Financial Forensic Analysis
Grounded in Jim Chanos' asset-liability and accounting-arbitrage framework. Monitoring the structural imbalances between front-loaded capital outlays and deferred downstream revenue across the 2025–2026 AI infrastructure build-out.
Every figure here is sourced and dated in the reference log → · Working notes →
Memory is now the biggest line in the datacentre bill — a projected 47% of cloud capex this year, 68% nextR84. Margin shows scarcity being priced. Inventory shows whether it is ending. more →
The buildout is justified by the premium that the most capable models command. more →
The capability/cost frontier, September 2026. Capability is the Epoch Capabilities Index; cost is list price per million tokens, blended 3:1 input:output, on a logarithmic axis. more →
The open-weight side moved this month, and it moved onto the line. DeepSeek V4 Flash 0731 scores 154.5 on the index — downloadable, MIT-licensed, and the cheapest thing you can buy at that capability, at $0.094 per million tokensR97R99. The April release of the same model scored 146.1. Same architecture, same size: the gain is post-trainingR98.
The model directly above it costs 4.8 times as much for 1.2% more capability. That is GPT-5.6 Luna at $0.45, and it is where this curve starts to leave the ground.
Above that, September stretched the top without moving the ceiling price. GPT-6 Astra enters at 166.6 and Claude Fable 5.1 at 164.2 — both at the same $20 blended as the models they replaceR93R94. Capability moved; the ceiling price did not.
And to the side sits the odd one. Muse Spark 1.2 scores 155.5 and lists at $2.00 — or $0.125 if you let Meta train on your prompts R91. A sixteenth of its own list price, and it is the amber point on the chart. It is not on the frontier line, because it is not the same transaction.
Correction, 9 September 2026. An earlier version of this panel priced DeepSeek V4 Flash 0731 at DeepSeek’s own $0.66, placed it below the line, and called the cheaper hosted rates an “understatement”. That was wrong. The weights are MIT-licensed and eight independent hosts serve them between $0.09 and $0.18 — the creator was the outlier, not the feed. The panel now prices every model at the cheapest place it can actually be bought, and uses a vendor’s own list price only where that vendor is the only source. On an open-weight model it never is.
| Percentile of capability range | Index | Cheapest available | Step |
|---|---|---|---|
| 0% (floor) | 113.0 | $0.015 | — |
| 25% | 126.4 | $0.028 | 1.9x |
| 50% | 139.8 | $0.055 | 1.9x |
| 75% | 153.2 | $0.094 | 1.7x |
| 90% | 161.2 | $10.000 | 106.6x |
| 100% (ceiling) | 166.6 | $20.000 | 2.0x |
Read the right-hand column. Three quarters of the entire capability range costs under ten cents per million tokens — 1.9x, 1.9x, 1.7x across the whole of it. Then the curve does not bend, it breaks: 106x across the next fifteen percentiles. more →
The composition splits on exactly the same line. All five frontier models below the break are open-weight or Chinese. All five above it are closed and American. Not a tendency — every single one. more →
September moved the top of this chart, and moved it down. Fable 5.1 scores 73.4% at $9.64 a task; Fable 5 scored 70.5% at $17.32R95. A higher score for 44% less money, and every effort level improved on both axes at once — the whole of Fable 5 is now off the frontier. Astra and Muse Spark are not in this benchmark yet, so they do not appear hereR96.
One of these prices has an expiry date printed on it. Gemini 3.7 Flash sits on the frontier at $1.50 blended, but Google's own pricing page states that rate holds only through 31 December 2026, rising to double on 1 January 2027R77. more →
Capability: Epoch AI, "AI Benchmarking Hub", published online at epoch.ai, licensed CC-BY. more →
Both charts above price capability per token, which is the number vendors publish and the wrong number for anyone buying work. more →
The cheapest configuration reaching each score, against measured cost per task. more →
The shape of the first chart survives on measured money. Going from 70.8% to 73.4% — under three points of score — costs 3.4x, from $2.81 to $9.64 a task. more →
It also shows something list prices cannot. The effort dial is the cost dial. Claude Fable 5.1 costs $3.53 a task at medium effort and $9.64 at maximum — a 2.7x range for one model at one published price, which is more than switching vendor buys across most of the range. more →
And it corrects something this page had been assuming. Work back from the measured bill and these tasks turn out to be enormous: a single task consumes at least a million and a half tokens, most of it context re-read across dozens of agent steps — 76 steps for the most expensive configuration.R81
Which means the per-token charts above are not understating the cost of this work — if anything they overstate the rate, because agentic tasks are dominated by input tokens and input is a fifth the price of output. more →
Source: CursorBench, cursor.com/cursorbench, and ARC-AGI-2, both via Epoch AI, "AI Benchmarking Hub", epoch.ai (licensed CC-BY). more →
The panel above prices capability against an index. more →
The cheapest model able to sustain a task of each length, against blended list price per million tokens. more →
| Task length | Cheapest model that can sustain it | Step |
|---|---|---|
| 15 minutes | $0.100 | — |
| 30 minutes | $0.100 | 1.0x |
| 1 hour | $1.925 | 19.2x |
| 2 hours | $3.438 | 1.8x |
| 4 hours | $4.500 | 1.3x |
| 8 hours | $10.000 | 2.2x |
Crossing from half-hour to hour-long tasks costs 19.2x. Going from one hour to eight costs about five times in total. Once a model can sustain an hour of autonomous work, extending that horizon is comparatively cheap. more →
Now the part that matters more than the price curve. Everything above is measured at a 50% success rate — a coin flip. more →
| Model | 50% horizon | 80% horizon | $/1M |
|---|---|---|---|
| Claude Opus 4.6 | 12.0h | 1.17h | $10.000 |
| Gemini 3.1 Pro | 6.4h | 1.50h | $4.500 |
| GPT-5 | 3.4h | 0.64h | $3.438 |
The twelve-hour model is a seventy-minute model at a threshold you would rely on. And the ranking inverts: at 80%, the most expensive model on the chart drops off the frontier entirely, beaten on horizon by one costing less than half as much. more →
Capability: METR, "Measuring AI Ability to Complete Long Tasks" (arXiv:2503.14499) and "Task-Completion Time Horizons of Frontier AI Models" (Time Horizon 1.1), metr.org/time-horizons. more →
Astra, Fable 5.1 and Muse Spark are not on this chart. METR has not published a time horizon for any of them yetR96. They appear on the capability and cost-per-task panels above, where the measurements exist.
Capex, guidance, and whether operations still cover it.
← swipe to see all columns →
| Company | Q2 CapEx | FY26 Guide | Free Cash Flow | The tell | Tape |
|---|---|---|---|---|---|
| Alphabet GOOGL | $44.9B | $195–205B ↑ | −$5.9B | First negative FCF as a public company. Coverage 0.87×.R45 | −7% |
| Microsoft MSFT | $41.0B | ~$175B ↓ headline only | +$19.6B | Two accounting levers at once: building lives 15→25yr, and ~$15B of leases reclassified out of headline capex with no change to spend.R31R37 | +8% |
| Meta META | $31.1B ×1.8 YoY | $130–145B floor ↑ | $784M | Ad impressions decelerated to +14% (from 19%, 18%). Revenue carried by price, not volume.R41R42 | −7% |
| Amazon AMZN | $53.1B net PP&E | $220B ↑ on memory | −$7.6B TTM | Buildout now partly debt-funded. Offsetting: AWS +36.7%, margin +650bps, backlog $496B.R31aR33R34 | +8% |
The tell this quarter is not the size of the spend — it is that operating cash flow stopped covering it. Two of the big four are free-cash-flow negative in the same quarter, and Microsoft's relief rally came from holding the envelope flat while extending useful livesR39. more →
Meta is the exception that qualifies the rest. At Amazon, Microsoft and Alphabet the financial leg deteriorated while demand accelerated. more →
An accounting story becomes a credit story at the point where the marginal dollar has to be borrowed. That point has passed: Amazon confirmed debt issuanceR33, the FOMC held on a 9–3 vote with three dissents for a hike, and the 30-year pushed past 5.19%R38.
Related-Party Disclosure
The claim is that chip vendors and hyperscalers fund their own customers, so some reported revenue is the seller's own money returning. more →
The structures are built to stay non-voting and below the significant-influence threshold, so that ASC 850 related-party disclosure never fires. more →
The three peers holding comparable positions disclose none of itR21. more →
Oracle's $300B OpenAI contract and AMD's ~$90B could not be traced to any filing by either party. more →
Rule
In this market, the larger the headline number, the less likely any issuer has ever said it.
The $0.01 warrant is an industry instrument, not a one-off: AMD→OpenAI, AMD→Meta, Google→TeraWulf, Google→Cipher, CoreWeave→Core Scientific. more →
The margin test came back clean — AWS margin expanded ~645bps while growth accelerated, Google Cloud margin roughly doubledR28. more →
Method. Every figure is quoted verbatim or marked as inference. more →
The forensic question is not how much the hyperscalers spend — it is whether the business still generates the cash to pay for it. more →
Figure 4 — Operating cash flow as a multiple of capital expenditure, trailing twelve months. more →
| Company | Coverage, 2022 | Coverage, latest | TTM Op. Cash Flow | TTM CapEx | TTM Free Cash Flow | Period end |
|---|---|---|---|---|---|---|
| Microsoft MSFT | 3.71x | 1.58x | $182.9B | $115.9B | $67.0B | 2026-06-30 |
| Alphabet GOOGL | 3.42x | 1.40x | $185.7B | $132.4B | $53.3B | 2026-06-30 |
| Amazon AMZN | 0.61x | 0.98x | $148.5B | $151.0B | −$2.5B | 2026-03-31 |
| Meta META | 3.00x | 1.64x | $124.0B | $75.7B | $48.3B | 2026-03-31 |
| Oracle ORCL | 2.12x | 0.57x | $32.0B | $55.7B | −$23.7B | 2026-05-31 |
Source: SEC EDGAR XBRL company filings. Most quarters are derived by differencing year-to-date cumulative filings, since few issuers tag discrete quarterly cash-flow figures. more →
Current values against bubble-top thresholds. Hover any metric for what it means. more →
← swipe to see all columns →
| Monitoring Metric | Underlying Vulnerability | Current Live Metric | Thresholds | Status |
|---|---|---|---|---|
|
CIP to Net PP&E Ratio
Deferred Depreciation Shield
Construction-in-Progress represents capital spent on data centers not yet active. more → |
Accounting shield via deferred depreciation. Reads a floor since Aug 2026 — SPE-financed compute never enters this ratio.R18 | 28.4%R70 floor · computed | < 15% (Historical mean, below is acceptable) | Impaired |
|
Long-End Financing Cost
Accounting Story → Credit Story
The build-out has crossed from internally-funded to externally-funded. more → |
External funding meets a rising cost of capital. more → | 30Y 5.19% · 5Y CDS ~75bp | 30Y sustained > 5.5%, or IG tech spreads +75bp from trough | Elevated |
|
Ad-Engine Volume Growth
Spending More to Sell Less
For ad-funded capex (Meta, Alphabet), impression volume is the demand signal underneath the revenue line. more → |
Demand-side crack beneath a price-carried revenue print | Meta +14% (from 19%) | Volume growth −5pp over two quarters while capex guidance rises | Triggered |
|
Tech Buyback Volumes
Capex Cannibalization Risk
When massive capital expenditure requirements eat into operating cash flow, companies must slow discretionary share buybacks to protect liquid reserves. |
Free cash flow exhaustion under capex strain | GOOGL $0 big tech −17% YoYR66 |
> 0% (Organic growth, above is expected) | Breached |
|
Operating Cash Flow CapEx Coverage
The Self-Funding Test
Operating cash flow divided by capital expenditure. Above 1.0x the build-out is paid for out of the business; below 1.0x it must be funded from cash reserves, debt, or leases. |
Build-out no longer self-funded — shift to debt & lease financing | 0.87× (GOOGL) | > 1.0× (Self-funded, below requires external capital) | Breached |
|
GPU Spot Leasing Prices
Organic Demand Reality Check
While primary contracts mask real demand, real-time hourly secondary rental prices reflect true industry utilization. |
Secondary capacity oversupply. Blended index invalid — prior gen collapsing while current gen is rationed; the read is the spread.R55 | H100 −64–75% Blackwell/Rubin rationed |
Per generation. Widening current-vs-prior spread = obsolescence | Spread widening |
|
Grid Utility Lead Times
The Power Lead-Time Paradox
High lead times strand capital: Completed facilities can't get power. more → Rapid decreases trigger crash: Easing bottlenecks allows massive backlogged compute online instantly, flooding the market and collapsing leasing margins. |
Physical transmission bottlenecks & stranded asset risk | 24–72 mo 5–7 yr where constrainedR69 |
< 30 mo = pricing pressure · > 30 mo = stranded capital | Strained |
|
Neo-Cloud Interest Burden
Replaces the utilisation read
This monitor originally watched utilisation and lease rates for signs the leveraged intermediaries could not service their debt. more → |
Debt service outrunning rental incomeR9 | Loss widening on 2× revenue | Interest/revenue rising across quarters, or widening adj-EBITDA-to-GAAP gap, while revenue grows | Triggered |
|
Inference Efficiency Transfer
Re-pointed, not re-scaled
The original instrument assumed an efficiency breakthrough would arrive from outside and damage Nvidia. more → |
Obsolescence risk assigned to cohorts 2–3 and SPE creditorsR49 | Realised $/token vs $45/M claim | Falling realised price = deflation thesis · rising = Nvidia thesis | Unresolved |
Macroeconomic Precedents
The current expansion mirrors the speculative patterns of the 1998–2000 telecommunications build-out and the 2005–2007 subprime credit expansion. more →
Figure 1 — Comparative infrastructure spending. A single hyperscaler's 2026 guide is already 2× the entire 1998–2002 Dot-Com telecom build-out; the big four combined are roughly 8×.
Industry Cohort Risk Matrix
Five distinct layers present unique balance sheet exposures — from CIP deferrals in hyperscalers to heavy leverage in neo-cloud intermediaries.
MSFT · GOOGL · AMZN · META
CRWV · NBIS · Fluidstack
EQIX · DLR · CORZ
GEV · Siemens · Bloom Energy
NVDA · TSMC · AMD · MU
By early 2026, Construction in Progress (CIP) balances have reached unprecedented levels. more →
CIP balance by entity — the deferred depreciation wave.
Capital Efficiency
Incremental Return on Invested Capital evaluates the profitability generated by each additional dollar of capital. A sustained decline toward 10% indicates that returns on new hardware may no longer cover the weighted average cost of capital.
Microsoft's cumulative incremental ROIC across the AI capex era (FY2022–FY2026) is 24.5%, against 51.2% for the pre-AI era (FY2019–FY2022)R61 — roughly a halving. This is a monitor computation on a consolidated basis from SEC XBRL filings, not a reported figure and not a segment figure: Microsoft discloses segment operating income but not segment invested capital, so no segment-level iROIC is checkable from disclosure. The pre-AI window starts at FY2019 to avoid an invested-capital tagging discontinuity in FY2017–FY2018.R70
Return Compression
All four hyperscalers earn a lower return on invested capital than they did at the start of 2024. more →
Pretax ROIC, annualised from the quarter.R74 Q4 is absent by construction — 10-Ks report full-year durations. more →
Assets under construction sit in the denominator earning nothing, so a falling return could be timing rather than decline. more →
Excluding it lifts the level by roughly 10–12 points — but does not flatten the decline. more →
Alphabet’s purchase commitments and other contractual obligations went from $332.4bn to $811.0bn in one quarter. more →
Read from the 10-Q text. This figure is not a tagged XBRL value — the concept API returns an unrelated $7.7bn.R73
The transition from speculative expansion to normalization is inevitable. more →
"When the supply of new issuance exceeds institutional demand, the market becomes highly vulnerable to a sharp valuation adjustment."R56 As of August 2026 the harder problem is that CIP and reported capex have both been engineered downward — the monitoring has to follow the risk into the credit markets.R18