QWE AI Academy Here’s the surprising bit: nobody – not even Paul Graham, YC’s co-founder – seems to have a clean, public answer for how much of OpenAI Y Combinator actually owns. The number making the rounds (0.6%) came from a tipster, not a filing. And yet that fuzzy little percentage is worth somewhere north of $5 billion.
That gap between “big number everyone’s quoting” and “actually verifiable fact” is the real story. If you’re building, investing, or just trying to understand who owns the AI companies shaping your tools, this matters. So let’s use the YC/OpenAI stake as a hands-on case study for reading private cap tables – without falling for the headlines.
The problem: AI cap tables are mostly invisible
The number going viral right now: Y Combinator owns about 0.6% of OpenAI, and at OpenAI’s $852 billion valuation (confirmed in OpenAI’s March 31, 2026 announcement), that works out to over $5 billion. John Gruber broke it on Daring Fireball – after, in his words, asking around.
That’s the issue. A multi-billion-dollar stake in the most-talked-about company on earth, and the best public source is “a little birdie.” Carta’s documentation (as of 2025) is blunt about why: for private companies, the cap table is a highly confidential document, typically shared only with board members, investors, and lawyers during specific events like fundraising or legal reviews.
Translation: unless you’re inside the deal, you’re guessing. Most articles about “who owns OpenAI” stitch together leaks, secondaries, and SEC fragments into a picture that looks confident but isn’t.
Why the existing coverage falls short
Same three facts, every piece: Altman ran YC, YC Research seeded OpenAI in 2016, the stake is worth ~$5B today. None of them teach you to verify any of it. None flag what’s still unknown.
The actual reporting is messier. TechCrunch reported in May 2024 that YC’s late-stage fund put $10 million into OpenAI’s for-profit subsidiary – and Graham said he wasn’t aware of the investment at the time. So “YC’s stake” isn’t one stake. It’s at least two: the original YC Research seed position from 2016, plus a later $10M check from a separate fund that the co-founder didn’t know about.
Headlines collapse all of this into a single percentage. That’s where readers get fooled.
How to actually investigate a private AI company’s cap table
Here’s a four-step process you can run on any private AI company. We’ll use OpenAI as the worked example.
Step 1: Anchor on the latest official valuation
Skip the rumor mills. OpenAI’s March 31, 2026 post confirmed a $122 billion funding round at a $852 billion post-money valuation – that’s the source of truth. Anything older is stale.
Step 2: Identify named investors and their dollar amounts
Cross-reference the official post with reputable financial press. Amazon committed up to $50 billion, Nvidia $30 billion, SoftBank $30 billion – those are the anchor checks per Bloomberg and CNBC reporting on the round. Smaller LPs and earlier investors fill in around them.
Step 3: Calculate the implied percentage – but flag what you don’t know
Dollar invested ÷ post-money valuation gives a rough ownership floor. But it’s a floor, not a fact. Two reasons:
- Anti-dilution and preference terms. Investors negotiate clauses that change effective ownership in down rounds or exits.
- Non-cash investment. Here’s the one that catches most people: Nvidia’s stake. A 2026 reconstruction by StartupHub puts Nvidia at roughly 3.47% of OpenAI, valued at approximately $29.6 billion against a cost basis of $30 billion – technically 1.0x, barely breaking even. But Jensen Huang isn’t losing sleep over it, and the footnote explains why: a “significant portion” of that investment came in the form of GPU compute credits, not cash. Same percentage on paper, completely different economic reality underneath.
The percentage on a leaked cap table doesn’t tell you what anyone actually paid.
Step 4: Treat reconstructed cap tables as estimates, not gospel
You’ll find Twitter threads and finance blogs publishing “the OpenAI cap table” as if it’s a leaked document. It almost never is. They’re reconstructions built from public funding announcements plus inference – useful, but treat them as roughly 80% confidence, not 100%.
Pro tip: When you see a private-company ownership percentage cited online, ask one question: “Is this from a primary source (the company, an SEC filing, a court document) or a reconstruction?” If it’s a reconstruction, treat the number like a weather forecast – directional, not exact.
Real-world example: applying the method to YC’s stake
Run the four steps on the YC/OpenAI question:
- Latest valuation: $852B post-money, March 31, 2026. Confirmed by OpenAI directly.
- Named investors with dollar amounts: Amazon, Nvidia, SoftBank in the latest round. YC is not in this round – its position dates to 2016 and a later $10M from the late-stage fund.
- Implied ownership: Gruber’s source says 0.6%. We can’t confirm this against any primary document. The 0.6% × $852B = $5.1B math works, but there is one source and one source only.
- Reconstruction risk: High. Reconstructed spreadsheets have been circulating online showing who owns OpenAI at its current valuation, what they paid, and what their stakes are worth. Useful starting point – not a filed document.
So the honest answer: “YC owns roughly 0.6% of OpenAI, worth approximately $5B at the current valuation, per a single unverified source. The exact figure could be lower (dilution from later rounds) or higher (if early shares carried pro-rata rights that were exercised). No one outside YC and OpenAI’s legal team knows the precise number.”
That’s a more useful sentence than what most outlets are running.
The math gotcha nobody’s talking about
OpenAI reportedly restructured in late 2025 – splitting ownership between a nonprofit foundation and a public benefit corporation. The OpenAI Foundation’s stake, per a 2026 reconstruction, sits at 25.80%, worth approximately $219.8 billion at the March 2026 valuation. That would make it potentially the largest charitable foundation stake in corporate history.
If YC’s 0.6% was measured before that restructuring, the figure today might sit in a different legal entity, with different rights, different payout terms, different IPO mechanics. Nobody has clarified which. Worth keeping in mind before you bet a worldview on the headline figure.
Think of it like measuring a river’s width before and after a dam splits it into two channels. The water is the same, but “width” means something different depending on which channel – and which side of the dam – you’re measuring from.
Pro tips for tracking AI company ownership
- Bookmark the company’s own newsroom. Most leaks turn out to either confirm or contradict an official post that came out months earlier.
- Read the secondary sale reports. Secondaries (employees selling existing shares) reveal valuations between primary rounds and often surface more detail than press releases.
- Track the foundation/PBC split for any restructured AI company. Anthropic, OpenAI, and others have unusual structures that break standard cap-table assumptions.
- Don’t conflate “stake worth X” with “will receive X.” Liquidation preferences, lock-ups, and conditional tranches mean paper value rarely equals exit value.
FAQ
Is the 0.6% figure for YC’s OpenAI stake confirmed?
No. One anonymous source, via John Gruber. Best public estimate available – not a verified number.
Why does any of this matter to me if I’m not an investor?
Say you’re choosing which AI tool to build a product on. Knowing that YC has a multi-billion stake in OpenAI tells you something about why YC-backed startups often default to the OpenAI API – there’s an incentive structure underneath, not just a technical preference. The same logic applies when Anthropic-friendly accelerators push Claude, or when an investor in your stack also funds your competitor. Cap tables are alignment maps. Reading them changes how you interpret recommendations, partnership announcements, and “unbiased” tool comparisons.
Where can I learn the basics of cap table math before tackling private AI companies?
The American Bar Association’s primer on cap table math covers pre-money, post-money, and dilution in plain language – one or two manual calculations on a simple example and the OpenAI-scale numbers stop feeling magical. They’re just bigger versions of the same arithmetic.
Next action: Pick one AI company you use daily (Anthropic, Perplexity, Mistral – your choice). Run the four-step method on it this week. Anchor on their latest official round, list the named investors, calculate implied ownership, and write down what you can’t verify. You’ll know more about the AI industry’s incentive structure in 30 minutes than most people who use these tools every day.