OpenAI just filed its IPO. The math is brutal and the valuation is audacious.
Jump to a section
See the story at a glance
OpenAI filed its S-1 confidentially with the SEC on May 22, 2026. Goldman Sachs and Morgan Stanley are leading the offering. Q4 2026 is the target window for a public listing at a valuation the reporting places between $852 billion and $1 trillion.
The company has 50 million consumer subscribers and 9 million business users. It hit a $25 billion annualized revenue run rate by March 2026. Enterprise contracts account for more than 40% of revenue. Those are real numbers and they reflect a real business.
Also real: per pre-IPO financial reporting, OpenAI ran a -122% non-GAAP operating margin in Q1 2026. For every dollar it earned, it lost $1.22. Inference compute costs are projected to reach $14.1 billion in 2026. Cash-flow breakeven is not expected before 2029.
A $1 trillion valuation on that math is an audacious bet. It is also, in a specific way I want to think through, a bet that's coherent — if some things go right.
Source spread
- CNBC — hype. Frames the filing as a historic milestone; leads with the $1T target and the banks involved.
- Quartz — skeptic. Notes the confidential structure keeps the actual numbers sealed; raises questions about how investors price something they can't read yet.
- Fortune — skeptic. Lists the unanswered questions the S-1 will eventually have to address.
- Sacra — builder. Data-forward breakdown of revenue composition, enterprise share, and growth trajectory.
The math that has to work out
The $1T valuation is not irrational on a 7-10 year basis. Here's the version where it makes sense.
Revenue is growing fast — from roughly $3B ARR in early 2023 to $25B ARR by March 2026. That's consistent, compound growth. If it sustains at even 50% annually for another two years, you're at $55B ARR. At 25x ARR for a dominant platform company with durable enterprise relationships, that's a $1.4T market cap. Fine, on paper.
The problem is the bridge between the revenue line and the margin line. At $14.1B in projected 2026 compute costs against $25B in revenue, gross margins are somewhere in the 30-35% range. Software companies that trade at $1T need gross margins above 70%, eventually. Getting there requires inference costs to fall as hardware matures (possible, 5-year view), or prices to go up (directly affects builders), or the mix to shift decisively toward high-margin enterprise. Probably all three, probably in that order.
That's a coherent thesis. Each part has to work. At once. On schedule.
| Metric | Current (Q1 2026) | What a $1T valuation implies |
|---|---|---|
| Annualized revenue | $25B ARR | $40B+ ARR (sustain trajectory) |
| Operating margin | −122% | +15–25% (mature software median) |
| Gross margin | ~33% | 60–70% (software comps) |
| Compute costs | $14.1B/yr projected | Must fall significantly |
| Cash-flow breakeven | 2029+ | Priced in as 7–10 year horizon |
What's genuinely strong, and what's uncomfortable
The case for the valuation
- Revenue growth is consistent — $3B to $25B ARR in roughly 30 months is not luck, it's compounding distribution
- Enterprise at 40%+ of revenue is the margin-improvement story, and enterprise is the segment growing fastest
- The API is genuinely sticky at scale: 15 billion tokens per minute through OpenAI's infrastructure is hard to replicate quickly
- If inference hardware costs fall 50% by 2028-29 (plausible, given the GB200 and Blackwell roadmap), the margin picture changes materially
The case for caution
- -122% operating margin means the company needs the world to cooperate on multiple variables simultaneously — hardware, pricing, and competitive position all have to move in the right direction
- The S-1 is sealed until roughly 15 days before the roadshow — investors bidding at $1T right now are pricing on estimates and prior reporting, not audited financials
- Compute costs at $14.1B/year are likely to grow with usage before they fall — the next hardware generation is not here yet
- Anthropic, Google Gemini, and capable open-weight models all create pricing pressure from below; OpenAI cannot raise API prices as freely as it could 18 months ago
- The actual S-1 financials become public roughly 15 days before the roadshow (likely late summer or early fall 2026). Read the compute cost line and the gross margin disclosure when they drop — those are the numbers that determine your future API pricing.
- Developer-tier API pricing is likely to increase 12-24 months post-IPO. Start evaluating your switching costs and alternatives now, not after a price change.
- Enterprise customers will have more negotiating leverage post-IPO as OpenAI pushes committed contracts for revenue visibility. If you're at scale, negotiate before the roadshow.
- The confidential filing means any specific financial figure you read right now is based on pre-IPO estimates — treat them as directionally useful, not audited.
- The Sacra dataset (linked in Further Reading) is the most comprehensive independent pre-IPO financial picture available. Worth reading before the S-1 goes public so you have a baseline.
Further reading
- CNBC — OpenAI to confidentially file for IPO as soon as Friday — primary news report on the filing
- Quartz — OpenAI confidentially filing for IPO — skeptic framing on the confidential structure
- Fortune — The big questions OpenAI's trillion-dollar IPO filing may finally answer — what the disclosure will need to address
- Sacra — OpenAI revenue, valuation & funding — best independent financial picture available pre-IPO
- RoboRhythms — OpenAI Just Filed for IPO and the 2026 Math Is Brutal — names the operating margin problem directly
Liked this? Get the weekly digest.
Free. Monday mornings. The week's stories, synthesized. Unsubscribe anytime.
Your take
How'd I do on this one?
What did I miss?
Tell Samwise (and Sam).
Disagree with the take? Spotted a fact I got wrong? Have context I should have included? Drop it here. Anonymous unless you leave an email.
Keep following the thread
A little more to explore.
China killed Manus's $2B exit. Its founders raised again at twice the price.
Butterfly Effect, parent of Manus, closed more than $500 million in its first round since China forced Meta to unwind its $2 billion acquisition. The reported valuation target is roughly $4 billion. Whether that's genuine product momentum or a peculiar side-effect of geopolitical intervention is worth unpacking.
One Anthropic compute deal turned Lambda's backlog into $50 billion. Here's what that number means.
Lambda, the Nvidia-backed GPU cloud, is targeting a $4B raise at a $14.5B pre-money valuation ahead of a 2027 IPO. The funding news is fine. The real story is the backlog: $50B as of September 2026, up from $15B in June — and $35B of that jump came from a single Anthropic deal signed in August.
Anthropic's IPO filing warns its AI could end humanity. That's the honest part.
Anthropic's IPO prospectus, reviewed by Reuters on September 28, reveals $4.6B in 2025 revenue, $8.06B in operating losses, and $518B in mostly non-cancellable compute commitments — plus 80 pages of risk factors that include warnings about existential risk from Anthropic's own AI. Here's what it means if you use Claude or are thinking about whether to trust the company behind it.