OpenAI

Phase A — Understand the business

Company Overview

OpenAI is the frontier AI lab behind ChatGPT, the GPT model family, the OpenAI API, Sora (video), and a forthcoming Jony-Ive-designed consumer device. Plain-terms business model: it spends enormous sums training frontier models, then monetizes them three ways — (1) consumer subscriptions (ChatGPT Plus/Pro/Business), (2) a developer API priced per token, and (3) enterprise seats and deployments. As of March 2026 it ran at a $25B annualized revenue run-rate ($2B/month) — though other sources peg the run-rate lower (~$20B late-2025), a spread that itself flags how unaudited these figures are.

Scale of the franchise:

• 900M+ weekly active users on ChatGPT (March 2026)
• 50M+ consumer subscribers; 9M+ paying business users (grew 4x in under six months)
• 1M+ business customers (surpassed Nov 2025); 92% of the Fortune 500 are customers; 7M+ enterprise seats

Corporate structure (post-restructuring, Oct 2025): a recapitalization converted the for-profit arm into OpenAI Group PBC (a Delaware public-benefit corporation) controlled by the nonprofit OpenAI Foundation. Reported ownership: Foundation ~26% (an equity stake valued ~$130B), Microsoft ~27%, and ~47% held by current/former employees and other investors. This is the single most important structural fact for an investor — it is what makes a tradeable security conceivable at all, and the named private-watch catalyst.

Customers/suppliers/competitors: customers span consumers, ~1M businesses, and developers; key suppliers are NVIDIA (GPUs), Microsoft Azure + Oracle (cloud/DC), and increasingly AMD/Broadcom (custom silicon). Competitors: Anthropic, Google DeepMind, xAI, Meta, plus open-weight challengers (DeepSeek, Mistral). Contract structure: mostly recurring (subscriptions + usage-based API), with the cost side dominated by multi-year take-or-pay-style compute commitments — the asymmetry (recurring-but-cancelable revenue vs. locked-in capexCapital expenditureMoney spent on long-lived things — buildings, machines, servers — rather than on running costs.) is the core tension of the whole story.

Supply Chain

The AI stack runs electrons → accelerators → data centers → frontier lab → API/product → end user. Named stakeholders along OpenAI's chain:

The defining feature is circularity. OpenAI's suppliers are increasingly also its investors and its revenue-recognition counterparties: NVIDIA invests in OpenAI which buys NVIDIA GPUs; AMD grants OpenAI warrants for ~10% of AMD against future GPU purchases; Oracle books OpenAI as a giant backlog customer while OpenAI funds Oracle-built capacity. Names or it didn't happen — and here the names recur on both sides of the ledger, which is exactly what the bear case is built on.

Competitive Advantages (moats)

Durable moats:

1. Consumer brand + distribution. ChatGPT is the default noun for "AI chatbot" — 900M WAU is a distribution asset no competitor except Google (via Search/Workspace) can match. This is the strongest moat.
2. Scale economics on training. First-mover compute scale and a multi-year head start on RLHF/reasoning data.
3. Capital access. The restructuring + the ability to raise $40B (2025) and reportedly $100B+ rounds (2026) is itself a moat in a game where compute = capital.
4. Microsoft partnership — Azure compute + GTM + IP cross-license through 2032.

Bargaining power — and where it's thin. Over consumers OpenAI has pricing power (subscriptions). Over suppliers it has very little: it needs NVIDIA's allocation more than NVIDIA needs any single buyer, and it needs cloud capacity badly enough to sign $1.4T in commitments. The moat that is visibly eroding is enterprise API share: ChatGPT's app market share fell from 69.1% (Jan 2025) → 45.3% (2026) as Gemini rose 14.7%→25.2%, and multiple enterprise-LLM trackers now put Anthropic ahead in enterprise API (Anthropic ~32–40% vs OpenAI ~25–27%). OpenAI leads on revenue and consumer; it is losing the enterprise-developer moat to Anthropic and the distribution war to Google. That divergence is the whole investment debate.

Segments

No audited segment disclosure exists (our figures empty — private). Reconstructed revenue mix from public reporting:

• Consumer subscriptions (ChatGPT Plus/Pro/Business): the largest line — 50M+ subs is the bulk of the ~$25B run-rate.
• API / developer: per-token usage; under share pressure from Anthropic and price competition.
• Enterprise: fastest-growing — 9M+ paying business users (4x in <6 months), 7M+ seats.
• Emerging: Sora (creator revenue-share being introduced), forthcoming hardware device (ships no earlier than ~Feb 2027 per a court filing).

Trend: consumer + enterprise accelerating; API decelerating in relative share as competition compresses token pricing. Geography is undisclosed; ChatGPT WAU is global with the US the largest paid market. Treat all segment splits as — directionally reliable, precisely unsourced.

---

Phase B — Measure performance (+private overlay: funding & traction in place of SEC earnings)

Funding & Valuation Trajectory (overlay swap for "Earnings Result")

The funding story is the financial story for a private company. Round history:

The Feb 2026 round is striking on two axes: the size ($100B+ in a single private round is unprecedented) and the investor list — Amazon and NVIDIA writing the biggest checks is the circular-financing pattern in plain sight, and a T. Rowe Price entry is a textbook crossover-fund / IPO-proximity tell.

Burn — the counterweight:

• 2025 actuals: ~$34B spent; reported net loss ~$39B (incl. non-cash charges).
• 2026: internal projection of ~$14B net loss; ~$27B cash burn; Q1 2026 cash burnCash burnHow much more cash goes out than comes in, per period. The clock on a company with no profits. $3.7B on $5.7B revenue (≈65% burn ratio).
• Through-2029: company's own documents reportedly guide to ~$115B cumulative losses before profitability in the 2030s, with burn staying 57% of revenue through 2027 ($63B cash burn in 2027).

Conflict flagged: run-rate figures range $20B–$25B; loss figures range $14B (P&L) to $34–39B (cash/total spend) depending on whether the source means net loss, cash burn, or gross spend. I treat ~$25B run-rate / ~$14B 2026 net loss / ~$27B 2026 cash burn as the central case and label the spread explicitly. A miss already happened: OpenAI missed its internal Q1 2026 revenue and user-growth targets.

Founder & Narrative Sentiment (overlay: founder interviews/comms in place of earnings calls)

Management's public posture in 2026 has shifted from "scale at all costs" to defending the spend and managing IPO expectations. Sam Altman is publicly guiding to an IPO "within the next year" and floating a ~$1T listing valuation. The tonal shift worth flagging: Altman has introduced recursive self-improvement (RSI) as a reason an IPO could be delayed ("the faster the potential RSI takeoff looks, the more it could be advantageous to delay") — simultaneously a capability flex and a hedge. Meanwhile the CFO is audibly off-message (see Lens 9). The recurring phrases: "compute," "scale," "AGI verification panel." What they've stopped saying: the original nonprofit-purity framing — the Oct 2025 restructuring retired it.

Cap Table & Secondary Marks + Frontier Comps (overlay swap for "Comps")

Syndicate quality: tier-1 across the board — SoftBank (anchor), Microsoft (strategic, 27%), NVIDIA + Amazon (strategic/circular), Thrive, a16z, Coatue, Altimeter, plus crossover/mutual money (T. Rowe Price, MGX, TPG). Crossover entry is the IPO-proximity signal the +private overlay is built to catch.

Frontier-lab comp set (private-market valuations —, unaudited; no P/E exists pre-profit, so the comparable is valuation vs. revenue run-rate):

The headline comp fact: Anthropic passed OpenAI in valuation (~$965B vs $852B) on roughly 2x the revenue run-rate ($47B vs ~$25B) as of May 2026. OpenAI is now the more expensive frontier lab on a revenue multiple while losing enterprise share — the single most important relative-value fact in this dossier. (Note: the Anthropic $47B run-rate figure is itself a steep claim and should be treated as unaudited.)

Funding & Product Catalysts (overlay: funding/product events in place of stock-price catalysts)

Events that moved OpenAI's private valuation or narrative materially:

• Jan 2025 — Stargate announced at the White House ($500B / 10GW intent) → infrastructure-scale repricing.
• Mar 2025 — $40B round at $300B.
• Oct 2025 — for-profit/PBC restructuring + Microsoft 27% → unlocked the path to tradeable equity.
• Oct 2025–Jan 2026 — Broadcom (10GW custom ASIC), AMD (~6GW + warrants), Cerebras chip deals → ~33GW of compute commitments in three weeks; cumulative infrastructure commitments reported >$1.4T.
• Feb 2026 — ~$110–122B round at ~$852B.
• May 2026 — Anthropic leapfrogs to ~$965B → competitive-narrative inflection.
• Early Jun 2026 — reported confidential S-1 filing (Goldman Sachs / Morgan Stanley) → IPO-by-2026 odds jumped (prediction markets ~76% YES by year-end).

Pattern: OpenAI's marks react to (a) compute-commitment scale, (b) structural/governance unlocks, and (c) competitive relative-position vs. Anthropic — not to model-release cadence per se. The market is pricing the capital + governance story more than the technology.

---

Phase C — Judge people & books

Management

• Sam Altman (CEO) — founder archetype, promotional. Track record: built OpenAI from nonprofit to ~$850B; unmatched at fundraising and narrative-setting. Skin in the game: famously held ~no direct equity historically (the restructuring may change this — unconfirmed). The defining 2026 trait is promotional behavior — pushing a $1T IPO and a $600B+ data-center bet while the CFO publicly counsels caution.
• Sarah Friar (CFO) — the red flag is the friction. Multiple outlets report Friar no longer reports to Altman (since Aug 2025; reports into Fidji Simo's applications org instead), is reportedly excluded from key financial meetings, and is publicly at odds with Altman over the IPO timeline — she favors 2027, citing unfinished organizational/procedural work and the risk in the ~$600B spend; he wants Q4 2026. A CFO sidelined and openly skeptical immediately before an IPO is a material governance flag.
• Fidji Simo (Applications head) — announced an extended medical leave (worsening neuroimmune condition); combined with other April 2026 executive departures, this compounds key-person/strategic-direction risk.

Capital-allocation history: aggressive to the point of being the entire bear case — $1.4T in compute commitments against a P&L losing $14B+/yr, partly financed through circular supplier-investor structures. Whether that is visionary (buying the scarce input early) or reckless (committing fixed costs against cancelable revenue) is precisely the bull/bear axis. Archetype: founder-led, mission-narrative, capital-maximalist — high-variance for a company about to face public-market scrutiny.

Forensic Red Flags

`, unaudited per public sources.]

Accounting/structural risks (no audited statements exist, which is itself the headline risk for a company filing toward an S-1):

1. Circular revenue recognition. Supplier-investors (NVIDIA, Amazon, AMD, Oracle) sit on both sides — capital in, GPU/cloud purchases out, equity/warrants exchanged. Auditors and an S-1 review will scrutinize how much "revenue"/backlog is effectively round-tripped vendor financing.
2. Off-balance-sheet-style commitments. $1.4T of multi-year compute obligations vs. a tiny revenue base — the gap between committed capex and recurring, cancelable revenue is the central forensic question.
3. Going-concern-adjacent burn. ~$27B 2026 cash burn funded round-to-round; any funding-market stall is existential.
4. Governance opacity — nonprofit-controls-PBC structure, AGI-verification clause that toggles Microsoft's revenue share, CFO reporting changes.

Legal proceedings (web; no 10-K Item 3 exists):

• NYT (+ other publishers) copyright suits vs. OpenAI/Microsoft — ongoing; OpenAI denies; two federal judges have separately ruled model training "highly transformative"/fair-use, a favorable trend but not a final resolution.
• Elon Musk / xAI litigation — OpenAI reportedly won a recent trial; depositions (Altman, Brockman) are being produced into the copyright cases.
• FTC/DOJ: the targeted search surfaced no confirmed material federal enforcement action as of June 2026, though governance/insider-selling/SEC-review chatter exists around the pre-IPO posture. Treat as unconfirmed.

Verdict on the books: No audited financials and no SEC enforcement record exist — verified that OpenAI has no CIK (so EDGAR LR/AAER is N/A) and via web search as of 2026-06-18. The material forensic risk is not a discovered fraud; it is the absence of audited statements beneath $1.4T of commitments and a circular financing web — exactly what an S-1 audit will stress.

---

Phase D — Project & stress-test

IPO-Readiness & Path-to-Tradeable (overlay swap for "Forward Projection")

private-watch.json carries ipo_readiness: 4 (pre-IPO/secondary-active) and catalyst "for-profit restructuring toward capital access." That reading is now stale on the upside: the restructuring completed (Oct 2025) and a confidential S-1 was reportedly filed in early June 2026 with Goldman Sachs and Morgan Stanley — which moves the name to effectively readiness 5 (S-1 filed / IPO imminent).

Milestones that unlock the listing — and their state:

• ✅ For-profit/PBC restructuring (done, Oct 2025)
• ✅ Microsoft agreement modernized (Apr 2026, removed restrictive terms)
• ◐ Confidential S-1 filed (reported, early Jun 2026)
• ✗ Audited financials sufficient for public markets (the CFO's stated gating concern)
• ✗ Internal alignment on timing (Altman Q4-2026 vs Friar 2027 — unresolved)

Estimated window: prediction markets imply ~76% chance of IPO by 31 Dec 2026; Altman targets Q4 2026 at up to ~$1T; the CFO pushes 2027. Base case: a listing in the Q4 2026 – H1 2027 window at a $1T+ valuation, contingent on clean enough audited financials and funding-market stability. The binary that actually matters is not EPS — it's does the IPO (or continued private capital) close before the funding treadmill — ~$27B/yr burn against $1.4T of commitments — outruns investor appetite?

Per --watchlist rules, no Brier forecast is logged (breadth loop). The scoreable binary, were one logged: "OpenAI completes an IPO before 2027-06-30" — base ~55–65%, handicapping Altman's push against the CFO's caution and audit gating.

Write-back: private-watch.jsonopenai.dossier should be set to this file and ipo_readiness bumped 4→5 (S-1 reportedly filed). Per strict wave boundaries, I am not editing that file in this unattended run — flagging it for the central post-wave update.

Bull vs Bear

Bull case. OpenAI is the category-defining consumer AI brand (900M WAU), the revenue leader (~$25B run-rate, ~$2B/mo), growing enterprise 4x in six months, with tier-1 capital access and a completed governance structure that clears the path to a ~$1T IPO. It has locked up scarce compute early (Stargate, $1.4T) — if AI demand compounds, owning that capacity ahead of rivals is the winning move, and operating leverage on a fixed model-training cost base could flip the P&L hard once revenue laps the spend (the company guides to 2030s profitability). The Ive device + Sora + agents open new monetization surfaces. If you believe AI demand is under-, not over-, estimated, OpenAI's pre-committed scale is the asset.

Bear case (permanent-impairment risks).

1. The unit economics may not close. ~65% of Q1 2026 revenue went to burn; $1.4T of fixed commitments sit against cancelable, increasingly price-competed revenue. If token prices keep falling and enterprise share keeps leaking to Anthropic, the spend never earns its return.
2. Anthropic is winning the part that compounds. Anthropic passed OpenAI in valuation on ~2x the revenue run-rate and leads enterprise API — OpenAI risks becoming the consumer leader (lower-margin, churnier) while losing the enterprise/developer franchise (stickier, higher-margin).
3. Circular financing + funding dependence. The buildout is underwritten by suppliers who are also investors; a single funding-market stall (note the reported SoftBank liquidity-crunch concerns) could break the chain.

Pre-mortem (18 months out, thesis broke): the IPO slipped past 2027 on audit/governance problems; a funding round priced flat-to-down as Anthropic's enterprise lead widened and token economics compressed; one anchor (SoftBank) pulled back under its own leverage; the $1.4T commitments converted from "scarce-asset moat" to "fixed-cost millstone."

Are the marks too high? On a revenue multiple OpenAI (~34x) is richer than Anthropic (~21x) while growing enterprise slower — so relative to its closest comp, yes, the mark looks full.

Contrarian view of what the market refuses to see: the consensus treats OpenAI as the AI franchise. The data says it is becoming the B2C AI franchise while losing the B2B/developer war — and B2C AI is exactly where price competition, churn, and Google's distribution bite hardest. The market is paying an enterprise-winner multiple for a company whose enterprise position is eroding.

Devil's Advocate (short-seller)

Dismantling the bull case:

• Revenue concentration & fragility: consumer subscriptions (cancelable monthly) are the largest line; API is under direct price war; the "$1.4T moat" is a $1.4T liability if demand or pricing disappoints by 20–30%.
• The moat bulls overrate: "ChatGPT brand" is real but didn't stop app share falling 69%→45% in a year. Brand ≠ switching cost when the competitor is one tab away and Google owns the defaults.
• Most dangerous competitor bulls underrate: Anthropic — higher valuation on 2x revenue, enterprise-API leader, and it's also nearing an IPO. The bull narrative ("OpenAI is the leader") is already factually wrong on valuation and enterprise share.
• Worst capital allocation: committing $1.4T against a $25B revenue base via circular supplier-investor deals (AMD warrants for GPU purchases; NVIDIA/Amazon investing then selling in compute) — accounting that an S-1 audit will not flatter.
• What must hold for a $1T IPO mark: AI demand keeps compounding and OpenAI captures enough of it and token economics stop deflating and funding markets stay open and the audit is clean and the CFO/CEO resolve timing. That's a long conjunction.
• 20–30% growth disappointment: with burn at ~57–65% of revenue, a growth miss doesn't dent margins — it threatens runwayRunwayHow long the cash lasts at the current rate of spending. It shortens the moment spending rises, which is why a figure taken from a quiet quarter flatters. and forces a down-round or a rushed IPO. Q1 2026 already missed internal targets.
• Single permanent-impairment scenario (most plausible): a funding-market stall (e.g., SoftBank retrenches under its own leverage) mid-buildout, forcing OpenAI to honor fixed compute commitments without the capital to do so — a down-round at best, a distressed restructuring at worst. Plausibility: moderate and rising, the single biggest risk to the equity.

Management Questions (ordered by information value)

1. Of the ~$1.4T in announced compute commitments, how much is firm take-or-pay vs. cancelable/optional, and what is the cash schedule by year?
2. What is the audited 2025 net loss and 2026 run-rate revenue — and when will audited financials be available for the S-1?
3. How much of reported revenue/backlog involves supplier-investors (NVIDIA, Amazon, AMD, Oracle) on both sides of a transaction, and how is that recognized?
4. Why does the CFO favor a 2027 listing, and what specifically is unfinished — is the disagreement about readiness or about valuation?
5. Enterprise API share has shifted toward Anthropic — what is the concrete plan to defend the developer franchise, and what's the current net-revenue-retention by cohort?
6. What is the path and timeline to positive free cash flow, with the key assumptions (token pricing, mix, utilization)?
7. If the next funding round priced flat or down, what is the contingency for honoring the compute commitments?
8. What are the AGI-verification clause's exact triggers, and what happens to the Microsoft revenue share and IP rights when invoked?
9. What is gross margin by segment (consumer / API / enterprise) today and at scale?
10. What is the realistic ship date, BOM, and revenue model for the Ive hardware device (the court filing says no earlier than ~Feb 2027)?
11. How exposed is the buildout to a single anchor's liquidity (SoftBank), and what diversification of capital sources exists?
12. What is insider/employee selling activity into secondaries, and what does the cap table look like post-IPO (lockups, Foundation control)?
13. How is the nonprofit Foundation's control reconciled with public-shareholder fiduciary duties post-IPO?
14. What is the customer concentration in the enterprise book, and the churn/expansion trend?
15. How does management think about the risk that OpenAI becomes the consumer leader but the enterprise loser, and what would change that trajectory?

---