F.F

AI

NeoCloud Bare-Metal Unit Economics: A Three-Year Rental Model

31 July 2026

TL;DR

A NeoCloud is not a miniature hyperscaler. In its purest form it is a balance-sheet business that buys accelerators, parks them on contracted power, and sells GPU-time—often as bare metal, sometimes as Kubernetes or slurm-wrapped clusters. The public narrative emphasizes backlog and megawatts. The private spreadsheet emphasizes four clocks that rarely align: hardware purchase ramp, colo energization lag, customer occupancy ramp, and price decay across a GPU generation.

This article does three things. First, it places NeoCloud in the supply stack between wholesale AI data centers and hyperscale GPU clouds. Second, it decomposes bare-metal rental into hardware cost, data-center host cost, team cost, and revenue, separating fixed parameters from assumptions you should be allowed to edit. Third, it ships an interactive three-year model—summary KPIs plus a full monthly revenue and cash-opex table—so a reader can see when a “high EBITDA margin” story still fails cash payback.

The default case is intentionally tight: new-build H100-class capex near $30k per GPU, list rental near $1.68 per GPU-hour, 36-month economic life, and ordinary colo and payroll drag. Under those inputs the model often prints strong mid-period EBITDA while cumulative free cash remains negative inside the horizon. That is not a bug in the calculator. It is the NeoCloud problem statement.

1. What a NeoCloud Actually Sells

1.1 Definition and boundary

NeoCloud is the market label for AI-native GPU clouds that grew outside the classic hyperscaler trio. Operators such as CoreWeave, Lambda, Crusoe, Voltage Park, Together, and a long tail of regional hosts typically share four traits:

The product customers think they buy is “H100s” or “B200s.” The product the P&L actually sells is billable GPU-hours on a financed asset, inside a power envelope, under an SLA, after sales leakage. Confusing the chip with the hour is how decks overstate returns.

1.2 Position in the AI infrastructure stack

Layer What is sold Dominant cost Typical buyer
Power / land / interconnect MW, water, fiber, entitlements Grid, generation, civil works Hyperscalers, developers, utilities
Wholesale AI colo / powered shell $/kW-month capacity Building, cooling, electrical NeoClouds, enterprises, clouds
NeoCloud bare metal / GPU cloud $/GPU-hour or cluster-month GPU + fabric + colo + ops Labs, AI apps, enterprises
Hyperscale GPU SKUs Managed instance families Fleet + software + sales Broad developer base
Model / agent platforms Tokens, apps, outcomes Model, data, product End users

NeoCloud gross margin is therefore a spread: rental price per hour minus the cash cost of keeping that hour alive. Depreciation is not optional intellectually even when lenders underwrite EBITDA. A six-year accounting life on a three-to-four-year economic life is a financing choice, not a physics result.

1.3 Bare metal versus managed GPU cloud

Bare metal is the cleanest unit-economic object. The customer receives nodes (or NVL racks), brings a runtime, and pays for reserved capacity. Managed Kubernetes, serverless inference, and multi-tenant slicing add attach revenue—and support cost, noisy-neighbor risk, and product complexity. The calculator below models bare-metal rental with a simple attach rate for storage, egress, and support. It does not model multi-tenant packing efficiency above the stated occupancy rate; if you believe packing beats 99% “sold occupancy,” raise occupancy rather than hiding the claim.

2. Three Cost Layers and One Revenue Engine

Every NeoCloud bare-metal deal can be forced into four blocks. Mixing them is how models double-count power or forget that payroll is mostly fixed.

2.1 Hardware (balance sheet)

Annualized failure rate (AFR) does not always reduce billable hours if hot spares exist; it does force a spares and maintenance cash budget. The model charges AFR × GPU price / 12 on the deployed base.

2.2 Data-center hosting (mostly operating cost)

2.3 Team and G&A

A thin bare-metal pod can run with a small site reliability and customer operations footprint. Defaults: 3 people at $100k average fully loaded cash cost, plus a fixed other-G&A stub and a sales commission on net revenue. This understates a public NeoCloud’s corporate overhead; it is closer to a single-cluster economic unit than to company-level SG&A.

2.4 Revenue

2.5 Metrics that are easy to omit

Beyond headline ROI, the model surfaces:

Not modeled, and material in real deals: debt draws and interest, GPU collateral advance rates, customer concentration, power curtailment, liquid-cooling premiums, import duties, and residual-value execution risk. Treat the sheet as a cluster underwriting sandbox, not a full corporate LBO.

3. Interactive three-year bare-metal calculator

Edit assumption fields (solid borders). Fixed commercial inputs use dashed borders but remain editable so quotes can be updated. Presets for H100 / H200 / B200 / GB200 NVL72 overwrite GPU price, power, and GPUs per node. The monthly grid is wide on purpose—scroll horizontally like a workbook.

Bare-metal rental model · 36 months

Cluster-level cash and P&L. Defaults assume H100-class hardware, locked $1.68/GPU-hour, staged delivery, and ordinary colo drag. Reset restores defaults.

Assumptions · Data center hostingEditable · operations & capacity

N production racks per management rack; expands reserved kW.

Assumptions · HardwareEditable · fleet & accounting life

Assumptions · Team & revenueEditable · people and price path

Fixed parametersQuote inputs · dashed fields still editable

Three-year return summary

Cash metrics ignore depreciation. Accounting payback asks when cumulative EBIT recovers initial capital. IRR is annualized from monthly free-cash flows. Financing (interest, advance rates) is excluded.

Initial invested capital GPU + ancillary + NRC
36-mo cumulative free cash EBITDA − capex − NRC
Cash ROI (36 mo) Cum. free cash ÷ invested capital
Payback (ex-depreciation) Cash recovery of capex
Payback (with depreciation) Cum. EBIT covers capital
IRR (annualized) On monthly free cash
Peak funding need Deepest cash trough
36-mo net revenue After SLA & bad debt
36-mo EBITDA Before dep / NRC amort.
EBITDA margin EBITDA ÷ net revenue
Billable GPU-hours 36-month total
Realized $/GPU-hour GPU rental ÷ billable hours

Annual roll-up

Period Net revenue Cash opex EBITDA EBIT Free cash

Monthly detail (M1–M36)

Rows follow a landlord-and-operator workbook: billable hours → revenue haircuts → colo and energy → spares → people → EBITDA → depreciation → EBIT. Memo rows show occupancy-independent cash opex and cumulative free cash.

4. How to Read the Model Without Fooling Yourself

4.1 The EBITDA trap

NeoCloud pitch decks often lead with adjusted EBITDA margins in the 50–70% range once a cluster is full. The default sheet can reproduce that optical: after ramp, monthly EBITDA is large relative to net revenue because GPU depreciation is non-cash. Cumulative free cash is the adult metric. If peak funding exceeds what your advance rate and equity can carry, the cluster dies of liquidity even while “unit economics work.”

4.2 Four ramps, one spreadsheet

Returns are path-dependent. Delay between purchase and deployment burns insurance-like and sometimes rent before revenue. A long occupancy ramp with fixed rent and payroll is a classic cash crush. Price decay on renewals can erase year-three cash that the locked case assumed. When you stress the model, change one ramp at a time and watch peak funding, not only terminal ROI.

4.3 Sensitivities that usually matter most

Lever Direction that helps cash Why
GPU capex / secondary purchase Lower Dominates invested capital
$/GPU-hour and lock length Higher, longer Revenue density and decay protection
Occupancy and fill ramp Higher, faster Fixed colo and payroll leverage
$/kW rent and energy Lower Largest variable + semi-fixed cash opex
Deployment gap Shorter Reduces pre-revenue cash burn
Economic life vs. depr. life Aligned Prevents fake accounting returns

A useful diligence habit: force depreciation life down to the commercial lease length and re-check accounting payback. If the equity story only works on a six-year life with a three-year customer, you are underwriting refinancing hope.

5. Market Context: Why NeoClouds Still Exist

Hyperscalers have compressed the list-price gap on popular SKUs, and GPU supply is less apocalyptic than in 2023–24. NeoClouds still clear demand when they offer faster cluster delivery, reserved multi-node fabrics, flexible contract sizes, or power that hyperscalers have not yet absorbed. Their fragile edge is procurement and energization speed; their structural risk is that the asset they finance is a depreciating semiconductor generation sitting on someone else’s building.

Public comparables illustrate the tension. Operators can post multi-billion revenue run-rates and enormous RPO while still showing thin GAAP profitability, heavy leverage, and concentrated counterparties. That pattern is consistent with a business that capitalizes GPUs aggressively, sells hours under multi-year contracts, and hopes residual value and refinancing cooperate. The calculator is the cluster-level version of the same argument.

6. Conclusion

NeoCloud bare metal is a financed hardware rental business with a power landlord and a short technological half-life. The right underwriting artifacts are not slogan margins but: invested capital, peak funding, cash payback without depreciation mythology, and accounting payback when life equals lease. Use the sheet to ask whether a given quote is a contract worth signing—or a generation tax dressed up as cloud.

If you extend the model, the highest-value additions are usually debt schedule and advance rate, contractual take-or-pay floors, residual sale at month 36, and a second GPU vintage for refresh. The version on this page stays transparent: every assumption is visible, every month is listed, and nothing is hidden in a black-box “AI infra multiple.”