2026 is shaping up to be the year AI went public — literally. Anthropic filed confidentially for an IPO. OpenAI is targeting $40 billion. SpaceX is aiming for $75 billion. Those aren't typos, and they're not competing with each other in the traditional sense. They're racing ahead of each other into public markets while the window is still wide open.
Anthropic: The Quiet Filing
Anthropic filed its IPO paperwork in stealth mode — exactly how you'd expect from a company that built its brand on being thoughtful about AI safety. The news came via the New York Times, confirming what had been rumored for months.
What makes this interesting: Salesforce holds a $5 billion stake — roughly 3-4% based on typical pricing. They first invested about $50 million in early 2023. That's a 100x return on paper, assuming the IPO prices even modestly above the last round.
Salesforce's $5B stake from a $50M bet in 2023 is exactly the kind of asymmetric upside that makes IPOs inevitable.
If Anthropic goes public, it will face the same pressure Cook and Ilya Sutskever talked about on the Investor Call — the tension between staying mission-driven and delivering returns to public market shareholders. That's the bind everyAI startup faces when size forces optics.
OpenAI: The $40 Billion Question
OpenAI is reportedly aiming for the largest IPO in history — $40 billion. That's more than Arm, more than Visa, more than almost anything tech has ever produced. The valuation reflects actual revenue traction (reportedly approaching $4 billion annually), enterprise adoption, and the o1 reasoning models shipping to production.
But Florida just sued Sam Altman personally, alleging deceptive trade practices, negligence, and public nuisance. The AG wants Altman's personal assets on the line. That's a novel theory — and if it gains traction, it could complicate the IPO narrative around governance and risk.
SpaceX: The $75 Billion Moonshot
SpaceX isn't an AI company, but it's in the race. Up to 5% of Class A shares will go to employees and their networks. 60%+ have extended lock-ups — meaning they're betting on longer-term value rather than flipping on day one.
The interesting link: Jensen Huang confirmed Anthropic, OpenAI, and SpaceX are all first-big users of Nvidia's new Vera CPUs. These companies don't just eat compute — they're reshaping what compute gets built.
Why This Matters
This isn't just about three companies going public. It's about what public markets are willing to pay for the most transformative technology since electricity. Every analyst will have models on growth, retention, margins. None of those models worked for the last generation of AI companies going public — and the ones that did, like Palantir, traded sideways for years before finding product-market fit.
Arm did well. Reddit did well. Snowflake did okay. The difference is those weren't trying to build artificial general intelligence while explaining to the SEC why their compute costs are exploding while revenue chases it.
The smart money watches for how each company structures the offering — dual-class shares, lock-up terms, governance carve-outs. The dumb money shows up on day one and wonders why the stock trades down.
We'll be watching the signal data around earnings, revenue beats, and GPU utilization. That's where the real story comes alive.