's acquisition of TMGcore IP in 2024-01—changed the actual attribution of the author roster: Shell implicitly doubled its influence through the acquisition, while Modine obtained IP through its acquisition but did not carry over personal OCP influence, revealing the two modes of success and failure in "accumulating OCP influence through acquisition." Fourth, the 2024-12 Charter Rev 1.0 is the largest single adjustment to the immersion space's governance structure—upgrading "scattered workstreams" into the two-tier nested architecture of "PSC + 4 TC + 17 workstream," marking immersion's maturation from an "experimental sub-project" into a "small mini-program." Fifth, the 2026-04 EMEA Power Distribution Base Spec is the Solutions TC's first major output after the Charter restructuring, foreshadowing the future all-round expansion of immersion specs from "fluids/materials" toward "systems/power/optical networking/cleaning chemistry."

3.10 Implications for the Ch4 Acquisition Framework

All of Ch3's analysis ultimately returns to one question: for a pure-immersion independent-player acquisition in the $100–200M valuation range in the immersion space, what unique due-diligence and pricing inputs can the OCP influence dimension provide? The conclusions of §3.1–§3.9 are distilled into five direct implications for Ch4.

Implication 1: the OCP asset value of acquirable independent immersion players is concentrated in 4 companies in Tier A+ + 1 in Tier S—Asperitas (Open Cassette + Young CTO + Solutions TC Co-Chair), GRC (Bean Co-Chair + Material Compat author), Submer (Power Distribution Base Spec + Marketplace ORv3 + multiple authors), and Iceotope (chassis IP + Mullick Industry Liaison chair) are the four platform-type acquisitions; plus 2 individual-centric entities only as acqui-hires—Promersion (Brink personally + the Requirements lead-author identity) + Avayla (Mullick personally). These 6 targets overlap heavily with the Ch4 §4.3 candidate pool, but Ch3 adds a new dimension of due-diligence points (see Implications 2–5).

Implication 2: cross-company migration decouples the "corporate entity" from the "OCP author identity"—Asperitas is still an OCP member with the Open Cassette spec legacy, but Brink has moved to Promersion; GRC, through Bean, inherits the network from the Schneider era; Vertiv, through Gore, inherits the immersion author identity from the Iceotope era. This means that in Ch4 due diligence one cannot look only at "the target company's OCP membership + spec byline record," but must also look at "whether the key individuals are still employed + whether they are willing to be retained." Both Brink and Bean are around 60 with no clear successor—a retention risk the acquirer must assess at the NBO stage.

Implication 3: the two acquisition precedents M&I Materials → Shell and TMGcore → Modine provide a reverse reference—the success path (Shell, by integrating M&I's engineering team, raised the immersion-coolant author count from 3 to 5, doubling its influence); the failure path (Modine obtained the TMGcore IP but did not carry over Shah / Margerison personally, and TMGcore's actual influence in OCP immersion went to zero). The acquirer must make explicit the "individuals transfer with the deal" carry-over provisions—especially for the three key individuals Brink (Promersion founder), Bean (GRC CTO), and Young (Asperitas CTO).

Implication 4: the unreachability of the immersion-coolant domain is a structural constraint—Tier A's 4 petrochemical majors BP / CPC / ExxonMobil / FUCHS collectively occupy all the chairs of the 4 workstreams under the Fluids Committee (100% of say over fluids), and combined with Tier A+'s Shell (which has integrated M&I) + the chemical collective (Chemours / Cargill / 3M exiting / Solvay / Lubrizol / KYZEN, etc.) they cover every coolant-related spec node. This means no Ch4 acquisition can obtain the fluid-spec definition right—immersion-coolant IP must be obtained through supply-chain collaboration rather than acquisition, and Engineered Fluids (no OCP author role but a Warranty Contributing Member) is one of the few tuck-in-able coolant targets (see the Ch4 §4.3 candidate dossiers).

Implication 5: the 4-TC architecture after the 2024-12 Charter Rev 1.0 makes the "workstream chair individual" a new anchor of acquisition value—Mullick (Avayla / Iceotope) alone holds the Industry Liaison chair + persists across 3 companies; Young (Asperitas) alone holds three lines, Solutions TC Co-Chair + TCO + SI; Bean (GRC) holds the Immersion Committee Co-Chair. Even when these individuals are not part of a company equity deal, the asset value of acqui-hiring them via their OCP governance positioning is on the order of $5–10M (per the §3.6 acqui-hire valuation anchor), with a marginal cost far below a platform-type acquisition. Therefore, Ch4's optimal path may not be a single platform acquisition, but a three-piece combination of "platform acquisition + key-individual acqui-hire + coolant tuck-in"—maximizing the capture of OCP governance + IP + team through a combined strategy while diversifying valuation risk. This is the most important unique contribution of the Ch3 OCP influence map to the Ch4 acquisition framework.

4. Acquisition Feasibility Study: Deep Dive on the Final 3 — Asperitas, Iceotope, GRC

Ch4 is this report's "decision chapter"—it brings Ch1's market map, Ch2's technology pathways, and Ch3's OCP immersion influence map and company ranking all to bear on one concrete M&A decision: in the immersion liquid cooling space, can a single Final platform acquisition be completed within the $100–200M valuation range via >50% controlling stake + consolidation onto a listed acquirer's financials? After R4 multi-track parallel deep research, the answer this chapter gives is a Final shortlist of three—Asperitas (NL) + Iceotope (UK) + GRC (US), each with a complete 12-part dossier decision package.

Important limitation disclosure: ① all three targets are private companies, and their valuations, cap tables, and ROFR / drag-along / right-of-first-refusal terms are all undisclosed, so all valuations in this chapter are tagged with "derived / range / comparison" basis and confidence; ② internal governance details such as key-individual shareholdings, retention terms, and board decision mechanisms require a second verification at the NBO + data-room stage; ③ 2026 H1 data show all three exhibiting an "unwillingness to sell" posture (see §4.2.5), and this chapter's timing judgment incorporates this dynamic.

4.1 Methodology Framework of the Acquisition Study

4.1.1 Acquisition Premise: >50% Controlling Stake + Consolidation onto a Listed Acquirer's Financials

This study's acquisition definition is locked to >50% controlling stake + consolidation onto a listed acquirer's financials—it does not consider the three paths of minority stake / acqui-hire / tuck-in as Final shortlist candidates. There are three reasons: ① a >50% controlling stake is the accounting precondition for consolidation onto a listed acquirer's financials (per the IFRS 10 / US GAAP ASC 810 control principle); ② a listed acquirer needs to reflect the target's full revenue and EBITDA in its own segment reporting to demonstrate strategic consistency to the capital markets; ③ the immersion space's current configuration is unstable (PFAS regulation + the AI inflection not yet arrived + giants buying up), so minority-stake interests are too exposed to the acquired party's future strategic choices.

Companies such as Promersion / Avayla / Engineered Fluids / LiquidCool / the DataQube-affiliated entity are not in the >50% platform-acquisition feasible pool because they are too small (estimated <$50M) or are merely consulting / IP-licensing entities—but they can be discussed separately as acqui-hire / tuck-in candidates after the Final platform acquisition is completed. Submer (Spain) is excluded—because of its 2024-10 completion of a $55.5M Series C + the $500M valuation rumor + its 2026-02 reverse acquisition of Radian Arc to pivot to full-stack [S-249]—a clear unwillingness-to-sell posture + exceeding the valuation ceiling.

4.1.2 Three Motives + 4 Hard Filters

The study's three motives inherit from the Q1 brief: ① financial return—control premium + subsequent exit / IPO; ② vertical integration—locking in cooling capability for self-built compute, with immersion as an option-style foothold (see §2.4); ③ technology / IP / team—patents, formulations, the engineering team, and OCP governance positioning.

The 4 hard filters constitute the candidate-pool screen: ① immersion-focused (not touching cold plate, inheriting the Q1 scope); ② valuation falling in the $100–200M range; ③ still independent—not acquired or asset-stripped; ④ excluding mega-scale platforms (including already-listed companies). Filter ④ removes the listed-company control-premium mechanism, compressing the candidate pool down to "private independent immersion players."

4.1.3 6-Dimension Scoring System

The scoring system is jointly constructed from the Q1 brief's three motives + the Ch3 OCP influence map, with 6 dimensions in total.

Table 4.1: The 6-Dimension Framework of the Ch4 Scoring System

4.1.4 Priority of the Three Execution Paths

Under the >50% controlling-stake definition, the study identifies three execution paths: ① friendly joint tender—bundling existing shareholders + founder cash-out + management earn-out, the lowest friction; ② majority control + existing shareholders retain a minority—the compromise path that defuses ROFR / drag-along; ③ 75%+ full acquisition—the strongest control but the highest premium. The applicable path differs for each of the Final 3; see the "feasibility of acquiring control" segment within each company's §4.3.4–§4.3.6 dossier.

4.2 M&A Environment and Comparable Transactions

4.2.1 Full Map of 2024-2026 Cooling M&A (incl. 5 new 2026 H1 deals)

The R4 research confirms a total of 12 major acquisitions + some PE / strategic investments in the 2024–2026 cooling space. The table below organizes the 7 known baseline deals + the 5 new 2026 H1 deals together.

Table 4.2: Full Map of 2024-2026 Cooling-Sector M&A

4.2.2 Control Premium + Listed-Company Consolidation Multiple: the Ecolab–CoolIT Anchor

In 2023-05 KKR Global Impact Fund II acquired a majority stake in CoolIT for $270M (Mubadala co-invested as a minority), and in 2026-03 resold it to Ecolab for $4.75B—total return 17.6× / equity return 15× / IRR 145-160% / 2.8-year hold—KKR Global Impact's first exit in North America [S-041].

This deal locked the cooling sector's valuation anchor at a historical high:

Table 4.3: Measured Valuation Multiples of 2024-2026 Cooling M&A

Data source: [S-040][S-041][S-042][S-044b][S-046b][S-054][S-243][S-244][S-245][S-246][S-247][S-248] + R4 sub-agent D financial estimates

The control-vs-full-ownership premium analysis shows that complete exits (100%) dominate the cooling sector—8 deals at 100% full ownership vs. only 1 staged deal (Motivair 75% + 25%). Buyers clearly prefer 100% acquisition to obtain the technology IP + integrate the supply chain, and the valuation gap between control and full ownership is not significant. The guidance this implies for Ch4: the Final 3 should prioritize a 100% full-ownership + management earn-out structure rather than a "75% first, 25% later" staged structure—unless an existing major shareholder (such as Asperitas's Shell Ventures or GRC's SK Lubricants) explicitly refuses to fully exit.

4.2.3 The Gap Between the Giants' Buying Range and the $100-200M Range

The valuation range of the 12 cooling acquisitions is $65M (Accelsius Series B) → $9.5B (Eaton-Boyd)—7 of them ≥$850M, giant-led and mainly cold plate / CDU / heat-rejection conglomerates. The $100-200M range is nearly empty—precisely the core opportunity window of the Ch4 mandate: pure-immersion independents have not yet been bought up by the giants, but only small and mid-sized independent vendors remain.

4.2.4 Two Contrasting Precedents of "Accumulating Influence Through Acquisition"

The R4 research confirms the two contrasting precedents described in Ch3 §3.10:

• The success path—Shell + M&I Materials (2023-12-31): Shell, by integrating M&I's MIVOLT team (Eleanor Jones + Sayan Sengupta), raised the author count of immersion Requirements Rev 2.10 from 3 to 5—doubling its influence. This way of "accumulating OCP influence through acquisition" cannot be replicated by the acquirer.
• The failure path—Modine + TMGcore IP (2024-01, $12M): Modine obtained the IP/assets but did not carry over Shah / Margerison personally (Shah went to MARA; Margerison still carries a Chain Enterprises dual byline) → TMGcore's actual influence in OCP immersion went to zero.

Implication for Ch4: the buyer must make explicit the "individuals transfer with the deal" carry-over provisions—especially for the four key individuals Brink (not inside Asperitas, but with a deep Shell fluid relationship), Bean (GRC CTO), Young (Asperitas CTO), and Mullick (Iceotope VP).

4.2.5 Timing Variable: the 2026 H1 Collective "Unwillingness-to-Sell" Posture of Independent Immersion Players—a Key Finding

The R4 data reveal a finding that overturns the Ch4 timing—in 2026 H1, all independent immersion players showed "reverse moves / strategic financing / commercial-collaboration" signals, clearly refusing to be acquired at a low valuation:

Table 4.4: 2026 H1 Final 3 + Peer Strategic Moves

Key implications for Ch4 timing: ① the 2026 H1 immersion-M&A window is already closed—none of the Final 3 is willing to be sold at the current valuation; ② genuinely feasible timing is most likely pushed to 2026 H2 – 2027, requiring an external catalyst (the start of ABC Impact's exit window / SK Enmove's strategic pivot / completion of the Invest-NL mandate, etc.); ③ PE entry (Blackstone Energy Transition acquiring a majority stake in ACT in 2026-03 [S-245] / Carrier Ventures' 2026-04 follow-on in ZutaCore [S-248]) suggests Advance Studio may face professional PE bidding rivals in 2026 H2.

Practical acquisition recommendation: first make a partnership / strategic-minority investment to establish a signal, and launch the NBO in 2026 H2 – 2027 depending on the external catalyst.

4.3 Candidate-Pool Screening and the Final 3 Deep Dive

This section converges the candidate pool in three steps—"full-market profile → 5 categories of exclusion → Final 3 shared selection criteria"—then gives the complete 12-part dossier for each of Asperitas, Iceotope, and GRC.

4.3.1 Full-Market Candidate Profile Snapshot

The R4 research covers all 12 immersion-related candidate entities across the market—laid out along three axes: business model + valuation magnitude + independence.

Table 4.5: Full-Market Candidate Profile of the Immersion Space (as of 2026-05-28)

Source: [S-040][S-041][S-046b][S-120]–[S-129][S-231]–[S-242][S-249] + R4 sub-agent A/B/C financial estimates

4.3.2 The 5 Categories of Exclusion

The 12 candidates are excluded in 5 categories:

① Over valuation (3 companies): Submer ($500M rumored + having become a consolidator itself) / Firmus ($5.5B, already excluded) / DUG Technology (ASX-listed + not pure liquid cooling).

② Insufficient scale (4 companies): Promersion (a 2-10-person consulting entity) / Engineered Fluids (no institutional funding, tuck-in) / LiquidCool Solutions (~$5.3M funding + the equity burden after the DataQube liquidation) / Avayla (a small company led by Mullick personally)—none constitutes a >50% platform-acquisition feasible pool, but they are retained as acqui-hire / tuck-in candidates after the Final platform is completed.

③ Already acquired (4 companies): LiquidStack → Trane (closed 2026-03-03) / CoolIT → Ecolab (announced 2026-03-19, closing 2026 Q3) / CoolTera + PurgeRite → Vertiv (closed) / Motivair → Schneider (75% closed 2025-02).

④ Already asset-acquired (1 company): TMGcore IP was taken by Modine for $12M in FY2024 and has exited as a legal entity [S-121].

⑤ Insufficient data (pending further screening): Midas Immersion Cooling (US) / DCX (PL, Tier 4 data centers) / Immersion4 (CH, Switzerland) / Synano (a nano-oil company partnering with GRC, newly appearing at EMEA 2026) / KYZEN (cleaning chemistry, newly entering OCP)—these are second-round screening candidates after the 2026 H2 – 2027 timing launch.

4.3.3 The Final 3's Shared Selection Criteria and Complementarity

After the 5-category exclusion screen, the remaining 3 form the Final shortlist: Asperitas (NL) + Iceotope (UK) + GRC (US). The three share 5 commonalities + 3 strategic complementarities.

5 commonalities: ① all are single-phase immersion (avoiding PFAS regulatory risk; see §2.5.2 + S-254 on the ECHA 13.5-year buffer); ② all operate independently + are not absorbed by giants; ③ all are in Tier A+ or Tier A of the Ch3 §3.7 final company influence ranking; ④ all already have strategic investors but have not lost control; ⑤ all have built deep collaboration with at least one chemical company or operator (Asperitas-Shell / Iceotope-Dell-HPE-Meta / GRC-Samsung-LG-SK).

3 strategic complementarities:

Complementarity 1: geographic coverage spread across NL + UK + US—forming full coverage of the three major markets Europe (Asperitas) + the UK (Iceotope) + North America (GRC), with no geographic overlap.

Complementarity 2: full technology-form spectrum—Asperitas leads with Open Cassette + the direct-forced-convection DFCX1 (an AI/HPC pivot); Iceotope leads with precision/chassis single-phase (1U/2U standard rack); GRC leads with the open-bath tank (ICEraQ + the modular ICEtank). The three cover the three major single-phase immersion forms "cassette / chassis / tank," with no product overlap.

Complementarity 3: division of labor in Ch3 OCP influence positioning—Asperitas leads on spec positioning (Open Cassette sole author + Solutions TC Co-Chair via Young); Iceotope leads on IP density (219 patents + Mullick Industry Liaison chair); GRC leads on governance participation (Bean Immersion Committee Co-Chair + FMEA WS chair). The three cover the three types of OCP immersion influence anchors "the standards pen + the patent wall + the governance seat," with no role overlap.

The next three sub-sections unfold the dossiers in the order Asperitas → Iceotope → GRC (the order matches the original brief; the §4.4 comprehensive recommendation will re-rank priority by the real data).

4.3.4 Asperitas (NL, the top pick for spec positioning)

Business overview —— Asperitas's legal entity is Aecorsis BV (KVK 60596376, Amsterdam Laarderhoogtweg 18; any SPA must target this entity rather than the trade name Asperitas) [S-232]. Founded in 2014 by Rolf Brink + Markus Mandemaker; in 2017 it made the global commercial debut of AIC24 (a 24U natural-convection oil-immersion cooling cabinet, ~30-32 kW/cabinet, pump-free, CE/UKCA); on 2024-06-26 it launched DFCX1 (Direct Forced Convection, 12U, 44-60 kW/cabinet, >1000 W/chip, targeting AI/HPC) [S-234]. Its fluid partner is Shell (Shell S5 X, a single-phase GTL synthetic hydrocarbon, co-developed with Asperitas). Eurofiber is DFCX1's first 12U customer. Headcount is 22-33 employees (Crunchbase 2026-04 reports 33; Growjo 2024-07 reports 22; an actual count is ~25-35 people).

Funding history and ownership structure —— Cumulative funding is severely under-reported: the Crunchbase $2.15M figure is distorted; the true cumulative is €5-15M (incl. the 2019 Seed + the 2023 bridge + the 2025-09 scale-up round) + a €34M IPCEI grant (non-equity).

Table 4.6: Asperitas Complete Funding History

Current cap-table inference (no public data, pure order-of-magnitude estimate): Brink + Mandemaker founders 15-30%; Shell Ventures 8-15%; PDENH (managed by ROM InWest) 5-10%; Invest-NL 5-12%; STECON Group / STECX Ventures 15-25% (2025-09 lead); Andy Young + the management option pool 5-12%; other employees + miscellaneous 3-8%. Core ROFR / governance risk: the stacked consent rights of the three parties Shell + Invest-NL + STECON = a >50% controlling stake cannot bypass any one of them.

Technology IP and product portfolio —— The core patent family US10716238B2 (granted 2020-07-14, inventor Dirk Roelof Brink, priority 2015-11-23, expires 2036-11-21) + EP3380907B1 + CN108351675B + NL2015841B1 + SA518391580B1, all belonging to Aecorsis BV, not transferred to Promersion. Aggregator counts for comparison: Crunchbase/IPqwery 7 / PatSnap 16 / actually verifiable 1 core patent family (5-6 grants across jurisdictions). OCP contributions: Open Cassette Spec v1.0 (2020-02-27, Koster sole author) + Fan Sim Spec (Brink) + Requirements v1.0 / Rev 2.10 + Solutions TC Co-Chair (Young) + TCO co-lead (Young + MSFT Roytman) + ITE-SI co-lead (Young). Promersion IP carve-out: no evidence of a formal carve-out; the core patents still belong to Aecorsis BV. Key certification gaps: ✗ NVIDIA Certified System missing (only an NPN partner, no H100/B200 certified); ✗ AMD formal certification missing; ✓ Intel/Dell PowerEdge "Ready Solutions" via UNICOM Engineering; ✓ Cisco AI-Ready Immersion-Cooled DC Solution Brief.

OCP influence positioning (inherits Ch3 §3.5) —— Asperitas ranks #3 (27 points) in the Ch3 §3.7 final ranking, tied with Shell. The key current contributor = Andy Young (CTO) alone: he holds three lines, Solutions TC Co-Chair + Solutions-TCO co-lead + ITE-SI co-lead; plus Rev 2.10 co-author + Fluid Base Spec ack. Brink has already become independent through Promersion (founded 2021, fully exited the Asperitas advisor role in 2023-04), and his Cooling Environments Immersion sub-project Lead and Requirements Rev 2.10 sole Author identities cannot transfer with an Asperitas acquisition—the biggest OCP-influence loss item in an Asperitas acquisition.

Key-individual retention framework ——

• Andy Young (CTO) —— the most critical retention: Leeds University (UK) BEng/PhD Mech Eng; former Iceotope Head of Fluids → Ansys → Asperitas R&D Director (2020) → CTO (2021-03); co-lead of multiple OCP seats; not a founder, equity probably 5-10%; highly portable (a direct competitor of Iceotope). Retention must include a 4-5-year earn-out + cash bonus ≥ €1M.
• Rolf Brink (founder / now shareholder only) —— not bound: handed over the CEO role in 2022-05 (to Ronald Monster), fully exited the advisor + non-exec role in 2023-04 [S-233]; Promersion is his next venture; estimated age ~55 (the baseline 65 was a misjudgment); the acquirer should arrange a Brink shareholding buyout separately and preserve the OCP relationship via a Promersion advisory contract.
• Markus Mandemaker (co-founder) —— low visibility: rare public appearances; TheOrg does not list him on the exec team; he may have faded out.
• Rutger de Haij (current CEO, Interim since 2023-01) —— TU Delft Civil Eng; former Shell Commercial Manager Renewable Heat → EscherCloud MD → Asperitas Non-Exec Director (2021-10) → Interim CEO; the shadow of a manager seconded by Shell Ventures.
• Ronald Monster (former CEO 2022-05 to ~2023-01): departed, no need to consider.
• R&D team: based on the single patent inventor Brink + the single-point OCP representative Young → the core team is inferred to be ≤10 people, with high concentration and large bus-factor risk.

Valuation basis and range judgment —— Asperitas's valuation most likely falls in the $40-90M range, with a midpoint of about $60M—❌ not in the $100-200M range (that is Submer's magnitude). Multi-path estimation:

Actual acquisition pricing: considering the strategic premium (OCP positioning + Andy Young + the Shell S5 X relationship + the follow-on of the IPCEI 5-year grant) + control premium → $70-120M is the reasonable bid range. It lands perfectly at the "lower bound of $100-200M" set in Ch3.

Feasibility of a >50% controlling stake + resolving existing shareholders' interests —— conditionally feasible, 60-70% probability of passage. The core obstacle has a three-layer structure:

Layer 1—Shell Ventures (the hardest): Shell is Asperitas's strategic anchor (the sole supplier of S5 X fluid + Rutger de Haij is a manager groomed by Shell); Shell is itself an upstream strategic player for Asperitas (including the MIDEL business from integrating M&I Materials in 2024). Shell has a high probability of opposing a change of control, because the acquirer could sever Shell's fluid exclusivity. Resolution: ① a direct dialogue with Shell Ventures in advance; ② the acquisition agreement includes a 5-10-year Shell S5 X preferred / exclusive procurement commitment; ③ Shell Ventures can continue to hold as a minority or cash out (its 2019 ticket is small, so cash-out is financially feasible).

Layer 2—Invest-NL (moderately hard): the Dutch SWF mandate = advancing the Dutch sustainable economy + the deep-tech ecosystem; charter consent rights + the Vifo Act + the political climate will slow things significantly. The shadow of the Solvinity case in 2026-05: the BTI for the first time fully prohibited a €100M+ cross-border critical-infrastructure acquisition [S-250]—but Asperitas is not critical infrastructure (not a DC operator). Resolution: ① Invest-NL most likely is willing to cash out (the mandate is fulfilled—the startup has been pushed to the scale-up stage); ② accept a conditional approval including R&D-stays-in-NL + employment commitments + IP-no-outflow terms; ③ commit that Aecorsis BV retains its Dutch HQ.

Layer 3—STECON Group (structurally hard): the 2025-09 lead investor + an SE Asia MoU; invested <1 year ago, not yet past lockup expiration; very likely has ROFR + drag-along + tag-along. Resolution: ① STECON cashes out (~1.3× return in the short term); ② STECON becomes the SE Asia regional distribution partner, retaining an independent SE Asia commercial arrangement; ③ a long-term earn-out with an SE Asia revenue-target clause.

Geopolitical / regulatory overlay (NL) —— the Netherlands' Vifo Act passage rate is about 81% (67% unconditional + 14% conditional + 1.4% prohibition); BTI 2024 case data: 69 filings / 70 cases concluded [S-250]. Data-center cooling is not within the Vifo Art. 7 "vital provider" enumeration (vital includes only heat transport / nuclear / Schiphol / Rotterdam port / banking / FMI / Groningen gas / gas storage); Asperitas is a cooling supplier rather than a DC operator, so it is by default outside the vital scope. The Sensitive Technology Decree 2023-172 covers quantum / photonics / semiconductor / high-assurance infosec, and immersion cooling is by default not included, though case-by-case is possible. The lesson of the Kyndryl-Solvinity case: even when the buyer is a U.S. entity, critical infrastructure can still be fully prohibited—but Asperitas is not in this minefield (an industrial-cooling OEM rather than a government IT contractor). The BTI review duration is expected to be 4-6 months; a pre-filing informal consultation is recommended. SER Merger Code + Works Council: Asperitas has 22-33 employees < 50 → no mandatory union notification / statutory Works Council consultation.

Due-diligence checklist for a >50% controlling stake + listed-company consolidation ——

• Financial DD: Aecorsis BV's 3-year audited accounts (KVK public scope is limited); revenue / EBITDA / cash burn; the IPCEI grant's booking timing; the IPCEI project's milestone-attainment status.
• Legal DD: the five-party shareholder agreement among Shell Ventures / Invest-NL / PDENH / STECON / TK & Partners; ROFR / drag-along / tag-along terms; the Shell S5 X fluid-supply exclusivity terms; the IPCEI 5-year grant-recovery terms.
• Technical DD: US10716238B2 + the international family freedom-to-operate; verification of the Promersion IP carve-out (confirming there is no carve-out); AIC24 + DFCX1 product-certification continuity; the NVIDIA / AMD certification gap-filling path.
• OCP governance DD: Andy Young retention terms; the Open Cassette spec maintenance commitment; continuity of the Solutions TC seat; the impact of the Inspired 2026-08-01 retirement on Asperitas's SP/Accepted eligibility.
• Listed-company consolidation accounting: PPA (Aecorsis BV → parent, functional currency EUR recommended); Goodwill $40-85M impairment risk; conditions for retaining the Dutch innovation box (5% corporate income tax); IFRS / Dutch GAAP difference adjustments (R&D capitalization + the IPCEI grant booking timing).

Timing and execution cadence —— Recommended cadence: Month 0-1 a private approach to Brink to confirm his willingness to exit and a Promersion coexistence arrangement; Month 1-2 a strategic dialogue with Shell Ventures (fluid-supply commitment); Month 2-3 communication with Invest-NL + PDENH on minority-continue or cash-out + a BTI informal consultation; Month 3-5 a two-way negotiation with STECON; Month 4-6 LOI + DD; Month 6-8 SPA signing + the formal Vifo filing; Month 8-12 the BTI review (4-6 months expected); Month 12 closing + locking in Young's 4-year earn-out; Month 12+ accelerating NVIDIA / AMD certification + a Shell 10-year fluid contract + landing the SE Asia partner. Key dates: 2026-08-01 Inspired retirement (Accepted-eligibility continuity is a key check item); 2026-10 OCP Global Summit San Jose (Asperitas already spoke at EMEA 2026).

Accounting implications of listed-company consolidation —— PPA: assuming a 100% acquisition @ $70-120M, identifiable IP (the patent family + Open Cassette + Fan Sim) ~$5-10M; customer relationships (Shell HPC + Eurofiber + Cyxtera COE) ~$3-8M; brand ~$2-5M; fixed assets + WC ~$5-10M; deferred tax assets (the IPCEI grant + the Dutch innovation box) ~$3-5M; total identifiable net assets ~$20-35M. Goodwill = purchase price − identifiable net assets = $40-85M (typically 60-70% of the purchase price is goodwill). The five goodwill-impairment triggers: ① NVIDIA-certification lag (the biggest risk; failing to obtain H100/B200 certified within 12 months → missing the AI inflection window); ② an interruption in Shell S5 X fluid supply; ③ Brink continuing to independently lead OCP via Promersion / Andy Young departing; ④ failure of the MISD IPCEI project's delivery milestones; ⑤ a price war among Submer/Iceotope/LiquidStack/GRC.

Overall judgment —— conditionally proceed:

• ✓ clean technology IP (the core patents are in Aecorsis BV, not carved out to Promersion)
• ✓ excellent OCP spec positioning (Open Cassette + Solutions TC Co-Chair)
• ✓ valuation most likely reasonable ($40-90M base, the $70-120M strategic premium is acceptable)
• ✓ high Vifo Act passage rate (81% unconditional + conditional)
• ⚠️ the stacked consent rights of the three shareholders are the core gating
• ⚠️ the missing NVIDIA Certified is the biggest product risk
• ⚠️ Brink's independence via Promersion = OCP influence cannot be transferred with the deal
• ✗ Andy Young is a single-point key talent; retention must be locked in up front

Top 5 key risk array: ① long-term absence of NVIDIA H100/B200 certified (high / very high; budget €2-5M / 12 months); ② Andy Young not staying (medium / very high; a 4-year earn-out); ③ Shell Ventures opposing (medium / high; advance communication + a 10-year Shell S5 X commitment); ④ a STECON SE Asia MoU conflict (medium / medium; partner conversion); ⑤ the BTI triggering a conditional approval with an R&D-stays-in-NL clause (medium / medium; pre-filing consultation).

4.3.5 Iceotope (UK, the IP flagship, a valuation-ceiling-risk candidate)

Business overview —— Iceotope Technologies is headquartered in Sheffield, UK, specializing in precision/chassis immersion (vs. Asperitas open-cassette / vs. GRC open-bath tank): a single server is itself an independent sealed chassis, each with its own dielectric-fluid loop; the 1U/2U/3U form factor is compatible with standard racks and does not require retrofitting the data-center floor / load-bearing; single-phase dielectric fluid (non-fluorinated); each chassis is hot-swappable. The group parent Iceotope Group Limited (Companies House #08783357) holds 75%+ equity + voting power in Iceotope Technologies Limited (#08004482) (the baseline #08099660 is a dissolved clothing company—incorrect) [S-237]. Registered address Amp Technology Centre, Sheffield, S60 5WG. SIC Codes 26200 (computer manufacturing) + 62090 (IT services). The most recent audited accounts are through 2025-06-30.

Funding history and ownership structure —— Cumulative funding $72-104M (Tracxn reports $81.4M over 10 rounds):

Table 4.7: Iceotope Complete Funding History

Key fact: when the 2022-06 ABC Impact round closed, all 25%+ single-shareholding relationships were dissolved—the group parent currently has No active PSC (since 2022-06-28, no single entity holds >25%) [S-237]—the target with the most dispersed cap table among the Final 3.

Current cap-table inference (estimated from combined public sources): ABC Impact 18-25%; Edinv 10-15%; Northern Gritstone 6-10%; British Patient Capital 4-7%; Pavilion Capital 4-7%; Two Seas Capital 4-8%; Barclays Climate Ventures 2-4%; Aster Capital (Series A holdover) 2-5%; Ombu Group (legacy) 0-5%; founders + ESOP (Hopton + management) 18-28%.

Technology IP and product portfolio —— 219 patents (the 2026-05 milestone update; the baseline 200 was too low) [S-235]; distribution: mechanical / chassis design 35% + fluid technology 25% + server integration / interfaces 20% + control systems 10% + application methods 10%. The KUL product line: KUL Box (a 1U/2U edge unit, integrated with the HPE ProLiant DL110); KUL 2 (a 2U precision chassis, integrated with the Gigabyte G293); KUL Extreme (desert / 5G edge, jointly launched with HPE+Intel+nVent at MWC 2023); KUL AI (high-density AI accelerator, 1MW+/rack spec); the Avnet KUL chassis (an OEM-channel commercial product + EcoStruxure BMS). Strategic OEM/Partner Integrations: Dell "Key Collaborator" (specific product undisclosed); HPE ProLiant DL110 + DL380 Gen10 already ship in KUL chassis (production-grade integration); Meta (2022-12 joint research on HDD storage immersion, R&D grade); Intel (2022 joint OCP immersion-warranty announcement); nVent (an OEM-level distribution agreement); Schneider Electric (EcoStruxure BMS + Aster Capital Series A strategic sponsor); Avnet (channel commercialization since 2021); NREL HPC reference; Gigabyte (KUL 2 OEM adaptation). Key certification gaps: ✗ NVIDIA Certified System undisclosed; ✗ AMD formal certification has not reached the Vertiv/Schneider level; the AI Tray is a universal accelerator tray, theoretically compatible with GB200/H100/H200/MI300/MI325, but NVIDIA OEM certification is undisclosed.

OCP influence positioning (inherits Ch3 §3.5) —— Iceotope ranks #20 (9 points) in the Ch3 §3.7 final ranking—relatively far back, but it has a special hub individual. Kelley Mullick (VP Technology Advancement & Alliances, joined 2023-05) holds the OCP Industry Liaison WS Chair + was a 2024 OCP board candidate; she also still serves as Avayla CEO (her own consulting/design house)—a dual identity that leverages OCP say but is also a potential conflict. Mustafa Kadhim (materials-technology lead) holds an OCP Material Compat ack. Iceotope as an organization has already exited the OCP Immersion Cooling spec author roster (since Rev 2.10, only Kadhim's acknowledgement)—after Nigel Gore moved to Vertiv in 2020, Iceotope's presence in the OCP author roster shrank sharply. Post-acquisition recommendation: consider acquiring Avayla (a small consulting firm, typically $5-10M) to fully lock in Mullick.

Key-individual retention framework —— the most critical retention signal is the 2025-01 executive change:

• Jonathan Ballon (current CEO + Chair, took office 2025-01-15) —— a textbook exit operator: former SkyHive President + CEO (acquired by Cornerstone in 2024, he ran the exit); former Intel AI/Edge VP; former GE / Cisco; currently non-executive chair of Axelera AI + Aspinity; SF based [S-236]. He was brought in to "do an exit," naturally aligned with a buyer.
• Simon Jesenko (CFO, joined 2023-06) —— the second key exit operator: former Senseye CFO (acquired by Siemens, he ran the acquisition + integration); 15+ years SVN/AUS/UK.
• Nathan Blom (Co-CEO + CCO 2024-09 to 2025-01 → CCO after 2025-01) —— former Lenovo NA ISG strategic transformation, 7+ years.
• Kelley Mullick, PhD (VP Tech Advancement & Alliances) —— former Intel data-center GTM; Chem/Mat E PhD, Ohio State; also still serves as Avayla CEO.
• Neil Edmunds (Chief Innovation Officer, promoted 2026-03) —— an internal Iceotope technical backbone, a long-term employee.
• Mustafa Kadhim (materials-technology lead) —— high retention—one of the core inventors of the 219 patents.
• Past / Already departed: Peter Hopton, founder (2005 VeryPC → 2011 Iceotope IP restructuring), currently not on the public C-suite list; David Craig, former CEO, retired 2024-09-30 (no longer CEO; the baseline "David Craig is current CEO" is outdated) [S-237]; Nigel Gore 2015-2019 Head of Product Management → 2020 Vertiv.

Retention-framework recommendation: Tier 1 (Ballon + Jesenko) 24-36 months + performance vest (they are the exit designers and should be tied to the deal close); Tier 2 (Blom + Mullick + Edmunds + Kadhim) 24 months + retention bonus; Tier 3 (the engineering team ~50) 12 months + ESOP refresh.

Valuation basis and range judgment —— Iceotope's valuation is most likely $150-260M, with a 55% probability of exceeding the $200M ceiling—the target with the greatest upside valuation risk among the Final 3. Multi-path estimation:

Probability assessment of the valuation "exceeding $200M": <$150M probability 15% / $150-200M probability 30% / $200-300M probability 40% / >$300M probability 15%—a net post-2026-05 probability of about 55% that the valuation is greater than $200M.

Feasibility of a >50% controlling stake + resolving existing shareholders' interests —— complex but conditionally feasible. Three budget tiers:

Tier A recommended: enter at $120M (50%+1 voting control); retain all current institutional investors until an IPO/secondary. Legal / regulatory path: the UK Companies Act 2006 has no takeover code (not a plc), so a PE-style private negotiation suffices; no public ROFR / pre-emption rights; after Series B-2, anti-dilution and tag-along are expected to have limited impact on a majority buyer.

Geopolitical / regulatory overlay (UK) —— under the UK NSI Act 2021, Computing Hardware is one of 17 mandatory sectors. On 2026-03-12 the UK Gov published its NSI Act amendment response—proposing to merge Computing Hardware into the Semiconductors category—a change that may lighten Iceotope's mandatory-notification risk (cooling equipment is clearly not a semiconductor); but voluntary notification + call-in power can still be used by the government. Iceotope's particular risk source: ABC Impact (Temasek ecosystem + ADB + SEA SWF affiliated), the exit of foreign state-affiliated capital, carries a medium-to-high call-in risk; Pavilion Capital (a Singapore GIC affiliate) is equivalent; Singapore is non-Five Eyes, non-EU—it enjoys no review exemption. 2024-25 NSI data: 1,143 notifications / 17 final orders / 1 unwind; Chinese parties at <5% of notifications account for 41% of call-ins [S-252]. Recommended path: a voluntary notification to the UK Cabinet Office Investment Security Unit (ISU), preparing a 60-90-day review window—even if mandatory notification is not triggered, the UK government's call-in risk for an overseas buyer when Temasek-affiliated capital exits sensitive technology is medium-to-high.

Due-diligence checklist for a >50% controlling stake + listed-company consolidation ——

• Financial DD: Iceotope Group Limited's complete 2024-2025 audited accounts (Companies House paid PDF download £3 / filing, P0 priority); revenue (Tracxn reports $3.45M ARR, to be verified); EBITDA / cash burn; the sentiment of the total raised over 5 years ($104M in vs. commercialization progress).
• Legal DD: the complete cap table + shareholding percentages (under NDA → request directly from ABC Impact / the CFO, P0); the ABC Impact and Pavilion exit agreements / tag-along terms (P0); Avayla acquisition willingness + valuation (bundled when negotiating the Mullick retention package).
• Technical DD: the specific Dell PowerEdge integration product codenames / OEM agreements; the current state of NVIDIA OEM certification (GB200/H100 chassis-level); the 219-patent freedom-to-operate.
• OCP governance DD: Mullick retention + handling of the Avayla dual-identity conflict; continuity of the Industry Liaison chair; the impact of the Inspired 2026-08-01 retirement (Iceotope has already exited the OCP author roster, relying mainly on the Mullick chair).
• The final classification of cooling equipment under the 2026 NSI Act reform: early informal engagement with the UK ISU.
• Listed-company consolidation accounting: PPA (Iceotope Group Limited → parent); the five goodwill-impairment triggers (Mullick departing / NVIDIA certification / the undisclosed depth of Dell integration / the $3.5M revenue dragging down ROIC / a Submer-GRC price war).

Timing and execution cadence —— Key timing risk: Series B-2 was just completed on 2026-05-14, and the newly arrived Two Seas Capital (an infrastructure/growth fund) just entered and will strongly oppose an early exit—a secondary bid will be needed. ABC Impact's willingness to exit is strong, with a window of 2026-2028 (it has held for 4 years; after Fund II's $600M final close in 2025-04 [S-238] the team's energy turns to the new fund, and the Fund I portfolio enters its harvest period). Recommended cadence: Month 0-3 approach ABC Impact under NDA + explore a secondary; Month 3-6 LOI + data room; Month 6-9 a joint negotiation with Pavilion / Edinv / Ombu; Month 9-12 SPA + UK NSI voluntary notification; Month 12-15 closing. Key dates: 2026-08-01 Inspired retirement (Iceotope has already exited the author roster, so the impact is small); 2026-10 OCP Global Summit (Mullick connects on-site about NVIDIA OEM certification).

Accounting implications of listed-company consolidation —— PPA: a $120M (Path A) acquisition of 50%+1 → controlling but not full ownership → full consolidation per IFRS 10 + minority interest disclosed. Identifiable IP (the 219 patents) ~$15-25M; customer relationships (HPE + Dell + Meta) ~$5-10M; brand ~$3-5M; deferred tax assets (UK R&D tax credit) ~$2-4M; fixed + WC ~$3-5M; total ~$30-50M. Goodwill estimated ~$70-90M. The five goodwill-impairment triggers: ① Mullick departing → OCP standing immediately weakens (the biggest risk); ② long-term absence of NVIDIA Certified; ③ the undisclosed depth of Dell integration → customer relationships cannot be quantified; ④ a small revenue base ($3.5M ARR) significantly drags down the buyer's ROIC; ⑤ a Submer / GRC price war.

Overall judgment —— retained in the Final 3, but repositioned from a "mid-price IP flagship" to a "ceiling-risk candidate," with a preset fallback to a partnership-only or 50%+1 JV structure:

• ✓ the strongest IP density (219 patents + integration depth across five OEMs HPE/Dell/Meta/Intel/nVent, the deepest of the Final 3)
• ✓ the appointment of the dual exit operators Ballon + Jesenko is a textbook exit signal
• ✓ UK jurisdiction (high legal certainty vs. NL/ES) + reduced mandatory-notification risk for cooling equipment after the 2026 NSI reform
• ✓ ABC Impact's 4-year hold enters the exit window
• ⚠️ valuation upside risk touches the $200M ceiling (55% probability of exceeding $200M)
• ⚠️ a highly dispersed cap table (no 25%+ PSC, a joint negotiation of 6-8 institutions)
• ⚠️ Series B-2 was just completed on 2026-05-14, and some newly arrived LPs are unwilling to exit in the short term

Top 5 key risk array: ① valuation exceeding the $200M ceiling (55% probability; the budget must flex to $230M); ② Mullick departing → OCP standing weakens (medium / very high; acqui-hire Avayla to lock in); ③ undisclosed NVIDIA certification (medium / high; connect on-site at the OCP Summit); ④ failure of the cap-table joint negotiation (medium / high; have ABC Impact sign first); ⑤ Two Seas / Barclays newly entered and unwilling to exit in the short term (high / medium; a secondary premium).

4.3.6 GRC (US, the top platform pick)

Business overview —— Green Revolution Cooling, Inc. = a Delaware-incorporated private company (SEC CIK 1457101); headquartered in Austin, Texas; founded 2009-02-02; brand "The Immersion Cooling Authority®." Business model: single-phase tank-style immersion liquid cooling systems (hardware + service + fluid program), an OEM/integrator hybrid. Headcount ~39 (self-reported 2026Q1) / PitchBook reports 61. Revenue 2025E about $8M (RocketReach estimate, unaudited—on the low side relative to the 22-country deployment + $43M cumulative funding magnitude). The ICEraQ product line: ICEraQ Series 10 (>200 kW/rack) / ICEraQ Quad / ICEraQ Micro+SX (2024-2025) / ICEraQ Nano (10U, ≤13 kW, liquid-to-air, released 2025-11-19) / ICEtank (containerized modular DC) / ReliaSys IR500 CDU (500 kW, DLC + RDHX, added 2024-2025—GRC crosses into DLC support, no longer only doing tanks). The ElectroSafe fluid Partner Program: certification of multiple oil/fluid suppliers (ENEOS / SK Enmove / Innovate Immersion)—GRC does not produce fluid itself, only certifies, an asset-light ecosystem model.

Funding history and ownership structure —— Cumulative equity funding $43M (through the 2022 Series C) + the 2025 Samsung strategic investment undisclosed:

Table 4.8: GRC Complete Funding History

Key observation: before 2022 there was no institutional VC participation at all (no Mercury Fund / Energy Foundry—the baseline assumption was wrong and should be removed); the 2025 Samsung round did not disclose an amount → possibly a small strategic investment ($5-15M range), not separately listed as a Series D.

Current cap-table inference: SK Lubricants / SK Enmove 15-25% (lead since 2022, high probability of a board seat + ROFR/drag-along/tag-along/liquidation preference); HTS 5-15% (2016 strategic investment, board representative Derek Gordon); Samsung Ventures 3-8% (2025-06, financial + strategic); ENEOS 3-8% (a fluid-program partner upgraded to a shareholder); Christiaan Best (founder) 10-20%; Peter Poulin (CEO) 1-5%; John Bean (CTO) <2%; management team + ESOP 5-10%; other angels 5-15%. Confirmed board members: Christiaan Best (independent director, former CEO/CTO); Mark Tlapak (independent director—possibly an early-investor representative or legal counsel); the HTS representative (Derek Gordon); the SK Lubricants representative (the $25M lead should have a seat).

Technology IP and product portfolio —— 33 registered patents (the accurate Justia figure; the baseline 30 was too low) [S-124] + Bean's 48-US-patent experience from his Schneider era (already moved to GRC). The main class is 'Electric Techniques Not Otherwise Provided For'; key IP directions are multi-tank coolant distribution / CDU / manifold / immersion racks / modular-based design. Strategic Alliances (key in 2024-2025): ① Cisco Engineering Alliance (2025-02)—a joint Cisco UCS + ICEraQ SX/Micro solution [S-240]; ② Dell Global Alliance Partner (12-year partnership); ③ Vertiv Global Alliance Partner; ④ Intel fluid-compatibility certification; ⑤ LG Electronics + SK Enmove + GRC three-party MOU (2025-10-27)—a Korean three-party AI DC immersion solution [S-242]; ⑥ Samsung C&T BCA (2025-06)—an AI/HPC/Edge reference design [S-239]; ⑦ Gigabyte single-phase immersion collaboration.

OCP influence positioning (inherits Ch3 §3.5) —— GRC ranks #9 (18 points) in the Ch3 §3.7 final ranking—mid-positioned among the Final 3, but the strongest key-individual OCP governance seat. John Bean (CTO, moved in from Schneider in 2021-04) holds the OCP Immersion Committee Co-Chair—the highest OCP governance seat among the Final 3 (vs. Asperitas's Young Solutions TC Co-Chair / Iceotope's Mullick Industry Liaison WS chair). Bean is also a Rev 2.10 co-author + Material Compat author + Fluid Base Spec ack + Design Guidelines reviewer. Nicholas Beardsley (Solutions-FMEA WS Chair, an Electrical Engineer) + Larry Kosch (Design Guidelines reviewer)—together form GRC's "engineering core" in OCP governance. Bean's OCP governance seat is one of the most valuable assets in a GRC acquisition.

Key-individual retention framework ——

• Christiaan Best (founder / now independent director, founded 2009 → handed over the CEO role in 2016 to become CTO → handed over the CTO role in 2021 to become an independent director) —— GRC's soul figure, but has been out of operations for 5 years; estimated age 45-55; his "yes" has a huge impact on management and the OCP ecosystem; Minimus Servers (his own custom-server brand) could become an additional post-acquisition asset.
• Peter Poulin (CEO, 2016–present, 9-year tenure, 30 years of IT-industry experience) —— named to Data Economy's "Top 50 Most Influential Climate Leaders in Data Centers and Cloud" in 2024; drove the SK strategic investment, the Samsung strategic investment, the 22-country deployment, the Cisco alliance, and the Series 10 + Nano iterations; his options should already be cliff-vested.
• John Bean (CTO, 2021-04–present, 5-year tenure) —— his Schneider Electric tenure was 2001-2020 (about 19-20 years, not the baseline 30)—Director of R&D Cooling → Director of Innovation → Director of Innovation and Technology; 48 US patents; an ASHRAE voting member; OCP Immersion Committee Co-Chair; estimated age ~50-55 (the baseline 55-65 was an overestimate) → medium retention risk, still in his career prime; the 5-year tenure still has strong room to grow his technical + OCP influence.
• Nicholas Beardsley (Solutions-FMEA WS Chair) + Larry Kosch (Design Guidelines reviewer) —— together form the OCP-governance "engineering core."

Retention-framework recommendation: the core trio (Best + Poulin + Bean) on a 3-year earn-out, with 30-40% of the deal consideration anchored to operational retention; Bean locked in separately (his OCP role = GRC's say in the immersion-IP ecosystem = the key to the acquisition premium); Best's transfer-consent right may be constrained by a controlling-shareholder agreement under the Delaware-corporation framework.

Valuation basis and range judgment —— GRC's 2026 fair valuation base case is $140-180M post-money—a 75-85% confidence of falling in the $100-200M range, the target with the highest valuation certainty among the Final 3. Multi-path estimation:

Probability of falling in the $100-200M range: 75-85%.

Feasibility of a >50% controlling stake + resolving existing shareholders' interests —— complex but resolvable, 55-70% confidence.

The core gating—SK Lubricants / SK Enmove: as the $25M Series C lead, it almost certainly has a board seat ≥ 1 + ROFR / Right of First Offer + tag-along + drag-along participation + a preferred liquidation preference (inferred 1× non-participating). Samsung Ventures (entered 2025, mainly financial, small amount) + HTS (a dual strategic + financial role, early in 2016) + ENEOS (a fluid-program partner upgraded to a shareholder) are highly likely to follow.

Controlling-stake path analysis: ① the ideal path—negotiate a "trigger drag-along" or share purchase agreement with SK Lubricants, with Samsung + HTS + ENEOS exiting alongside SK simultaneously (Asian strategics are often bundled); founders + management stock + earnout; ② the main obstacle—the two Korean conglomerates SK + Samsung together hold ≈ 28-33%, and they must ALL agree simultaneously for clean control; after its 2022 strategic investment SK has been upgraded to GRC's core fluid supplier + built a three-party AI DC alliance with LG in 2025 → SK regards GRC as the global distribution window for Korean fluids, so its willingness to exit may be low; ③ the practically feasible controlling model—50%+1 control + SK / Samsung retaining a minority stake (more feasible; the consolidation basis is unchanged); or 51-66% control + existing-shareholder LP rights (sub-optimal); 75%+ clean control requires a premium +/- a full exit by the Korean conglomerates (the hardest).

Geopolitical / regulatory overlay (US-CFIUS) —— CFIUS applicability under the 31 CFR Part 800 framework:

• Critical Infrastructure: DCs are explicitly listed in Appendix A; as a DC component supplier (not a DC operator), GRC is medium risk.
• Critical Technologies: immersion cooling is not on the BIS Emerging Technologies list, low risk.
• Government customers: NSA (a 2012 evaluation deployment) + USAF (a 2019 public modular DC) + DoE labs—medium.
• Proximity to sensitive facilities: headquartered in Austin TX, not within a "covered real estate" 35-mile zone, low.
• NDAA Section 889: as a US company, GRC is not directly triggered; if the buyer is PRC-controlled → GRC's DoD-supplier eligibility is affected (a secondary impact).

Key finding: the NSA "validation" is actually a public, non-classified evaluation—the 2012 NSA Advanced Computing Research team at the LPS (Laboratory for Physical Sciences) at the University of Maryland—this is a public NSA research unit; a 2012 evaluation deployment ≠ a current sensitive production deployment; no ongoing NSA service contract is disclosed [S-241]. The CFIUS risk level is far lower than the baseline "NSA classified-program validation" assumption.

CFIUS decision matrix: a US-controlled buyer has no mandatory filing / high passage rate; UK/CA/AU/NZ (Excepted Foreign States) mostly exempt / high 85-95%; South Korea / Japan / Western Europe (allies) + the 2025 Trump America First Investment Policy [S-251] streamlined review / 75-85%; Israel / India / Saudi SWF mandatory filing / 55-75%; PRC / Hong Kong / Russia / Iran / North Korea mandatory filing + near-certain blocking <10%. The meaning of the Trump 2025-02-21 EO: easing for allies + escalating for China—implication for the Final 3: if Advance Studio is a U.S. entity + a Five Eyes coordinated acquisition → GRC is the easiest of the Final 3 to pass CFIUS.

Due-diligence checklist for a >50% controlling stake + listed-company consolidation ——

• Financial DD: GRC's 3-year financials (private, unaudited, require an NDA to obtain); a federal-%-of-revenue audit—determines the necessity of a mitigation agreement; EBITDA / working capital.
• Legal DD: the complete terms of the SK Lubricants 2022 SPA (ROFR / drag-along / board-seat details); the Samsung Ventures 2025 investment agreement; the HTS / ENEOS agreements; Christiaan Best's personal shareholding → the Delaware controlling-shareholder-agreement constraint.
• Technical DD: the 33-patent freedom-to-operate; the ownership of Bean's 48 patents from his Schneider era (confirm they moved to GRC or that the former-employer Schneider interest is retained); the ElectroSafe Partner Program fluid contracts (the SK/ENEOS strategy).
• OCP governance DD: continuity of Bean's Immersion Committee Co-Chair; Beardsley + Kosch retention; the impact of the Inspired 2026-08-01 retirement (GRC's ICEraQ Nano + Series 10 will need to re-apply for Accepted certification).
• CFIUS voluntary filing (even if not mandatory) to obtain a safe harbor.
• Listed-company consolidation accounting: PPA + Goodwill; FOCI mitigation for NSA / DoD customer contracts (if the buyer is a foreign entity).

Timing and execution cadence —— SK is the key gating: first approach SK's leadership separately to confirm GRC's long-term positioning + willingness to exit; if SK refuses a full clear-out, downgrade to 51-66% control + the Korean conglomerates retaining a minority; prepare the CFIUS voluntary filing 75-90 days ahead; conduct a customer-concentration audit (federal % of revenue) in advance. Key dates: 2026-08-01 Inspired retirement (GRC's ICEraQ Nano + Series 10 will need to re-apply for Accepted certification); 2026-10 OCP Global Summit San Jose (Bean Co-Chair connects on-site); verification of the landing status of the 2025-10 LG-SK Enmove-GRC MOU.

Accounting implications of listed-company consolidation —— PPA: a $150M acquisition of 50%+1 → controlling but not full ownership → full consolidation + minority interest. Identifiable IP (33 patents + the ElectroSafe Partner Program) ~$10-20M; customer relationships (22 countries / Cisco / Dell / Vertiv / Samsung C&T / LG / NSA-LPS / USAF) ~$10-20M; brand (the "Immersion Cooling Authority®" trademark) ~$3-5M; fixed + WC ~$5-10M; total ~$30-55M. Goodwill estimated ~$95-120M. The five goodwill-impairment triggers: ① Bean departing → loss of the OCP Immersion Committee Co-Chair (the biggest risk); ② a change in the Korean-conglomerate relationship (SK strategic pivot); ③ the Cisco alliance / LG MOU failing to land; ④ NVIDIA certification not mentioned (the same risk as Asperitas/Iceotope); ⑤ a change in the ElectroSafe Partner Program fluid contracts.

Overall judgment —— the top pick to proceed:

• ✓ the most certain valuation in range (75-85% probability of $100-200M, base case $140-180M)
• ✓ the strongest OCP governance seat (Bean Immersion Committee Co-Chair)
• ✓ CFIUS risk far below baseline (NSA-LPS is a public, non-classified research laboratory; the 2025 Trump EO is streamlined for allies)
• ✓ the broadest product portfolio (Tank + Modular ICEtank + DLC ReliaSys IR500 CDU + Edge Nano)
• ✓ dense strategic partnerships (Cisco / Dell / Vertiv / Intel / Samsung / LG / SK / Gigabyte)
• ✓ Bean retention risk below baseline (Schneider 19-20 years not 30, age ~50-55 not 55-65)
• ⚠️ the Korean / Japanese conglomerates SK + Samsung + ENEOS hold ≈ 28-33% cumulatively, requiring top-level political + commercial persuasion in parallel
• ⚠️ undisclosed NVIDIA certification (a risk shared with Asperitas/Iceotope)

Top 5 key risk array: ① SK Lubricants refusing a full exit (medium / very high; downgrade to 51-66% control + the Korean conglomerates retaining a minority); ② Bean departure risk (low / high; a 3-year earn-out + an OCP retention bonus); ③ internal disagreement among the Korean conglomerates (whether SK + Samsung + ENEOS exit as a bundle, medium / medium); ④ CFIUS customer sensitivity (low / medium; a customer-concentration audit first); ⑤ re-application for ICEraQ Nano + Series 10 Accepted certification after 2026-08-01 (low / medium; communicate with the OCP foundation in advance).

4.4 Three-Company Comparison, Final Ranking, and Comprehensive Recommendation

4.4.1 Three-Company Key-Parameter Comparison Table

Table 4.9: Final 3 Key-Parameter Comparison

4.4.2 6-Dimension Scoring Matrix

Table 4.10: Final 3 × 6-Dimension Scoring (H=3 / M=2 / L=1)

4.4.3 Complementarity vs. Substitution Analysis

The three are fully complementary, with no substitution on geography / form / OCP influence:

• Geographic substitutability = 0: NL / UK / US spread, with no market-coverage overlap.
• Form substitutability = 0: the three forms cassette + direct-forced-convection (Asperitas) / chassis (Iceotope) / tank (GRC) do not compete with each other.
• OCP-influence substitutability = 30%: all are within the Cooling Environments / Immersion project, but Asperitas holds the Solutions TC + historical spec IP, Iceotope holds the Industry Liaison WS + the patent wall, and GRC holds the Immersion Committee Co-Chair and the Material Compat + FMEA engineering line.

Theoretical combination strategies: ① acquire 1 company + acqui-hire Brink/Promersion + tuck-in Engineered Fluids → a complete influence chain; ② acquire GRC + a subsequent 5-10% LP investment in Iceotope → transatlantic influence + reducing GRC's single US-CFIUS exposure.

4.4.4 Final Ranking and One-Line Positioning

Priority 1: GRC (US, the top platform pick) —— 41 points
"Certain valuation + the strongest governance seat + the broadest product + actual CFIUS risk below baseline—the only target among the Final 3 satisfying all three axes 'certain valuation × OCP strategic seat × feasible control,' recommended to enter at $150M 50%+1 + a Bean 3-year earn-out to lock in the Immersion Committee Co-Chair."

Priority 2: Asperitas (NL, the spec-positioning alternative) —— 35.5 points
"Cheap valuation ($70-120M) + the Open Cassette spec legacy + an 81% Vifo Act passage rate + Young's CTO retention can be locked in—the most cost-effective 'spec-positioning' target among the Final 3, but the missing NVIDIA certification requires 12 months to fill and the three shareholders' (Shell + Invest-NL + STECON) stacked consent rights require 4-6 months of negotiation."

Priority 3: Iceotope (UK, the IP flagship, valuation-ceiling risk) —— 34.5 points
"219 patents + the dual exit-operator appointments of Ballon/Jesenko + the ABC Impact exit window already open—but having just taken a $26M B-2 in 2026-05, the valuation is most likely $150-260M (55% probability over $200M); recommended to enter via Tier A 50%+1 voting control ($120M); if the valuation exceeds the $200M ceiling, adjust to partnership-only or a 30-49% strategic minority."

4.4.5 Pricing Stance: Option vs. Cash Cow (echoing §2.4 + §3.10)

Immersion is still "an inflection not yet arrived" rather than the current mainstream—Ch1 §1.3.2 shows immersion is only ~1% of the frontier AI rack BOM, Ch2 §2.4 shows cold plate is the current AI-build workhorse, and Ch3 §3.10 shows the real influence of immersion specs is still highly individual-centric (anchored by the four people Brink + Bean + Young + Mullick).

These targets should be priced as an "option" rather than a "cash cow"—the value lies in locking in scarce independent pure-immersion assets + spec positioning while the landscape is undecided and valuations have not yet been pushed up by the giants. The three companies' pricing should be anchored as:

GRC's "option / cash cow = 2:1" is the most balanced of the Final 3—and the reason it ranks first in §4.4.4.

4.4.6 Risk Array and Monitoring Points

Risk array shared across the three: ① NVIDIA GB200/B200 chassis-level certification is undisclosed for all (the biggest shared risk; on-site contact needed at the 2026 H2 OCP Summit); ② PE bidding rivals entering (the Blackstone-ACT 2026-03 + Carrier-ZutaCore 2026-04 signals; the immersion segment may follow in 2026 H2); ③ the 2026 H1 unwillingness-to-sell posture (all three of the Final 3 are on an "operate independently + strategic financing / collaboration" path, and the window is most likely pushed to 2026 H2-2027); ④ the PFAS regulation timeline is favorable to single-phase but requires clear marketing (the 13.5-year buffer is irrelevant to single-phase, but readers of the acquirer's segment report need to understand this); ⑤ the 2026-08-01 Inspired retirement affecting the three companies' SP/Accepted-eligibility continuity (Asperitas Open Cassette + GRC ICEraQ Nano/Series 10 + Iceotope's 2026 Marketplace listing plan all need to be reviewed).

3 catalyst monitoring points (determining the launch timing for 2026 H2 – 2027):

• ① A formal signal of ABC Impact's willingness to exit (Iceotope) —— the Fund I portfolio enters its harvest period; an NDA contact is expected to be possible in 2026 H2.
• ② SK Enmove's strategic pivot (GRC) —— the three-party coordination of the Korean conglomerates SK / Samsung / ENEOS; if SK Lubricants strategically reviews its GRC holding, an NBO can be triggered.
• ③ Completion of the Invest-NL mandate + STECON SE Asia landing (Asperitas) —— Invest-NL's 2025-09 ticket has a cash-out intention within 2-3 years.

4.4.7 Comprehensive Recommendation + Watch List

Core recommendation: enter in two steps between 2026 H2 – 2027—Phase 1 (2026 H2) first establish an NDA signal with the three parties ABC Impact / SK Enmove / Invest-NL + monitor catalyst triggers; Phase 2 (2027 H1) launch the NBO according to the priority of catalyst signals, with the goal of completing 1 Final platform acquisition (highest priority GRC, alternatives Asperitas / Iceotope) + simultaneously acqui-hiring Promersion (Brink) as an OCP spec-pen lock-in + tucking in Engineered Fluids (coolant IP) as a single-phase fluid backup.

Watch list (independent immersion players not selected but to be monitored continuously): ① Midas Immersion Cooling (US); ② DCX (PL); ③ Immersion4 (CH); ④ Synano (a nano-oil company partnering with GRC, newly appearing at EMEA 2026); ⑤ KYZEN (cleaning chemistry, newly entering the OCP Cleaning Guidelines WS). These targets are smaller in scale but diverse in form, and after further data disclosure in 2026 H2 – 2027 it can be re-evaluated whether they enter a Final 4 or 5.

Final stance: these targets should be priced on "option + timing discipline"—the value is not in current revenue (all three have ARR < $10M) but in OCP governance positioning + the IP option. Timing + valuation discipline is the decisive factor of this mandate. GRC is the only target among the Final 3 satisfying all three axes (certain valuation × OCP governance seat × feasible control), and is recommended as Priority 1 to launch during 2026 H2 – 2027 H1.

Conclusions and Limitations

Conclusions:

1. Liquid cooling is a high-growth, low-share slice of AI infrastructure; cold plate dominates the present, while immersion is a future bet constrained by coolant and operations.
2. The dominance of OCP immersion specs is highly concentrated and individual-centric (Brink: Asperitas → Promersion) + Intel; this defines "whom you acquire to obtain real IP and a standards seat."
3. Under the "$100–200M, pure-immersion, still-independent" filter, the shortlist = GRC / Iceotope / Asperitas (+ tuck-in); the wave of giant acquisitions makes pure-immersion independent targets a scarce opportunity window, but they should be priced with option thinking, strictly observing timing and valuation discipline.

Limitations and items to verify:

• Valuation: the valuations of GRC, Iceotope, and Asperitas are all undisclosed by first-hand sources; Submer's $500M is a rumor, not first-hand. All "range-fit" judgments must be confirmed at the NBO / data-room stage.
• Inflection uncertainty: immersion has not yet been widely deployed in AI clusters, and there is substantive uncertainty about when the sector will take off (a bet by nature).
• Definitions: some performance/cost figures (PUE, $/rack, heat transfer coefficient) are vendor or textbook figures, and the body has marked them "illustrative only."
• Dynamic items: CoolIT → Ecolab is a definitive agreement, not yet closed (expected 2026Q3); the candidate pool still needs further screening of smaller independent vendors such as Midas / DCX / Immersion4.
• Geopolitics: feasibility varies with the buyer entity's nationality (CFIUS / UK NSI / EU-NL FDI / Australia's FIRB); this report serves as an overlay and does not substitute for legal advice.