Home Newsletter All-Access Portfolios Blog Insights Video Library About

Ten Years in Six Months

The returns on AI capacity are far better than they look — inference earns roughly three times its full cost. The only real question is the path, and whether the financial system stays healthy long enough to finish paying for it.

Raman Bindlish · Research Note · 16 September 2026
Revenue vs all-in rent, per GW
~3x
Share of ceiling to break even
~10%
Enterprises with measured ROI
7%
📄  Get the complete Proflex report — free

“Ten Years in Six Months” — 36 pages, 10 figures and 10 tables, every number sourced. Read the summary below, or have the whole report sent to you now.

Previous edition: Tesla — the three-legged thesis →

On 8 September, OpenAI announced that roughly 10,000 autonomous agents, running an unreleased internal model, had resolved the Navier–Stokes existence and smoothness problem — one of the six Millennium Prize Problems still open. Time to a proof: 88 hours. Time for a separate model to formalise it in Lean: a further 17.

The result is disputed on provenance, and barely reviewed. But notice what it actually was. Ten thousand agents for eighty-eight hours is not a training run. It is inference — on the order of 880,000 agent-hours of frontier compute spent on a single problem. That is a demand signal as much as a capability one, and inference demand is exactly what has to fill the capacity this industry is building.

Our last report, The Rental Bill, spent fifty pages on why the capital behind the AI buildout is ahead of the revenue. Everything in it still stands. But the models got dramatically better while we were all arguing about the balance sheets — and when you redo the arithmetic on the right denominator, the answer changes.

The argument in one line. The return on AI capacity is not the risk — it looks exceptional. The risk is the path: adoption outside the technology sector is shallow, and the buildout is being financed by an economy that is not yet receiving the benefit.

1. “AI labs can’t make money” is a measurement error

Here is the most common calculation in AI commentary. Take the labs’ revenue, divide it by all the AI computing capacity that exists, observe that the result is small, conclude the business does not work.

The error is in the denominator. A large share of that capacity is doing training, and training generates no revenue by design — it is R&D. Including it is like dividing a drug company’s revenue by its laboratories plus its factories and calling the result a margin.

Chart 1
“AI can’t make money” is a denominator error
The denominator effect: revenue per total GW versus per inference GWGrouped columns. Revenue per total GW rises from $10B in 2023 to $25.5B in 2028E purely because inference share rises from 33% to 85%; revenue per inference GW held at $30B.$ BILLIONS REVENUE PER GIGAWATT PER YEAR$0B$10B$20B$30B2023: revenue per TOTAL GW $9.9B (what gets quoted)$10B2023: revenue per INFERENCE GW $30B$30B2023inference 33% of compute2025: revenue per TOTAL GW $15.0B (what gets quoted)$15B2025: revenue per INFERENCE GW $30B$30B2025inference 50% of compute2026: revenue per TOTAL GW $20.1B (what gets quoted)$20B2026: revenue per INFERENCE GW $30B$30B2026inference 67% of compute2028E: revenue per TOTAL GW $25.5B (what gets quoted)$26B2028E: revenue per INFERENCE GW $30B$30B2028Einference 85% of computePer total GW — what gets quotedPer inference GW — the operating reality

Source. Holds revenue per inference GW constant at $30B, so every movement in the lighter bars is mix, not performance. Inference share of compute: Deloitte cycle-based estimates (spend-based measures run as high as 80–90%).

Inference was about a third of AI compute in 2023, half in 2025 and roughly two-thirds in 2026. Hold revenue per inference gigawatt constant and revenue per total gigawatt still climbs from $10B to $26B — purely from mix. Which cuts both ways: reported AI economics are about to look considerably better without anything fundamental changing, and anyone extrapolating that improvement is extrapolating an accounting artifact.

The same error runs through the margin line. 2026 estimates of Anthropic’s gross margin range from 44% to 60% — sixteen points that turn entirely on how training cost is treated.

2. What a gigawatt actually earns

The cost side comes from The Rental Bill: roughly $38B to build a one-gigawatt AI datacentre plus $0.9B a year to run it. Amortised over a five-year hardware life at an 8% cost of capital, that is a rent of $8.5–10.4B per gigawatt per year.

The revenue side now comes from three independent directions. A bottom-up token model — about 7,700 NVL72-class racks per gigawatt at 30,000 tokens per second — produces a theoretical ceiling near $100B a year at full utilisation and list prices. Broadcom’s Hock Tan, at Goldman Sachs Communacopia this month, offered an illustration of $30B of annual recurring revenue against about $10B of annual cost per gigawatt. And $30B is exactly where our model sits at 30% of its ceiling.

Chart 2
Inference capacity earns roughly 3x its full all-in cost
Revenue per gigawatt of inference capacity against the all-in rentColumn chart: rent about $9B per GW per year; Hock Tan's illustration $30B; Proflex model at 30% of ceiling $30B; theoretical ceiling about $100B.$ BILLIONS PER GIGAWATT PER YEAR$0B$25B$50B$75B$100BAll-in rent per GW per year, $8.5–10.4B$9BRentall-in costHock Tan illustration: $30B ARR per GW$30BHock TanBroadcom illustrationProflex bottom-up token model at 30% of ceiling: $30B$30BProflex model30% of ceilingTheoretical ceiling at full utilisation and list prices: ~$100B~$100BTheoreticalceilingthe rent line

Source. Rent: Epoch AI 1 GW datacentre cost, amortised over five years at 8% ($8.5–10.4B). Hock Tan, Goldman Sachs Communacopia, Sep 2026 (Broadcom sells silicon to these labs — an interested party). Proflex bottom-up token model. Ceiling assumes full utilisation at list prices, which no operator achieves.

A caveat we would put in bold if we could: Broadcom sells silicon to these labs, so Tan is an interested party. His figures are used because they agree with a cost model and a token model built without reference to them. And nobody publishes fleet utilisation, which is the single largest unknown here. But look at how wide the margin for error is.

Chart 3
It pays for itself at ~10% of its ceiling
Coverage of all-in rent by realisation rateHorizontal bars: 10% of ceiling covers rent 1.1x; 20% 2.2x; 30% 3.3x; 50% 5.6x; 100% 11.1x. Dashed line at 1.0x break-even.REVENUE ÷ ALL-IN RENT, BY REALISED SHARE OF THE CEILING10% of ceiling$10B / GW / yr10% realised: $10B, 1.1x coverage1.1x20% of ceiling$20B / GW / yr20% realised: $20B, 2.2x2.2x30% — Tan's illustration$30B / GW / yr30% realised: $30B, 3.3x3.3x50% of ceiling$50B / GW / yr50% realised: $50B, 5.6x5.6x100% — theoretical$100B / GW / yr100%: $100B, 11.1x11.1xbreak-even 1.0x

Source. Coverage of the full all-in cost of capacity — depreciation, cost of capital and operating expense — against a ~$9B/GW/yr rent. Fleet utilisation is published by nobody; it is the single largest unknown in this analysis.

Inference capacity needs to realise only about 10% of its theoretical ceiling to cover its entire cost — depreciation and cost of capital included, not just electricity. You can be wrong about utilisation by a factor of three and the capacity still pays for itself.

📄  Get the full report — free

“Ten Years in Six Months” — the complete 36-page Proflex research note. The per-gigawatt model with every assumption exposed, the open-source question with production routing data, the circle problem, the macro transmission chain, the straight-line-versus-sawtooth framework and the seven-indicator watch list. Every figure sourced, including where sources conflict.

3. The labs are already demonstrating it

Chart 4
Anthropic’s gross margin: −94% to ~60% in two years
Anthropic gross margin 2024 to 2028EColumns around a zero line: −94% in 2024, about 60% in 2026, about 77% projected for 2028.ANTHROPIC GROSS MARGIN-100%-50%0%+50%+100%2024 gross margin: −94%−94%2024reported2026 gross margin: ~60% (44–60% range)~60%2026est. 44–60%2028 projection: ~77%~77%2028Ecompany projection

Source. Published estimates and company-attributed figures; 2028 is a company projection. The 2026 figure is estimated at 44–60% depending on how training cost is treated — 60% is the upper bound. Driven by inference efficiency, not price increases.

Anthropic’s gross margin moved from −94% in 2024 to roughly 60% in 2026, driven by inference efficiency rather than price increases — the more durable of the two mechanisms. It reported more than $11.5B of revenue in Q2 2026, turned adjusted operating income positive, and is tracking toward more than $1B of operating profit in Q3. OpenAI looks worse for an explicable reason: about 900 million free users drag its gross margin down by twenty to thirty points. That is a business-model choice about consumer free tiers, not a difference in the unit economics of inference.

And if capacity were heading for a glut, the price of renting it would be falling. It is doing the opposite.

Chart 5
Capacity is repricing up: 2.6x in under a year
IREN contract pricing per megawatt, Nov 2025 to nowColumns rising from $9.70M per MW in Nov 2025 to about $25M in current discussions, roughly 2.6 times.IREN CONTRACT VALUE, $ MILLIONS PER MW OF IT LOAD$0M$10M$20M$30MNov 2025 Microsoft 5-year: $9.70M/MW$9.70MNov 2025Microsoft, 5yrMay 2026 NVIDIA 5-year: $11.33M/MW$11.33MMay 2026NVIDIA, 5yrAug 2026 3-year deals: over $20M/MW$20M+Aug 20263yr dealsCurrent customer conversations: ~$25M/MW~$25MNowlive discussions

Source. IREN contract pricing per MW of IT load; company statements and reporting, Sep 2026. These appear to be total contract value per MW, not annual (a $25M/MW three-year deal is ~$8.3M/MW/yr). The level is ambiguous; the direction is not.

IREN priced a five-year Microsoft contract at $9.70M per megawatt in November 2025, NVIDIA at $11.33M in May, three-year deals above $20M in August — and about $25M is now showing up in live conversations. Rising prices on multi-year contracts are what scarcity looks like, not the early stage of an overcapacity cycle.

4. Open weights won the tokens. Frontier labs kept the dollars.

Chart 6
Open-weight models won the tokens, not the dollars
Open-weight share of tokens versus share of spendingLine chart: open-weight token share rose from 11% in April to 29% in June and 62% on 22 August 2026, while share of spending stayed under 4% (dashed line).OPEN-WEIGHT SHARE OF VERCEL AI GATEWAY0%20%40%60%Apr 2026: 11% of tokens11%Apr 2026Jun 2026: 29% of tokens29%Jun 202622 Aug 2026: 62% of tokens62%22 Aug 2026Share of spending: under 4%share of SPENDING: under 4%share of TOKENS

Source. Vercel AI Gateway data via OpenRouter analysis and industry reporting, 2026. Open weights run at roughly a tenth of the platform’s average token price; in June the four leading US frontier labs took 95% of the money. What customers spend is not what they earn.

Open-weight models went from 11% of tokens on Vercel’s AI Gateway in April to a record 62% by 22 August, while taking under 4% of the spending. That confirms the structure Tan described. It does not confirm his inference — a tenfold price gap fully explains that spending split without any difference in delivered value. What enterprises spend is not what they earn, and we could find no study measuring enterprise ROI by model class. For the investment case the point is simpler: a gigawatt serving open models is still a gigawatt earning its rent.

5. The contradiction in the extinction week

In the same week Dario Amodei published “We Must Pace the Frontier” — asking Washington for a narrow antitrust waiver so rival labs could coordinate on safety — The Information reported what Anthropic has been contracting for.

Chart 7
Up to $517B of compute, contracted in 11 months
Anthropic compute commitments by counterpartyHorizontal bars: reported ceiling up to $517B; named deals $275B; AWS $100B, Fluidstack $50B, Nscale $45B, SpaceX $45B, Lambda $35B.ANTHROPIC COMPUTE COMMITMENTS SINCE OCT 2025, $ BILLIONSReported ceilingincl. options / LOIsReported ceiling: up to $517Bup to $517BNamed 2026 dealsthe firm floorNamed 2026 deals: $275B+$275B+AWS10-yearAWS: $100B over 10 years$100BFluidstackFluidstack: $50B$50BNscaleNscale: $45B$45BSpaceXSpaceX: $45B$45BLambdaLambda: $35B$35B

Source. The Information via Forkast, TechCrunch and secondary reporting, Aug–Sep 2026. $275B of named deals is the firm floor; $517B is the reported ceiling and may include options and letters of intent. At $8.5–10.4B/GW/yr, 14.8 GW implies $126–154B a year of eventual rent — Proflex arithmetic, not a disclosed figure.

Up to $517B of compute and at least 14.8 gigawatts in eleven months, with an IPO prospectus expected within weeks. Apply the rent arithmetic and that capacity implies $126–154B a year of eventual rent against a tracked run-rate of about $74B. Our reading: the risk concerns are not insincere, but a pace limit plus an industry-funded pre-release testing body is, operationally, a compliance regime — and compliance is a fixed cost that favours the incumbents who can already afford it. That is ordinary American regulatory capture, not a scandal. It does not change the capacity forecast. It does raise the political risk premium on the sector.

6. The circle problem: the heaviest AI user is AI

The deepest, best-measured penetration of these tools is in software engineering — where junior developer demand is down roughly 40% and Gartner expects 75% of developers to orchestrate rather than write code by year end. Software engineering is also, to a large degree, the activity of the technology sector itself, including building AI. Capability compounds inside the loop. The economy-wide effect does not, because the benefit has not yet left it.

Chart 8
In production at 75%. Measured ROI at 7%.
Enterprise AI outcomes, 2026Horizontal bars: 75% AI in production, 66% productivity gains, 20% revenue growth, 7% measured ROI, 6% at least 5% of EBIT from AI.SHARE OF SURVEYED ENTERPRISES, 2026AI in production75% have AI in production75%Report productivity gains66% report productivity gains66%See AI-driven revenue growth20% see AI-driven revenue growth20%Established, measured ROI7% established measured ROI7%≥5% of EBIT from AI6% attribute 5%+ of EBIT to AI6%

Source. Deloitte survey of 3,235 leaders; KPMG pulse of 2,145 leaders across 20 countries, June 2026; enterprise AI production survey, August 2026. Seven per cent is not a verdict on the technology — it measures how far diffusion has travelled.

Three-quarters of enterprises now run AI in production. Only 7% report established, measured ROI, and only 6% can attribute 5% or more of operating profit to it. AI demonstrably compresses discovery — the search across chemical space, the hunt for a counterexample, the first draft of a system. It has not yet shown that it compresses validation — the Phase III trial, the regulated deployment, the organisational confidence to act. That is where most of the cost and nearly all of the time sits.

7. Two per cent of everything

Chart 9
The buildout is now a macroeconomic object
Macroeconomic footprint of AI spendingHorizontal bars: 68.8% of the change in private fixed investment in Q2 2026; about 25% of real GDP growth since 2023; about 2% of GDP.THREE MEASURES OF THE AI BUILDOUT'S FOOTPRINTAI share of the change inUS private fixed investment, Q2 202668.8% of the change in real private fixed investment, Q2 202668.8%Tech investment shareof real GDP growth since 2023About 25% of real GDP growth since 2023~25%AI + datacentre spendas % of US GDP, 2026About 2% of GDP in 2026~2%

Source. Different bases, so read each bar on its own: BEA-derived analysis (private fixed investment, Q2 2026); Deloitte (tech investment share of real GDP growth since 2023); analyst estimates via Forbes, June 2026 (share of GDP). Estimates of AI’s contribution to GDP growth range from ~20% to 65% and do not reconcile.

The US is on track to spend about 2% of GDP on AI and datacentre infrastructure this year. In Q2 2026, AI accounted for roughly 69% of the change in real private fixed investment. Hyperscaler AI capex has gone from about $235B in 2024 to over $700B projected for 2026. Whatever happens to AI capital spending now happens to the investment component of GDP more or less directly — with AI-linked debt heading toward $570B at the margin.

The mitigant is real: this buildout is financed mostly from internally generated cash flow. Amazon, Microsoft, Alphabet and Meta can absorb a disappointing few years. WorldCom could not. The exposure sits in the marginal financing — the SPVs, neocloud debt and vendor guarantees — and that is where a repricing would start.

Straight line or sawtooth

Both paths end in the same place. On the straight line, adoption broadens into ordinary industries roughly on the schedule the financing assumes, and 2026 capacity looks like one of the best capital allocations of the decade. On the sawtooth, capex expectations break before adoption matures, the unwind takes a chunk of GDP with it — and the capacity that would have produced the abundant economy stops being built. Interconnection queues run four to seven years; large transformers have 128-week lead times. A project that loses its financing in 2028 does not resume in 2029.

The technology does not need more capability to justify the spending. It needs more customers outside its own industry, sooner than the debt schedule implies. The technology is no longer the variable. The path is.

📄  The watch list is in the full report

Seven indicators that tell the straight line from the sawtooth — enterprise measured ROI, inference share and utilisation, neocloud contract pricing, Anthropic’s operating profit, non-tech adoption, AI-linked credit spreads and AI’s share of private fixed investment — with the specific reading that would change our mind in each direction.


Figures are from the Proflex Finance research note “Ten Years in Six Months” (10 September 2026), which cites every source and shows its arithmetic. Neither OpenAI nor Anthropic publishes audited accounts; fleet utilisation is published by nobody. Figures marked as estimates or model outputs are Proflex arithmetic, not disclosed numbers. This field moves faster than any document — verify anything you intend to act on against current sources.

⭐ All-Access — one month for $99

Every trade in this book reaches you in real time — entry, strike, credit target, and every roll after. You place the orders in your own brokerage, at your own size.

Real-time trade alerts by email, Discord and WhatsApp — entries, rolls and exits across the whole model book, including the AI names this research feeds.

A premium networking community — members alongside Silicon Valley executives and family offices.

Subsidised Bay Area events — member benefits worth up to $1,000 a year.

One conservative portfolio run through the year, with real-time alerts and real-time allocation adds.

One growth portfolio — the AI trade, the Bitcoin trade and the commodity thesis in detail, with full views and conviction picks.

One-month trial · $99 · cancel any time

Past performance is not a guide to future results and does not guarantee future returns. Performance figures refer to Proflex Finance strategy results and are not audited; individual results vary with entry timing, position sizing and account constraints. Nothing in this note is personal investment advice or a recommendation to buy or sell any security. Proflex Finance and its staff may hold positions in the securities discussed.