Two frontier AI labs are simultaneously pursuing IPOs for the first time. The question is what happens when the companies powering the agent economy have to show public market investors their actual numbers — not projections, not narrative, but 90-day GAAP financials, gross margin breakdowns, and compute cost transparency.
Anthropic has the early lead. The company confidentially filed its S-1 with the SEC on June 2, 2026, coming off a $965 billion post-money valuation from its May Series H round — $65 billion raised from Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. It is targeting an October 2026 Nasdaq listing under the ticker ANTHR. The numbers backing that timeline: a $47 billion revenue run rate as of May 2026, with 80% derived from enterprise customers. By Q1 2026, Anthropic had surpassed OpenAI in both overall LLM revenue share — 31.4% to 29% — and enterprise market share at 34.4% versus 32.3%, according to the Ramp AI Index. The enterprise-first bet is yielding results. Breakeven target: 2028.
OpenAI is hesitating. The company filed its own confidential prospectus roughly one week after Anthropic, around June 8, but advisers have presented the board with a choice: wait until 2027 for a $1 trillion valuation, or adjust pricing downward now. The New York Times reported on June 25 that SpaceX’s post-IPO jitters were a factor in the deliberation. OpenAI maintains an $852 billion private valuation and is targeting the $850 billion to $1 trillion range, but the market has shifted under its feet. Revenue run rate sits at approximately $25 billion — roughly half of Anthropic’s — and remains heavily anchored in consumer-led growth via ChatGPT. Profitability target: 2030, two years behind Anthropic’s timeline.
The structural question for both labs is the compute ownership gap. Neither company owns the infrastructure required to host its models. Both pay heavily for capacity — Anthropic has announced a $50 billion US AI infrastructure build-out for 2026 — which creates margin pressure that public investors will scrutinize in ways private markets never did. Once these companies enter the public domain, every compute commitment, every stock-based compensation grant, every gross margin shift becomes visible in quarterly 10-Q filings. Wall Street must learn to value businesses where the primary cost of goods sold is tied to token volume — a metric that has no standardized valuation framework yet.
SpaceX offers a cautionary precedent. After its June 12 IPO at $135 per share — raising $75 billion at a $1.77 trillion valuation — the stock peaked at $192 by mid-June, then suffered a 16% single-day sell-off in late June. As of July 9, SPCX trades at approximately $147.78, still above its IPO price but well below its debut high. The volatility illustrates how quickly narrative-driven valuations can correct when the market starts pricing in fundamentals. As one PitchBook analyst noted via CNBC on June 5: “The 2026 window either becomes the most consequential IPO cycle since the dot-com era or the most expensive lesson in narrative-versus-fundamentals that public markets have ever taught.”
The agent economy implications are direct. These are the two labs whose models power the majority of autonomous agent workflows — from coding assistants to enterprise automation. Public reporting will force them to clarify how they intend to move from massive infrastructure spending to sustainable profitability. OpenAI’s 2030 target and Anthropic’s 2028 goal are no longer internal projections; they become benchmarks public investors will hold them to every quarter. Both operate as Public Benefit Corporations, adding governance complexity — particularly for OpenAI, whose conversion from non-profit status remains contentious with regulators.
The divergence between enterprise stability and consumer scale will define how these offerings land. Anthropic’s 80% enterprise concentration offers more predictable revenue but less explosive growth. OpenAI’s consumer-led model delivers scale — ChatGPT remains the dominant interface — but with more volatility. The part that gets hidden behind the headline valuations is the infrastructure dependency: neither lab owns its compute, both are burning billions to scale, and both will now have to explain those numbers to a market that has not yet figured out how to value a token.