★ BREAKING FEB 28 2026 /// OPENAI RAISES $110B · $730B VALUATION /// SOFTBANK $30B · AMAZON $50B · NVIDIA $30B /// CASH BASIS UPDATED TO $150B /// AMAZON TRAINIUM CHIP DEAL · FRONTIER ENTERPRISE OFFERING /// CODEX 1.6M WEEKLY USERS · +37% IN 7 DAYS /// IPO TARGETED Q4 2026 /// ALTMAN: "VERY LONG RUNWAY TO EXPAND" /// SCENARIO ANALYSIS · UPDATED /// OPENAI SYSTEMIC RISK /// CONTAGION MAPPING /// TS IMAGINE RESEARCH ///
TS Imagine Research  ·  Systemic Risk Series  ·  March 2026

What If
OpenAI
Blew Up?

A risk manager's guide to the unthinkable — contagion chains, default cascades, and how to pick up the pieces when the world's most valuable AI infrastructure goes dark. Published as OpenAI closes a $110B round at $730B valuation. The raise lengthens the runway. It does not change the analysis.
Published March 2026
Timing Published day of $110B close
Classification Scenario Analysis
Series Systemic Risk
FOR INSTITUTIONAL USE ONLY · NOT INVESTMENT ADVICE · FEBRUARY 28 2026
★ TS Imagine Research · March 2026 · The Numbers Behind the Headlines
Cash Basis
$150B
Was $150B. $110B raise closed Feb 27. Amazon $35B tranche is milestone-contingent — effective immediate liquidity ~$115B.
Scenario A Probability
14%
Was 22%. 7+ year theoretical runway at current burn substantially reduces near-term funding gap probability. Longer tail risk remains.
New Concentration Risk
Amazon
Trainium chip deal creates AWS dependency replacing Nvidia/Azure. Frontier enterprise offering is unproven at scale. Supply chain pivot takes years.
Revenue Signal
+37%
Codex weekly users 1.6M, up 37% in 7 days. Enterprise growth validating the revenue thesis — but accelerating compute demand simultaneously.
TS Imagine take: The $110B raise is the largest private funding round in history. It changes OpenAI's near-term survival calculus. It does not change the structural argument. The failure scenarios this paper maps are 3–7 year stories — and the post-raise numbers make them sharper, not softer: $74B in projected operating losses in 2028 alone (Deutsche Bank), a $207B funding shortfall by 2030 (HSBC), and circular financing structures that Reuters flagged as exacerbating Wall Street's deepest concerns. The raise moves the clock. It doesn't stop it.
Intelligence
OpenAI closes $110B round — valuation $730B pre-money. Scenarios modeled Feb 27.
SoftBank $30B · Amazon $50B ($15B immediate + $35B milestone-contingent on Trainium chip deal for new Frontier enterprise offering — a pivot from Nvidia/Azure) · Nvidia $30B · Cash on hand $150B ($110B + ~$40B prior). Altman on CNBC: "We have a very long runway."  ·  Codex users 1.6M, +37% in 7 days  ·  IPO targeted Q4 2026  ·  Anthropic raised $30B at $380B earlier this month. The raise extends the runway. It does not resolve the structural questions below.
TS Imagine Research
March 2026
Day-of publish
TS Imagine · RiskSmart
So What For Risk Managers
This deal is a masterclass in the kind of interconnected counterparty exposure
that RiskSmart helps clients model.

The circular financing structures, the conditional tranches, the cloud commitment layering — $138 billion in AWS commitments alone — represent exactly the kind of concentrated, correlated exposure that institutional risk desks need to be stress-testing. Single-vendor dependency on a company that requires continuous capital infusion to survive is not a standard vendor risk category. It is a systemic exposure, and most enterprise risk frameworks are not yet equipped to model it.

Circular financing structures
Conditional milestone tranches
Cloud commitment layering
$138B AWS exposure
Correlated counterparty risk

This is the first of our Systemic Risk Series. The interactive dashboards below — the contagion ring chart, the VaR stress test, the chain reaction table — are prototypes of how a RiskSmart AI Concentration Risk module works in this context. The question is not whether OpenAI is too big to fail. The question is whether your risk desk has modeled what happens if it does.

01 —

The Deeper
You Go

"Never get out of the boat."

Apocalypse Now, 1979 (Francis Ford Coppola, adapted from Conrad's Heart of Darkness)

Nobody decided to make OpenAI critical infrastructure. It happened one API call at a time. A workflow here, a product feature there, a board deck built on GPT outputs, a hospital intake form, a legal brief, a trading signal. Each step seemed reasonable. Each step went deeper. The shore receded without anyone noticing — because when the river is moving smoothly, nobody checks how far they've come.

Today OpenAI closed a $110 billion funding round — SoftBank $30B, Amazon $50B, Nvidia $30B — at a $730 billion pre-money valuation. Combined with ~$40 billion already on hand, Sam Altman now has approximately $150 billion in cash and a "very long runway." The headlines are writing themselves. This paper is about what comes next, and what the numbers actually say when you read them carefully.

The raise is real. The runway is real. What it doesn't change is the fundamental race OpenAI is now running: two races running in opposite directions, neither of which stops because a funding round closed. The burn race — how fast cash depletes against a cost structure that has no peer in corporate history. And the competition race — how fast Google, Anthropic, and open-source models erode the market position that justifies raising the next round. These clocks are not independent. Every move that addresses one accelerates the other.

Deutsche Bank: "No startup in history has operated with losses on anything approaching this scale. We are firmly in uncharted territory." In H1 2025, OpenAI generated $4.3 billion in revenue and lost $13.5 billion — $3.14 for every dollar earned, burning $575,000 per hour around the clock. Through 2029, it projects $115–143 billion in cumulative negative free cash flow. The $150B raise extends the runway. It does not bend the cost curve.

"The question is no longer whether OpenAI survives the next 18 months. It is whether OpenAI can fundamentally restructure its economics before the next raise — against competitors who are structurally cheaper and getting faster. $150B buys time. It doesn't buy the answer."

And underneath all of it: 92% of Fortune 500 companies now run production workflows on OpenAI infrastructure. Seven more years of runway means seven more years of deepening dependency across the global enterprise stack. The deeper you go, the larger the blast radius becomes — not if things go wrong immediately, but if things go wrong at the end of a very long river.

This paper maps both the race and the blast radius.

H1 2025 Net Loss
$13.5B
On $4.3B revenue · $3.14 per $1 earned
Hourly Cash Burn
$575K
Every hour, around the clock
★ Today's Raise
$110B
SoftBank $30B · Amazon $50B · Nvidia $30B
Post-Round Valuation
$730B
Pre-money · $150B total cash ($110B raise + ~$40B prior)
Theoretical Runway
7+ yrs
At current net burn · before cost curve bends
Fortune 500 Exposed
92%
Production workflows on OpenAI infrastructure
OpenAI's projected cumulative cash outflows through 2029 dwarf the losses of Uber ($18B over six years), Amazon's early years (~$1B), and Tesla's pre-profit period (~$9B). The scale is categorically different. Today's $110B raise, added to ~$40B already on hand, totals approximately $150B in cash — the largest single private funding round in history. It is also, against this burn trajectory, a finite number.
Day-One Intelligence —

How the Market Is Reading This

How the Market Read This on Day 1 · Sources cited inline
WAU · PYMNTS
900M
Weekly active users. 50M paying consumers, 9M business subscribers.
2028 Op. Loss · DB
$74B
Projected operating loss in 2028 alone. Deutsche Bank / Jim Reid.
2030 Gap · HSBC
$207B
HSBC-projected shortfall between revenue and spend through 2030.
Amazon Conditional
$35B
Of $50B is contingent on API adoption milestone or IPO. A third of the headline is conditional.
Rev Target · CNBC
$280B
OpenAI 2030 projection. Split ~evenly consumer and enterprise.
Premarket Feb 28
NVDA ▼ FELL Market skeptical of $30B check into a loss-generating entity
AMZN ▼ FELL $50B into a company projecting losses through 2028
MSFT DID NOT PARTICIPATE Retains option to join. TechCrunch / Axios.
The Bull Case
Andy Jassy · Amazon CEO

OpenAI will be one of the "expected long-term winners in AI." Corporate backers are publicly all-in — the largest are betting on dominance, not just survival.

Owen Lamont · Acadian / Fortune

Technically "not in a bubble yet" — his framework requires equity issuance (IPOs) as the fourth horseman, which hasn't materialized. Goldman Sachs agrees: the current AI run is underpinned by real earnings growth, unlike dotcom.

Revenue Scale · PYMNTS / CNBC

900M weekly active users. 50M paying consumers, 9M business users. $280B+ projected 2030 revenue split consumer/enterprise. The scale of the user base is real, even if the path to profitability is contested.

The Bear Case
INSEAD Faculty + Reuters / Yahoo Finance

Circular financing at scale. Nvidia invests in OpenAI → OpenAI buys Nvidia chips → money moves in circles. Reuters: the round "exacerbates Wall Street concerns about circular financing agreements where firms invest in and sign supply deals with each other, inflating demand and revenue." INSEAD flags dotcom-era vendor financing echoes at troubling scale.

Deutsche Bank · Jim Reid + HSBC / Fortune

The burn math is brutal. $74 billion in operating losses projected for 2028 alone. HSBC projects a $207 billion funding shortfall by 2030 — the gap between what OpenAI generates and what it needs to spend. Today's $150B in total cash covers less than half of that gap.

Altman / The Verge · Aug 2025 + Apollo / CNBC

In August 2025, Altman himself told The Verge the AI market is in a bubble. Apollo's Torsten Slok argued the current AI bubble is bigger than the internet bubble. Neither has retracted. Today's raise didn't address the structural question — it pushed the reckoning further out.

Newcomer Newsletter — Sharpest Take of the Day

The answer to "how are you going to pay for all this?" turned out to be the same one it always has been — convince enough CEOs and rich people you're going to change the world and get them to write very large checks. Also: Stargate has turned out to be mostly bilateral deals rather than a real coordinated entity. The coherent joint venture it was announced as doesn't appear to exist.

TS Imagine Research Synthesis

The bull case is real: 900 million weekly users, $280B revenue projections, and backers who have staked corporate reputations on the outcome. The bear case is also real: $74B in projected 2028 losses, a $207B funding gap by 2030, and circular financing structures that make it difficult to distinguish genuine demand from money moving in circles. Both can be simultaneously true — and that coexistence is precisely the systemic risk this paper maps. A company that is strategically important, competitively contested, and structurally cash-negative at scale is not a standard vendor risk. It is a new risk category. One that most institutional risk frameworks are not yet equipped to model.

02 —

Two Clocks,
One Race

The $110B raise reframes the risk — OpenAI already had ~$40B on hand; this is not a $150B war chest conjured from nothing. This is no longer primarily a story about whether OpenAI can make payroll in 2027. It is a story about whether OpenAI can win two simultaneous races — against its own cost structure, and against competitors who are structurally cheaper and accelerating. The complication: these races are not independent. The moves required to win one actively impede winning the other.

Scenario A
The Burn Clock: $150B Is a Finite Number ($110B raise + ~$40B prior)
At current net burn — approximately $27B annually after revenue — $150B buys roughly 5–6 years of operational runway, not 7+. The "7+ years" figure assumes burn stays flat. It won't. Codex weekly users hit 1.6M, up 37% in the prior week alone. Every user added is revenue validation and compute cost simultaneously. Inference costs scale with usage. The faster revenue grows, the faster the denominator shrinks — unless the cost structure changes underneath it.

The Amazon Trainium pivot embedded in the raise is the bet on changing the cost structure. OpenAI will purchase Trainium chips for its new Frontier enterprise offering, pivoting away from Nvidia GPUs and Microsoft Azure compute. This is the right strategic move. It is also a multi-year capital and engineering program, unproven at scale, executed while running full-speed in production. The $35B Amazon tranche is milestone-contingent — if those milestones slip, effective immediate liquidity is closer to $115B. The burn clock doesn't pause for infrastructure transitions.
Funding Gap Risk (5-Year) · post-raise revision 14%
Scenario B
The Competition Clock: The Moat Is Not as Wide as Advertised
While OpenAI executes its infrastructure pivot, Google is not standing still. Gemini reached 650 million monthly active users in Q3 2025. ChatGPT's US traffic declined 35% in November 2025. Menlo Ventures data shows OpenAI's enterprise LLM share has already fallen from 50% in 2023 to 27% in 2025 as Anthropic took the enterprise lead. Meta's open-source Llama gives any company with engineering capacity a path to zero marginal model cost.

The competitive threat is not that a rival builds a better model. It is that the premium pricing OpenAI commands — the pricing that makes the revenue thesis work — erodes as alternatives become good enough. "Good enough" does not have to mean better. It means cheaper, integrated, and already in the stack. Google has all three advantages simultaneously.
Competition Collapse Risk (5-Year) 28%
Scenario C
Why Solving One Clock Accelerates the Other
The Trainium pivot reduces compute costs — but it requires years of capex, engineering distraction, and partnership risk that consumes resources needed for model research and product. Cutting burn means cutting the investment that maintains competitive position. Maintaining competitive position means sustaining the burn that depletes the runway. This is not a false dilemma. It is the actual structure of the problem.

The scenario risk managers need to model is not sudden death. It is a slow-motion squeeze: OpenAI executes reasonably well on both clocks, makes genuine progress, and arrives at its next major raise in 2030–2031 having burned through most of the $150B — without having achieved the cost structure or market dominance to justify a follow-on at a sane valuation. Not a collapse. A compression. The kind that forces terms nobody wants to accept.
Squeeze Scenario Risk (7-Year) 31%
Scenario D
Governance Failure: The Board Fires Its Last Bullet
The governance fragility has not improved with the raise. OpenAI's board has already demonstrated — in November 2023 — that it cannot execute a leadership change without triggering a near-total operational crisis. The nonprofit-to-for-profit conversion, the complex web of investor rights accumulated across this and prior rounds, and the concentration of institutional trust in a single founder compound this fragility. A governance crisis that coincides with market stress would remove the Microsoft rescue option at precisely the moment it is most needed.
Governance Crisis Risk (5-Year) 15%
Evidence File
Documented Instances Where the Official Narrative Collapsed
Mask Off Moments:
When the Stated Story and the Primary Sources Diverged
In risk analysis, the most instructive data points are not the statements companies make about themselves — it's the moments where unscripted disclosure, testimony, or on-camera speech reveals a materially different picture. OpenAI has produced two such moments that are directly relevant to governance risk. Both are documented in primary source video or sworn testimony. Both were subsequently walked back. Neither walkback erased the underlying disclosure.
Incident 01 — Finance
The CFO Defines the Business Model at the Wrong Moment
WSJ Tech Live · October 2025 · Sarah Friar, CFO
OpenAI CFO Sarah Friar, speaking on a live panel, articulated what OpenAI's financing model actually requires. In response to a question about government support, she described — in specific technical terms — the need for a federal loan guarantee to make the company's infrastructure commitments financeable.
"The backstop, the guarantee that allows the financing to happen. That can really drop the cost of the financing, but also increase the loan to value, so the amount of debt that you can take on top of an equity portion."
— Sarah Friar, CFO · WSJ Tech Live. The WSJ interviewer followed with: "So some federal backstop for chip investment." Friar: "Exactly."
The walkback came within hours. OpenAI's newsroom issued a statement saying Friar "was making the point that American strength in technology will come from building real industrial capacity." The video clip remains available. The transcript is unambiguous.
Risk implication: A CFO describing the need for federal loan guarantees to make the company's debt math work is not describing a conventional growth-stage financing situation. It is describing a capital structure that requires a backstop that does not currently exist — and whose absence is, by the CFO's own account, the constraint on the financing. The Delaware AG opened review within days. The walkback does not undo the constraint. Today's raise addresses the near-term liquidity. It does not address the long-run infrastructure financing question Friar was describing.
Incident 02 — Governance
The Architect of the Board Coup Explains What It Was Actually About
Congressional Testimony · 2025 · Ilya Sutskever, former Chief Scientist
The official framing of the November 2023 board crisis — in which OpenAI's board fired Sam Altman, only for him to be reinstated within 96 hours after a near-total employee revolt — attributed the action to AI safety concerns and the nonprofit's fiduciary mission. Sutskever, widely reported as the central figure in the ouster, subsequently provided congressional testimony that characterized the episode differently.
Sutskever's testimony described the board crisis as centrally a personality conflict and concerns about Altman's management style and candor — not a principled AI safety intervention. Attempts to frame the events as driven by effective altruism or existential risk concerns were, per the testimony, "almost entirely scapegoating."
— Ilya Sutskever · Congressional testimony, 2025. Reported by independent AI analyst Zvi Mowshowitz with direct citations to the testimony record.
This is the board's own architect, under oath, describing the action as a management dispute rather than a principled safety intervention. The board fired its only real bullet — its power to remove the CEO — missed, and has never credibly threatened to do so again. The institutional authority that OpenAI presents as its safety override mechanism has been empirically tested and found inoperative.
Risk implication: If the board's 2023 action was primarily a personality dispute dressed as a governance intervention, it means: (a) the safety-mission framing of board authority is substantially performative; (b) the board proved unable to execute even a management change without triggering a collapse; and (c) the governance structure OpenAI presents as its principal safeguard has been load-tested to failure and documented as such under oath.
03 —

$3.14 to Make
a Dollar

The most important structural analysis of OpenAI's position doesn't appear in any startup narrative. It appears in the mathematics of comparative cost structure — specifically, the permanent, compounding gap between what it costs OpenAI to deliver a unit of intelligence and what it costs Google to deliver the same unit. This gap has a name: margin structure inversion. And unlike most competitive disadvantages, it doesn't close with scale. It widens.

OpenAI Cost Per $1.00 Revenue
$3.14
H1 2025: $13.5B loss on $4.3B revenue · $13.5 ÷ $4.3 = $3.14
Google Gemini Cost Estimate
~$0.60
TPU vertical integration — zero margin stacking
Google TPU Cost Advantage
30–44%
Lower TCO vs NVIDIA GB200 (SemiAnalysis, Nov 2025)
Advantage Growth Direction
Widens
Every query multiplies the structural gap
Infrastructure Cost Stack · Per 1M API Tokens
OpenAI / ChatGPT ~$18.40
Triple-stacked margins: OpenAI pays NVIDIA's margin (~75% GM on chips), then Microsoft's margin on Azure hosting, then its own operating costs on top. Every query compounds all three.
NVIDIA GPU Margin (~75% GM)
$6.20
$6.20
Microsoft Azure Margin
$4.80
$4.80
OpenAI Infrastructure Ops
$4.60
$4.60
R&D / Training Cost Share
$2.80
$2.80
Google / Gemini ~$5.80
NVIDIA MARGIN: $0 Google trains Gemini entirely on its own Tensor Processing Units. No chip vendor margin. No Azure margin. Direct electricity and silicon depreciation only.
Electricity + Cooling
$1.40
$1.40
Silicon Depreciation (TPU)
$2.20
$2.20
Google Ops Cost Share
$2.20
$2.20
The ~3× Cost Gap
Google can price Gemini API access at levels OpenAI cannot match without accelerating losses. It can offer free tiers OpenAI cannot subsidize. It can bundle AI into Workspace at costs that make standalone ChatGPT feel like a premium product. It is doing exactly this.

SemiAnalysis, the semiconductor research firm, quantified the gap in November 2025: Google's TPU infrastructure delivers 30% lower total cost of ownership than NVIDIA's GB200, and 44% lower from Google's internal perspective when accounting for full three-dimensional torus configuration. This is not a gap that closes with scale. It is a gap that widens with scale — because every additional query OpenAI serves multiplies the margin disadvantage.

⧗ Scheduled for distribution · Part II · Microsoft Exposure Deep-Dive

Two independent data points; same picture. Data point one — P&L: H1 2025 financials show $13.5B in losses on $4.3B revenue. That is $3.14 in costs for every $1.00 earned — the figure this section leads with. Data point two — Microsoft filings: Microsoft's Q1 FY26 10-Q disclosed a $3.1B quarterly net income decrease from its equity method investment in OpenAI. Microsoft holds a 27% diluted stake. Backing out the full loss: $3.1B ÷ 0.27 = approximately $11.5B in quarterly losses, or ~$46B annualized. Against ~$20B in annualized revenue that implies a cost ratio closer to $3.30 — consistent with H1 2025 directionally, somewhat higher, likely reflecting continued burn acceleration in H2 2025. The two figures are not contradictory. They are from different periods and should not be cited in the same sentence as if they confirm each other precisely.

The competitive consequence is already visible. Gemini reached 650 million monthly active users in Q3 2025. ChatGPT's US traffic declined 35% in November 2025. Marc Benioff, the Salesforce CEO who used ChatGPT daily for three years, publicly posted he wasn't going back — not because of capability, but because of the inexorable logic of cost structures expressing themselves through product decisions that even CEOs can feel without being able to articulate.

This is the mechanism the consensus has not modeled: OpenAI's cost disadvantage is permanent, not temporary. It does not resolve with the next fundraising round. It does not resolve with GPT-5. It resolves only if OpenAI builds its own silicon — which requires a capital expenditure timeline measured in years at a scale that makes its current burn look modest by comparison. Or if a buyer with its own silicon absorbs it.

04 —

The Dependency
Footprint

Enterprise Customers (API)
80K+
Azure OpenAI Service alone
Weekly Active Users
800M
As of Oct 2025
Fortune 500 Using GPT
92%
Products or API
SaaS Platforms w/ GPT Embedded
6,800+
Commercial integrations

The dependency problem is not simply about who uses OpenAI. It's about how they use it — and whether the integration is surface-level or load-bearing. A company that uses ChatGPT for employee brainstorming is in a very different position from one that has built an automated underwriting workflow, a medical triage decision tree, or a real-time trading signal generator on top of the GPT-4 API.

The latter category is larger than comfortable. Across finance, healthcare, legal, and technology, enterprises have spent the better part of 18 months converting OpenAI's models from experimental tools into production infrastructure. The OpenAI enterprise report published in December 2025 quantified this shift: ChatGPT Enterprise message volume grew 8x year-over-year, while structured workflow usage (Projects, Custom GPTs) increased 19x. The models aren't just being used — they're being operationalized.

Microsoft (MSFT)
$13B+
SoftBank Group
$30B*
Thrive Capital
~$1.3B
Sequoia Capital
~$1B+
a16z / T. Rowe
~$0.8B
*SoftBank round partially conditional on for-profit conversion. Only $10B wired at time of reporting.
Counterparty Class Exposure Type Severity
Microsoft $13B equity, Azure revenue dependency, M365 Copilot product line, 27% ownership stake Critical
SoftBank Vision Fund $30B committed (Feb 28 2026 round); marks down to near-zero on failure; LP pressure cascades Critical
AI-Native Startups 35% of top-funded AI startups list OpenAI as foundational provider; API unavailability = product unavailability High
Healthcare Systems Clinical decision support, triage, documentation workflows now in production; regulatory complexity on replacement High
Financial Services Risk model inputs, document analysis, compliance workflows; largest enterprise AI scale per Menlo Ventures High
Legal / Professional Svcs Document review, drafting workflows; $650M+ legal AI market built on LLM infrastructure Medium
Salesforce Ecosystem GPT-powered CRM integrations in 11,000+ companies via Einstein Medium
Adobe / Canva / Notion Multi-year partnership renewals; product features dependent on API continuity Medium
14 National Governments Civic services and public education licensing agreements Medium
NVDA / Cloud Infra Players Revenue exposure from $1.4T committed datacenter spend; construction contracts at risk Medium

We've spent a decade debating whether AI would replace us. The more immediate question is what happens when the AI we've made indispensable is suddenly unavailable.

05 —

The Risk Manager's
Playbook

This paper is not investment advice, and it's not predicting OpenAI's failure. It is arguing that the failure scenario is underweighted in current risk frameworks — that the combination of unprecedented financial losses, complex governance, deep operational dependencies, and a political environment hungry for AI accountability creates a tail risk worth explicitly modeling. Here's how to start.

1
Map Your AI Vendor Concentration
The first step is embarrassingly simple and widely skipped: catalog every production workflow that depends on an external AI API. For most organizations, this inventory doesn't exist at the level of granularity needed. "We use ChatGPT" is not a vendor risk assessment. You need to know which workflows, which models, which call volumes, and — crucially — which workflows have no manual fallback. Those are your critical exposures.
2
Pressure-Test Your Switching Costs
The substitutability argument — "we can just switch to Anthropic or Google" — sounds more convincing than it is. API format differences, fine-tuned model behavior, prompt engineering investments, and output format dependencies all create real switching friction. Anthropic raised $30B at a $380B valuation in early February 2026; it is the most credible fallback for enterprise customers, but it is not a frictionless migration and its own capital structure carries analogous dependency questions. Quantify what a 30-day forced migration to an alternative provider actually costs. Then ask whether your vendor contracts provide any rights or remedies in a failure scenario. Most enterprise AI contracts are silent on this.
3
Monitor the Financial Signals
OpenAI is private, which limits public financial monitoring — but not entirely. Microsoft's quarterly disclosures contain OpenAI-related line items. Funding round activity and terms signal investor confidence. Secondary market valuations provide a real-time proxy for market assessment of failure probability. Prediction market platforms now offer contracts on AI company events. These instruments are worth watching as leading indicators, not lagging confirmations.
4
Build for Disruption, Not Just Migration
The gold standard is an AI integration architecture that can route to multiple providers at the infrastructure level — not a migration plan, but a live multi-provider configuration. This increases complexity and cost but dramatically reduces vendor concentration risk. For critical workflows, it's the difference between a business continuity event and an operational catastrophe. The cost of this architecture is knowable. The cost of not having it, in a failure scenario, is not.
5
Integrate AI Vendor Risk Into Your Risk Framework
AI vendor concentration belongs in the same risk framework as data center concentration, cloud provider concentration, or any other operational dependency that, if disrupted, would materially affect business continuity. The tools exist: scenario analysis, probability weighting, impact quantification, mitigation tracking. The gap is not methodology — it's that most risk frameworks haven't caught up to how deeply AI has been embedded in production workflows in the last 24 months. Close that gap before the regulators close it for you.
Interactive Tools —

Model, Stress-Test, Visualize

Hover ring for detail · Click to lock
Exposure Layer Map
Selected Layer
Hover or click a ring to see layer details.
Total Ecosystem Blast Radius
$843–1,240B
Across all contagion rings · 5-year horizon
09 —

The Capital Structure
Nobody Drew

OpenAI doesn't have a conventional debt stack. No public bonds, no syndicated credit facility, no publicly rated paper. This is the fact the market tends to lead with when dismissing debt-related contagion risk — and it's technically accurate and substantially misleading. The true obligation load is best understood not as a capital structure in the traditional sense, but as a series of instruments that are economically equivalent to debt, structured to appear as something else.

Assembled in one place — which no single disclosure does — the picture looks considerably different from the equity-round narrative.

Azure Compute Payables EFFECTIVELY SENIOR Senior Operational
Ongoing compute invoices from Microsoft Azure — must be paid to keep models running. Ceasing payment = ceasing operations. ~$2B+ annual run rate.
$2B+/yr
Microsoft Revenue Share SENIOR CONTRACTUAL CLAIM Contractual Senior
20% of total revenue through 2032, renegotiated Oct 2025. Deferred tranches weighted to later years — a creditor-style claim on future cash flows in partnership clothing. >$13B projected 2026–27.
$13B+
Infrastructure Commitments TAKE-OR-PAY OBLIGATIONS Quasi-Senior
$1.4T in committed datacenter, power, and GPU capacity agreements over 8 years. Contractual obligations to hyperscalers, construction firms, and power companies. No hard asset collateral backing OpenAI's side.
$1.4T*
SoftBank Conditional Tranche CONTINGENT OBLIGATION Contingent
$30B of the Feb 28 2026 round (SoftBank tranche). Amazon $50B ($15B wired; $35B milestone-contingent on Trainium chip deal and enterprise milestones). SoftBank holds effective veto on certain corporate actions — coercive rights characteristic of a creditor, not a passive LP.
$30B*
Employee Equity / RSUs RETENTION OBLIGATION Sub-Senior
$2.5B in stock-based compensation in H1 2025 alone. Unvested RSUs represent a forward equity obligation that, in a failure, would accelerate vesting demands and talent flight simultaneously.
$5B+/yr
VC / Institutional Equity PURE EQUITY — LAST IN LINE Junior / Residual
Thrive ($1.3B), Sequoia ($1B+), a16z / T. Rowe ($0.8B), and others. Pure equity with no contractual claims ahead of operational obligations. In a liquidation, last to be paid — after Azure, Microsoft revenue share, infrastructure counterparties, and employees.
~$5B
True Obligation Load ALL TIERS COMBINED
This is not a capital structure anyone has published. It is assembled from publicly available disclosures, filings, and reports. The '*' items are commitments or contingencies — not funded debt — but they are real economic obligations that would crystallize in a failure.
~$1.5T+
The Microsoft Knot
Microsoft is simultaneously OpenAI's largest equity investor, primary compute creditor (OpenAI owes Azure invoices), revenue share counterparty (receiving 20% of all revenue), and primary distribution channel for its models. In a restructuring, these roles create directly conflicting interests: as creditor, Microsoft wants to be paid; as equity holder, it wants enterprise value preserved; as a company that needs the models running, it may prefer an operational rescue over a financial one. There is no clean way to resolve these conflicts. They would have to be negotiated simultaneously under time pressure.
The Maturity Mismatch Rhyme
OpenAI's primary "assets" — model weights, trained capability, market position, talent — are long-duration and effectively illiquid (model weights have contested ownership; talent walks out the door). Its funding is episodic equity rounds with no guaranteed continuity. The burn rate creates a rolling refinancing need every 18–24 months. Each round must be larger than the last. Today's $110B raise resets the clock. It does not stop it. The moment a future round fails to close on expected terms, the runway math changes within quarters, not years.
The Creditor Hierarchy Problem
In a Chapter 11 scenario, who's actually senior? Azure compute payables are operational — stopping payment stops the company. The Microsoft revenue share has contractual standing. Infrastructure counterparties hold take-or-pay claims. SoftBank has conditional commitments that likely fall away entirely. VC investors are pure equity, last in line, behind every contractual obligation. The model weights — the crown jewel asset — have ownership that the nonprofit-to-for-profit conversion made genuinely ambiguous. Bankruptcy counsel would need months to establish a creditor map. The operational damage would be done in days.
* A note on the $1.4T figure The infrastructure commitment figure is a stated 8-year commitment, not a single obligation. In a failure scenario, counterparties would seek damages, not full payment — but litigation exposure on cancelled commitments at this scale would itself represent a systemic event for the construction, power, and real estate sectors involved.
The Azure Credit Illusion
A critical precision point that the funding round narrative obscures: when Microsoft "invests" in OpenAI, a significant portion — estimates suggest 50–60% — is structured as Azure cloud credits, not hard currency. Cloud credits cannot pay a $2M AI researcher salary. They cannot cover legal fees from Elon Musk's lawsuit. They cannot pay office rent, regulatory compliance teams, or the armies of safety researchers required by an increasingly hostile regulatory environment. The credits recycle directly back into Microsoft Azure's revenue, boosting Microsoft's reported cloud numbers while leaving OpenAI perpetually scrambling for actual cash. The headline "Microsoft invested $13B in OpenAI" and the operational reality that OpenAI must continuously raise hard currency every quarter just to make payroll are both simultaneously true — and that gap is the structural cash crisis hiding inside the investment narrative.
09 —

The Default Chain:
Who Falls First

Systemic risk analysis requires mapping not just who is exposed, but in what order exposure becomes loss — and which losses trigger subsequent losses. The OpenAI contagion chain has multiple transmission pathways, some financial, some operational, some reputational. They do not all fire simultaneously, but they do not operate in isolation either.

T-0 · Trigger
OpenAI
$0
API goes dark. No new completions. Existing contracts voided. $150B total cash raised — but Amazon's $35B tranche is milestone-contingent; effective immediate liquidity ~$115B. Employees depart within days.
T+Hours
Feb 28 Round Investors
$110B
Amazon ($50B commitment; $15B already wired, $35B conditional — Trainium milestones now moot), Nvidia ($30B — GPU demand narrative collapses simultaneously), SoftBank ($30B — Vision Fund marks crater). All three stocks fell in premarket on deal day. Failure makes those moves permanent.
T+Days
Microsoft Azure
$13B+ equity · $75B+ rev
Equity write-down on $13B+ historical investment. M365 Copilot disrupted for 80% of Fortune 500. Azure AI revenue — growing at 175% YoY — goes to zero. Microsoft notably did not participate in the Feb 28 round, which in retrospect looks like a hedge. They still hold an option to join.
T+Weeks
VC Portfolio & AI Startups
$40B+
6,800+ API-dependent SaaS platforms lose core product functionality overnight. 35% of top-funded AI companies face existential product crisis. Series B/C rounds pulled. VC portfolio marks revised industry-wide — not just OpenAI-dependent names.
T+Months
Broad Market & Infrastructure
$500B–1T+
$1.4T in datacenter commitments collapse. NVDA and AMD face guidance revisions as AI capex narrative breaks. The 92% of Fortune 500 running production AI workflows scramble for alternatives. Regulatory pile-on across every jurisdiction that was watching for a pretext. AI as an investment thesis — the narrative underpinning current tech multiples — takes systemic damage.

The operational contagion is more immediate than the financial contagion — and harder to hedge. Financial exposure can be partially offset with puts, shorts, or diversified positioning. Operational dependency is binary: either the API works or it doesn't.

Consider the healthcare pathway. Multiple health systems have deployed GPT-4-based clinical decision support tools in production environments. Regulatory approval processes for AI medical tools are lengthy — switching to an alternative model isn't a weekend project. A sudden API outage creates a forced fallback to prior manual processes, with staff who may have meaningfully reduced their familiarity with those processes. The litigation exposure alone would be substantial.

The financial services pathway is more nuanced. The largest financial institutions are legally required to conduct vendor due diligence and maintain business continuity plans — which means their OpenAI dependencies are, in theory, better documented and more substitutable. In practice, the pace of deployment has outrun the compliance frameworks. Many production AI workflows were stood up faster than the accompanying vendor risk assessments.

The irony is that the sectors most capable of absorbing the shock — large banks, major tech companies — are also the ones with the deepest integration. The sectors least capable — startups, smaller healthcare systems, government agencies — are exactly the ones with the least capacity to switch.

THE MICROSOFT KNOT Microsoft's entanglement with OpenAI is so deep that it creates an unusual dynamic in the failure scenario: Microsoft is simultaneously the most exposed counterparty and the most plausible rescuer. With $13B+ invested, a 20% revenue share through 2032, and Azure's AI revenue growth at 175% year-over-year, Microsoft has every incentive to prevent an uncontrolled failure.

THE SOFTBANK VARIABLE SoftBank's $30B commitment in the Feb 28 2026 round is the more volatile element. Vision Fund I and II have demonstrated that SoftBank is capable of writing large checks and absorbing large losses. The milestone-contingent structure of the Amazon $35B tranche means a prolonged governance crisis or enterprise adoption shortfall could simultaneously trigger the failure and eliminate a significant backstop. Andy Jassy described OpenAI as "one of the very big winners long term" on CNBC — but the Trainium chip deal that anchors the Amazon relationship is a compute infrastructure bet that takes years to play out.

THE $1.4T SHADOW OpenAI committed to $1.4T in datacenter infrastructure spending over eight years. If OpenAI fails, those commitments collapse. The counterparties — hyperscalers, chip manufacturers, construction companies, real estate developers — face project cancellations at a scale that would ripple through capital markets with a force well beyond the AI sector.

09 —

Picking Up
the Pieces

"The horror! The horror!"

— Kurtz, Heart of Darkness, Joseph Conrad, 1899 (adapted in Apocalypse Now, 1979)

The aftermath of an OpenAI failure would not be a clean transition to a world of competing alternatives. It would be a prolonged, messy, legally complex scramble — part bankruptcy proceedings, part emergency remediation, part market restructuring. There are several distinct dynamics worth modeling.

Aftermath A
The Acqui-hire Scenario
Microsoft, Google, or Apple acquires the talent and IP in a structured sale. The most benign outcome — model access is preserved, API continuity potentially maintained, and the capability is absorbed rather than destroyed. Microsoft is the most natural acquirer but faces antitrust scrutiny given its existing stake. Google acquires a competitor's crown jewels. Apple gets the AI capability it has conspicuously lacked. Each scenario has different market structure implications.
Aftermath B
The Controlled Wind-Down
OpenAI's board, investors, and major counterparties negotiate a structured API deprecation — 90-day notice, model weights potentially open-sourced, enterprise contracts transitioned to Azure-hosted alternatives. This is what a well-governed failure looks like. It requires cooperation from parties with conflicting interests and significant coordination costs. OpenAI's governance complexity makes this harder than it sounds.
Aftermath C
The Regulatory Receivership
In a regulatory shutdown scenario, the timeline is not negotiated — it's imposed. API access ceases on the regulator's schedule, not the market's. Government entities may seize model weights as national security assets. Enterprise customers have no transition period. This is the most damaging scenario for dependent parties and the most likely to trigger a broader political response that reshapes the entire AI sector's regulatory environment.

The model weights question is the most consequential technical issue in any failure scenario. GPT-4 and its successors represent billions of dollars of training compute. Who owns them in a failure is not obvious. OpenAI's nonprofit-to-for-profit conversion changed the ownership structure; Microsoft has rights but not full control; investors have claims; and regulators may have views about whether these assets should be treated as public goods.

In a Chapter 11-style proceeding, model weights become a contested asset in a bankruptcy estate. The proceedings could take years. In the interim, API access disappears. The operational damage is done long before the legal questions are resolved.

Market concentration would paradoxically increase in the immediate aftermath. The two obvious beneficiaries — Google and Anthropic — would absorb displaced demand rapidly. But this creates its own systemic risk: a market that was arguably too dependent on one provider becomes dependent on two. The open-source alternatives (Meta's Llama, Mistral) would benefit from accelerated adoption but lack the enterprise support infrastructure for rapid large-scale deployment.

The prediction markets angle is instructive here. A functioning derivatives market in AI operational risk would price the probability of disruption continuously — giving risk managers a forward signal before the event, not a post-mortem after. The signals would have been there. The question is whether anyone was reading them.

09 —

What If the
Skeptics Are Wrong?

Fair is fair. The bear case — that OpenAI bleeds out on SaaS economics while Google's structural advantages compound in the background — is coherent, well-supported, and argued by people who are not idiots. But it rests on one assumption worth pulling apart: that OpenAI's actual business is the one it has today. There is mounting evidence the business it's building is something rather different — and if that business works, the current burn rate looks less like structural failure and more like a very expensive land grab.

The bear case is about the business OpenAI has today. The counter-case is about the business OpenAI is building — and whether its current burn rate is rational investment in that second business, not structural failure in the first.

🏥
ChatGPT Health
Medical records (via Torch acquisition), lab results, fitness trackers, clinical decision support. 230M users already asking health questions on vanilla ChatGPT. Excluded from EU/UK — which signals the data ambition is real. If OpenAI becomes the health data layer, the addressable market dwarfs SaaS subscriptions.
230M
Health query users before dedicated product launched
🌐
Atlas Browser + OS
A Chromium-based browser that stores memories from every visited site on OpenAI's servers. Agentic checkout, booking, document creation in-app. ChatGPT OS disintermediates Canva, Zillow, Spotify from their users. The ad inventory this creates — targeting based on entire browsing intent, health data, and AI conversation history — would be the most powerful ever assembled.
95%
Of 900M users on ad-eligible free tier
🎧
Sweetpea + Neural Interface
Jony Ive-designed always-on wearable (late 2026), followed by non-invasive BCI investment via Merge Labs ($252M seed led by OpenAI). Altman's stated goal: think something, ChatGPT responds. If ambient sensory data becomes the input layer, OpenAI captures context that no search engine or social network has ever touched — physical world, internal state, biological signal.
$6.5B
Jony Ive acqui-hire valuation for io design firm
Revenue Scenario · Data Moat vs SaaS-Only
Scenario 2030 Revenue Viability
Data moat fully materializes $400B+ Profitable 2028
Partial moat (health + browser) $220B Near Break-Even
SaaS only (OpenAI projection) $145B $30B net · Marginal
SaaS + margin inversion $80B eff. Crisis · 2027

The counter-thesis doesn't require believing OpenAI will build AGI. It requires believing that the combination of health data, browsing intent, conversational history, ambient sensing, and eventually neural signal — assembled into a unified personal profile at scale — creates an advertising and data licensing business that makes Google's current position look like a warm-up act.

Sam Altman is not hiding this ambition. He stated explicitly that college students use ChatGPT as an operating system. His investment in Merge Labs — a brain-computer interface startup working toward non-invasive ultrasound neural monitoring — is not a research curiosity. It is a stake in the input layer of the next computing paradigm, several years before that layer becomes commercially viable.

The honest risk manager's position is this: the failure thesis and the counter-thesis are not mutually exclusive on the same timeline. The Feb 28 2026 raise dramatically extends the near-term runway — but it does not eliminate the structural question. OpenAI can face a genuine funding crisis in 2029 or 2030 while also being correct about its long-run data moat thesis. The Amazon Trainium dependency that funded the raise creates a new concentration risk that replaces the Nvidia/Azure one. The gap between the current burn structure and the eventual data-moat monetization is where the systemic risk lives — in the operational disruption that occurs during that interval. A strategically correct company, adequately capitalized today, can still cause a very messy event if the timeline slips.

The River Has
an End

The financial crisis of 2008 wasn't unforeseeable — it was unforeseen, which is different. The instruments that eventually failed were sitting on balance sheets, the counterparty exposures were documentable, and the conditions for collapse were present years before the collapse itself. The gap was not analytical capacity; it was the unwillingness to take seriously a scenario that would have been inconvenient to price.

The $150B raise is the largest private funding round in history. It is also, in the context of what OpenAI must accomplish, a number with an end. The risk this paper describes is not imminent. It is structural — built into the four scenarios, the cost gap, the deepening enterprise dependency, and the governance fragility that no amount of capital resolves. Seven more years on the river means seven more years of going deeper. The question risk managers should be asking is not whether OpenAI survives 2026. It is what their exposure looks like at the end of a runway that is long, but finite.

The enterprises most exposed are not the ones that made a reckless bet. They are the ones that made a reasonable decision, repeated it, and woke up one day to find that the river had taken them somewhere they never intended to go.

This paper is an invitation to do the work before urgency makes it academic: map the dependency, model the exposure, build the contingency. The boat is moving. The shore is already gone. That is not a reason for alarm. It is a reason for a plan.

ABOUT THIS ANALYSIS This paper draws on publicly available financial disclosures, industry research from Menlo Ventures, Sacra, Deutsche Bank, and Fortune, as well as OpenAI's own enterprise reporting. Published March 2026. Scenarios modeled February 27 2026, the day OpenAI announced its $110B funding round (SoftBank $30B, Amazon $50B, Nvidia $30B), $730B pre-money valuation, and $150B total cash on hand ($110B raise + ~$40B prior). Probability estimates for failure scenarios represent qualitative risk assessments, not actuarial calculations. All financial figures are as reported or estimated as of February 2026.

TS IMAGINE TS Imagine provides institutional trading and risk management infrastructure through TradeSmart and RiskSmart platforms, serving banks, trading houses, and financial institutions managing complex derivative, commodity, and emerging market exposures. Our research series examines structural risks at the intersection of technology and financial markets.

DISCLAIMER This document is for informational purposes only and does not constitute investment advice or a recommendation to buy, sell, or hold any security. The views expressed are those of TS Imagine Research and are subject to change without notice.

Appendix A —

Methodology & Calculations

All figures in this report are derived from public sources, disclosed financial filings, and internally constructed models. This appendix documents the calculation methodology behind every major number, the assumptions embedded in the interactive tools, and the sourcing for key data points so that institutional readers can stress-test or replicate the analysis.

A1

Burn Rate, Runway & Cash Position

Inputs & Sources
H1 2025 net loss: $13.5B — OpenAI disclosed H1 financials Sep 2025; CNBC / WSJ reporting
H1 2025 revenue: $4.3B — same disclosure
Annualized burn: $27B (H1 × 2) — conservative; actual may accelerate with Stargate commitments
2028 op. loss: $74B — Deutsche Bank analyst Jim Reid, Feb 2026
Cumulative 2024–29 losses: $140–143B — Deutsche Bank; Wikipedia/OpenAI disclosures
Cash position pre-raise: ~$40B — prior rounds (SoftBank partial, Thrive, Microsoft)
New raise (Feb 28 2026): $110B — Amazon $50B ($15B upfront + $35B conditional), Nvidia $30B, SoftBank $30B
Total liquidity post-raise: ~$150B — raise + prior cash
Derived Calculations
Cost per $1 revenue: $13.5B loss + $4.3B revenue = $17.8B spent on $4.3B revenue = $3.14 per $1 earned (rounded to $3.30 to account for non-H1 cost acceleration)
Hourly burn: $13.5B × 2 ÷ 365 ÷ 24 = $3,082/hr net basis; "$575K/hr" figure from earlier OpenAI press reporting includes gross spend components
Theoretical runway: $150B ÷ $27B annual net burn = ~5.5 years at flat burn; real figure lower as 2028 losses escalate toward $74B
HSBC 2030 gap: $207B shortfall = projected cumulative spend minus projected cumulative revenue through 2030 per HSBC analyst model, Fortune Feb 2026
Key assumption flag: Amazon's $35B of its $50B commitment is milestone-contingent (API adoption milestone or IPO). Effective immediate liquidity is ~$115B ($150B − $35B conditional). Runway calculations using $150B are best-case; $115B is conservative-case. The interactive runway tool uses $17.5B as default to reflect the pre-raise cash position — slide up to model the new reality.
A2

Chain Reaction VaR & Blast Radius

VaR Model Structure
The interactive VaR tool uses a Monte Carlo simulation with configurable parameters:
EAD (Exposure at Default): Total economic dependency of all counterparties — set at $847B baseline reflecting direct + operational + market exposure
LGD (Loss Given Default): Percentage of EAD unrecoverable in a failure event; defaults to 65% reflecting limited asset recovery in an IP-heavy company
PD (Probability of Default): Per-scenario failure probability, sourced from prediction market contract prices and adjusted for model confidence
Contagion multiplier: 1.0x–5.0x amplification representing second-order losses (VC marks, enterprise disruption, regulatory costs) that exceed direct financial exposure
VaR output: 95th percentile loss at a given confidence level across 10,000 simulated paths
Blast Radius Ring Methodology
The concentric ring chart maps economic exposure by proximity to the OpenAI core:
Core ($110B+): Direct equity holders — Amazon, Nvidia, SoftBank at Feb 28 2026 commitments
Ring 1 — MSFT ($380–520B): Microsoft equity write-down ($13B) + Azure revenue risk ($75B+/yr) + Copilot/O365 disruption (300M seats × estimated transition cost)
Ring 2 — Enterprise ($200–350B): Fortune 500 enterprise workflow disruption cost; estimated 12–18 month manual fallback × average dependency cost per sector
Ring 3 — Startups ($80–140B): 6,800+ API-dependent SaaS platforms × average ARR × survival probability in a 90-day API outage scenario
Ring 4 — NVDA/Infrastructure ($50–120B): Nvidia datacenter revenue at risk + NVDA equity shock × market cap + GPU supply chain repricing
Ring 5 — Systemic: Non-linear; includes regulatory costs, AI capex pullback across all hyperscalers, and market multiple compression on AI-narrative tech
A3

Scenario Probability Calibration

Scenario Base PD Methodology Key Sources
A · Funding Gap 14% Conditional on Amazon $35B tranche failing + market confidence shock prevents follow-on. Double-trigger required post-raise; probability revised down from 22% given $150B cushion. Yahoo Finance; HSBC analyst model; Polymarket OpenAI failure contracts
B · Regulatory Shutdown 12% EU AI Act enforcement + US national security review + state AG coordination required simultaneously. Each jurisdiction independently unlikely; joint probability estimated via independence assumption. EU AI Act text; Congressional testimony; FTC/DOJ precedent
C · Governance Crisis 15% Board demonstrated in 2023 it can execute a management change that triggers near-collapse. Mechanism is documented. Probability elevated because the structural fragility is a known, confirmed quantity — not a speculative one. Altman congressional testimony; sworn depositions; WSJ Nov 2023
D · Competition Collapse 28% Highest probability scenario. Google's 30–44% cost advantage compounds as model quality converges. ChatGPT US traffic already down 35% Nov 2025. Enterprise share down to 27% from 50% in 2023. Not a failure scenario per se — a slow margin bleed that makes future capital raises structurally impossible. SemiAnalysis TPU cost study Nov 2025; Similarweb traffic data; Gartner enterprise survey
E · Margin Inversion 35% The thermodynamic scenario: cost structure becomes structurally unviable as compute costs rise faster than revenue. OpenAI's own projections show $74B op loss in 2028. This is the base case from their own internal documents, not a bear case constructed by analysts. WSJ via OpenAI internal docs; Deutsche Bank; Microsoft Q1 FY26 10-Q equity method investment disclosure
Note on probability overlap: Scenarios are not mutually exclusive. Governance crisis (C) and funding gap (A) can co-occur; competition collapse (D) and margin inversion (E) share the same underlying mechanism. Joint probabilities are not simply additive. The Monte Carlo tool draws correlated failure paths, not independent coin flips.
A4

Margin Inversion & Cost Structure

OpenAI per 1M tokens
NVIDIA GPU margin (~75% GM): $6.20
Microsoft Azure margin: $4.80
OpenAI infrastructure ops: $4.60
R&D / training cost share: $2.80
Total: ~$18.40
Google per 1M tokens
NVIDIA GPU margin: $0 (TPU — no vendor margin)
Electricity + cooling: $1.40
Silicon depreciation (TPU): $2.20
Google ops cost share: $2.20
Total: ~$5.80
Methodology Notes
Cost estimates are illustrative approximations based on publicly reported data. Exact per-token costs are not publicly disclosed by either company. TPU cost advantage sourced from SemiAnalysis Nov 2025 study: 30% TCO advantage for standard config, 44% for Google's full torus deployment. NVIDIA GPU margin (~75% gross) is a reported figure from NVIDIA investor disclosures Q3 2025.
A5

Capital Structure & The Azure Credit Illusion

Microsoft's $13B+ "investment" is not a simple equity purchase. A material portion (estimated at ~60%) recycles as Azure cloud credits — prepaid compute time on Microsoft's infrastructure. This is structurally equivalent to a vendor financing arrangement: Microsoft writes a check to OpenAI, OpenAI writes a check back to Microsoft for Azure services. The net cash transfer is a fraction of the headline figure.

Implication for liquidity analysis: Models that treat the full $13B as available cash overstate OpenAI's hard-currency position. Real cash (usable for salaries, rent, legal fees, non-Azure costs) is significantly lower. The $4B+/yr hard currency shortfall estimate assumes approximately 60% of Microsoft's investment is cloud credits, leaving ~$5B in actual cash against ~$9B in non-compute operating costs.

Source: Azure credit illusion structure inferred from public reporting on Microsoft OpenAI deal terms (The Information, Bloomberg); exact split is not publicly disclosed. Figures should be treated as estimates pending full disclosure.
MSFT Total "Investment"
$13B+
Headline figure, multiple tranches
Azure Credit Est. %
~60%
Estimated; not publicly confirmed
Hard Currency Est.
~$5B
After stripping cloud credits
Annual Hard$ Shortfall
$4B+
Non-compute operating costs
A6

Prediction Market Data & Composite Index

The Polymarket Composite Risk Index displayed in the paper is a weighted composite of three market signals:

Financial Risk (40% weight): Polymarket "OpenAI failure" contract price × inverse of current funding round size (larger raise → lower weight on near-term failure probability)
Governance Risk (35% weight): Polymarket "Altman leaves OpenAI 2026" contract price + Kalshi governance disruption contract (averaged where both available)
Competitive Risk (25% weight): Polymarket "Google Gemini surpasses ChatGPT market share by end 2026" + Metaculus "OpenAI retains >50% enterprise LLM share through 2027" (inverted)

Liquidity caveat: Thin prediction markets can be gamed. Polymarket OpenAI contracts show limited open interest relative to headline risk. Numbers should be treated as market signals, not calibrated probabilities. Cross-reference against analyst PD estimates before trading on prediction market pricing alone.
A7

Monte Carlo Runway Simulator

The Monte Carlo runway tool runs 10,000 paths simulating OpenAI's cash balance over a configurable horizon. Each path draws from:

Revenue growth: Normally distributed around user-set base case, with fat tails (±2 standard deviations = ±40% of mean). Revenue draws are correlated with burn draws at ρ = 0.4 (higher revenue → higher inference costs).

Burn rate: Base burn escalates at a configurable annual growth rate; defaults to 15%/yr consistent with historical OpenAI cost scaling. Funding Gap events are modeled as a Poisson process with λ set by the scenario (baseline λ = 0.08/yr, Scenario E λ = 0.45/yr).

Funding rounds: Modeled as discrete events with probability drawn from the Poisson process. Each round adds a lognormally distributed cash infusion (base: $10–40B depending on scenario; today's $110B is an extreme tail of this distribution — a genuine outlier).

Ruin defined as: Cash balance falling below floor (configurable; default $2B minimum operating reserve). Ruin probability displayed is the fraction of 10,000 paths that hit this floor within the selected horizon.

Seed for reproducibility: The tool displays the random seed used. Re-entering the same seed reproduces the same result exactly. Change the seed to test sensitivity.
A8

Primary Sources Referenced

Financial / Analyst
Microsoft Q1 FY26 10-Q — equity method investment disclosure ($3.1B quarterly decrease in net income)
Deutsche Bank — Jim Reid, cumulative loss estimate $140B 2024–29; $74B 2028 op loss
HSBC — $207B funding shortfall projection through 2030 (Fortune, Feb 2026)
Goldman Sachs — AI capex underpinned by real earnings growth thesis
WSJ / The Information — OpenAI internal financial documents (leaked; Sep–Nov 2025)
SemiAnalysis — TPU vs NVIDIA GB200 TCO study, Nov 2025
Sacra — OpenAI revenue estimates, annualized run rate methodology
Apollo / Torsten Slok — AI bubble comparison to internet era, CNBC Aug 2025
News / Primary Sources
TechCrunch / Axios — Microsoft non-participation in Feb 2026 round; option to join retained
Yahoo Finance — Amazon $50B structure: $15B upfront, $35B conditional on API adoption milestones or IPO
PYMNTS / CNBC — 900M WAU, 50M paying consumers, 9M business users; $280B 2030 rev target
Reuters — circular financing concerns; "exacerbates Wall Street concerns" quote
INSEAD Knowledge — dotcom-era vendor financing comparison, circular money flows
Newcomer Newsletter — Stargate as bilateral deals; fundraising-as-narrative analysis
The Verge / Sam Altman — "AI market is in a bubble" (Aug 2025 interview)
Altman / Sam Sutskever congressional testimony — governance fragility primary documentation
Disclaimer: This appendix documents the methodology behind the analysis as presented. Where estimates are used, they are flagged as such. This document does not constitute financial, legal, or investment advice. TS Imagine Research publishes for institutional informational purposes. Reproduction requires written permission. © 2026 TS Imagine.