OpenAI is the category-defining consumer AI franchise — ~900M weekly users and a ~$25B annualized revenue run-rate reached in roughly three years. It is also asking public markets to absorb an entry valuation (~$1T target) that already capitalizes a 2029–2030 outcome, against record cash burn, eroding category share, and a gross margin moving the wrong direction. The question for an IC is not "is OpenAI a great company" — it is "is the success already in the price."
PARTICIPATE — SELECTIVELY, ON PRICE The franchise is real and the TAM is generational; the risk is entirely the entry multiple and the financing dependency behind it. We would underwrite a position only with a clear price discipline (a fill meaningfully below the reported ~$1T target, ideally ≤ ~25× forward revenue) and explicit sizing for binary financing/governance outcomes. Above ~40× run-rate revenue, risk-adjusted return skews unfavorable.
OpenAI completed a for-profit recapitalization on Oct 28, 2025, converting its for-profit arm into OpenAI Group PBC (a public-benefit corporation) controlled by the nonprofit OpenAI Foundation. Microsoft holds ~27% (~$135B at the time); the Foundation holds ~26% (~$130B). Microsoft retains IP and Azure access rights through 2032, including to any AGI verified by an independent panel. (Reported: CNBC, Fortune, TIME, Al Jazeera — Oct 2025.)
Apr-2023 → reported Sep-2026 IPO target. Final point is a reported target, not a closed round.
~30× step-up in ~3 years. Oct-2025 $500B was a secondary (no primary capital); Apr-2026 $852B was a $122B round anchored by Amazon, NVIDIA, SoftBank, Microsoft. (Reported: Visual Capitalist, CNBC, Sacra, startupbooted.)
Solid = reported/actual; hatched = company targets. Revenue is annualized run-rate.
Revenue ~$2B (2023) → ~$20B exit-2025 → ~$25B Q1-2026; targets ~$100B (2029) and ~$200B (2030). Cumulative burn reported ~$665B through 2030; cash-flow positive ~2029–2030. (Reported: Fortune, CNBC, Sherwood, RDWorld.)
Inference cost quadrupled in 2025; margin fell as the company scaled.
~40% (2024) → ~33% (2025). Company target band ~52–67% later in the decade — an inversion of the current trend, not a continuation. (Reported: Fortune / internal projection summaries.)
| Metric ($B) | 2023 | 2024 | 2025 | 2026E | 2027E | 2028E | 2029E | 2030E |
|---|---|---|---|---|---|---|---|---|
| Revenue run-rate (annualized) | ~2 | ~6 | ~20 | ~30 | — | — | ~100 | ~200 |
| Net loss (P&L) | — | — | — | ~(14) | — | — | — | profit goal |
| Cash burn | — | — | — | ~(25) | ~(57) | ~(74–85) | improving | +~39 |
| Adj. gross margin | — | ~40% | ~33% | — | — | — | ~52–67% (target) | — |
Figures are reported/estimated and mix full-year actuals with annualized run-rates and internal projections that have varied across reports; treat as directional, not audited. Dashes = not reliably reported.
OpenAI has accumulated roughly $1.4T of multi-year compute & energy commitments across seven vendors — recently reframed to ~$600B by 2030 as rentals (opex) rather than the capital builds the Stargate JV implied. HSBC estimates OpenAI must still find ~$207B to meet these obligations. The IPO is, in part, a financing event to close that gap.
Named multi-year commitments; sums to ~$1.15T of the ~$1.4T headline.
(Reported: Tom Tunguz, DCD/HSBC, Built In, Bloomberg.)
Annualized revenue vs. headline commitment stack ($B, log scale).
~$1.4T of commitments against a ~$25B revenue run-rate — a ~56× ratio. The underwrite is a bet that demand arrives before the obligations bite.
OpenAI still leads consumer AI but is losing share fast, and it is losing the enterprise segment it most needs — the stickier, higher-margin one. Three credible strategies now coexist: consumer scale (ChatGPT), ecosystem distribution (Gemini), enterprise precision (Claude).
Feb-2025 → Apr-2026. ChatGPT still #1, but the curve is bending.
ChatGPT 76.5%→54.7%; Gemini 5.6%→27.4%; Claude ~3%→8.2%. (Reported: Similarweb/Momentic, Fortune, Vertu.)
At ~34–40× annualized revenue, OpenAI would IPO at a multiple richer than peak-era hyperscalers and top SaaS names ever sustained at scale. Every comparable saw multiple compression as it grew; buyers here are underwriting the 2029–2030 revenue ($100–200B targets) today. Price, not product, is the thesis risk.
~$1.4T in commitments (reframed ~$600B) against a ~$207B reported funding gap means the public raise is structurally necessary to fund obligations — different from a mature company listing to give holders liquidity. That changes the negotiating posture and the downside.
Software is supposed to get more profitable at scale. OpenAI's adjusted gross margin fell from ~40% to ~33% as it grew, because inference cost quadrupled. The bull case literally requires margins to reverse direction to the 52–67% band — an assumption, not a trend.
The same dollars cycle among NVIDIA → OpenAI → Oracle/Microsoft and back (NVIDIA's ~$100B investment funds chip purchases; Microsoft's $250B cloud commitment is also its revenue). It flatters all parties' growth and creates correlated, cascading downside if AI demand disappoints.
A nonprofit Foundation controls a ~$1T PBC; the Microsoft AGI clause and IP access through 2032 are overhangs public shareholders cannot vote away. Governance is a structural discount that the reported target price does not appear to reflect.
A confidential submission means the headline numbers ($25B run-rate, $14B loss, margins) are unaudited and report-to-report inconsistent. The first true diligence event is the public S-1; until then, every model is built on sand. (This is why OloLand's filing-grounded engines returned nothing here — by design.)
Illustrative OloLand-framework scenarios on stated revenue multiples — not a filing-grounded DCF (no audited cash flows exist). Anchored on the reported ~$852B last round and ~$1T target.
Reported IPO target ≈ $1,000B+ sits at the top of the Base case — i.e. the market is pricing the base-to-bull path as the entry point, leaving little margin for execution slippage.
| Scenario | Driver assumptions | Multiple basis | Implied EV |
|---|---|---|---|
| Bear | Growth decelerates as Gemini/Claude take share; margin pressure persists; AI-capex sentiment cools | ~12–15× 2027E rev (~$45B) | $550–700B |
| Base | Reaches ~$100B by 2029; enterprise stabilizes; margins begin to recover | ~10–12× 2029E rev ($100B) | $950–1,200B |
| Bull | ~$200B by 2030; agents + ads + enterprise compound; margins hit target band | ~8–10× 2030E rev ($200B) | $1,600–2,100B |
Multiples are analyst-set illustrations for scenario framing; actual public-market multiples depend on rate environment, AI sentiment, and disclosed margins. Not investment advice.
OloLand-style risk framing; positions are analyst estimates pending the S-1.
| # | Risk | Category | Severity | Likelihood | Why it matters |
|---|---|---|---|---|---|
| 1 | Entry valuation / multiple | Market | Critical | ~65% | ~$1T target capitalizes 2029–30 outcomes; little room for slippage. |
| 2 | Financing dependency & burn | Liquidity | Critical | ~55% | ~$207B reported gap; burn peaks ~$74–85B (2028). |
| 3 | Competitive erosion | Commercial | High | ~70% | Consumer share 76.5%→54.7%; ~70% of enterprise deals lost to Anthropic. |
| 4 | Inference / gross-margin structure | Financial | High | ~60% | Margin fell 40%→33% while scaling; bull case needs reversal. |
| 5 | Compute concentration / circular financing | Structural | High | ~45% | Correlated downside across NVIDIA/MSFT/Oracle web. |
| 6 | Governance: nonprofit control + AGI clause | Governance | High | ~50% | Economics without control; MSFT rights through 2032. |
| 7 | Supplier dependence (MSFT, TSMC) | Operational | Med-High | ~35% | Single points of failure in cloud and fabrication. |
| 8 | Litigation / regulatory | Legal | Medium | ~45% | Musk/xAI, antitrust scrutiny, copyright (NYT), congressional probes. |
| 9 | Key-person (Altman) / talent | People | Medium | ~25% | Concentrated leadership; intense talent competition. |
| 10 | Monetization of free base / ads | Commercial | Medium | ~50% | ~5% pay; ads model unproven at this scale. |
| 11 | Disclosure / provenance | Diligence | Medium | ~40% | No audited public financials until the S-1 prints. |
| 12 | AI safety / model regulation | Regulatory | Medium | ~40% | Frontier-model rules could raise cost or constrain products. |
Sources (reported/estimated — public record as of June 8, 2026): Sacra; Visual Capitalist; CNBC (restructuring Oct-2025; $852B round; risk factors Mar-2026); Fortune (cash burn / margins; Microsoft–OpenAI debt); TIME; Al Jazeera; Reuters/Bloomberg (circular deals); Sherwood News; DataCenterDynamics / HSBC ($207B gap); Tom Tunguz (compute spend); Momentic / Similarweb / Vertu (market share); RDWorld; Polymarket; technologychecker.io. OloLand deal record: deal43de247c961c.
Important: This memo is a pre-S-1 underwriting read built entirely on public reporting and OloLand-framework analysis. It is not based on a filed or audited S-1, contains figures that vary across sources, and is not investment, legal, or financial advice. Re-underwrite on the published S-1. Prepared by Aleks Niebylski, OloLand — the verifiable intelligence layer for diligence.