Bloomberg confirmed late Friday that Anthropic is on track to close a new $30 billion funding round as early as the week of May 26, at a pre-money valuation above $900 billion. Sequoia Capital, Dragoneer Investment Group, Altimeter Capital, and Greenoaks Capital Partners are co-leading with roughly $2 billion checks each. Founders Fund and General Catalyst are following on.
If the deal closes on those terms — and the people involved say it will — Anthropic surpasses OpenAI's $852 billion valuation from March and becomes the most valuable private AI company in the world. For roughly nine weeks, OpenAI held the most valuable private company title in tech history. The crown is moving.
Three months ago this seemed structurally implausible. OpenAI had 5x the revenue, 3x the brand recognition, and twice as many enterprise customers. What changed is that Anthropic delivered the one data point nobody else in the frontier-model game has delivered: profitability.
A Year of Compounding That Most Investors Did Not See Coming
| Quarter | Anthropic Revenue | QoQ Growth | Annualized Run Rate |
|---|---|---|---|
| Q4 2025 | ~$2.3B | — | ~$9B |
| Q1 2026 | $4.8B | +109% | ~$19B |
| Q2 2026 (projected) | $10.9B | +127% | ~$43B |
| Q3 2026 (analyst est.) | $18-22B | +65-100% | $70-88B |
The Sacra estimate of $43 billion in annualized revenue for April 2026 is the number that makes the $900 billion valuation arithmetic work. At $43B annualized, the implied revenue multiple is roughly 21x. Three months ago, when Anthropic raised the first $30B round at $380B post-money on roughly $19B annualized, the multiple was 20x. The valuation has more than doubled while the multiple has barely moved. That is what happens when the underlying ARR doubles in a single quarter.
For context: I covered the $5B+ Anthropic ARR milestone earlier this year — that number was the quarterly run rate. The annualized figure has moved from $9B at year-end 2025 to $43B in April. The compounding is real and it is happening faster than any model lab has ever grown.
The First Profitable Frontier Lab
The single most important number in the deal is not the $900 billion valuation. It is the $559 million.
Anthropic is projecting its first ever operating profit in Q2 2026 — roughly $559 million on $10.9 billion of revenue, for an operating margin of 5.1%. That number does not sound large until you remember what every other frontier model lab is reporting: OpenAI is still burning multibillion-dollar amounts annually despite $5.7B in Q1 revenue. xAI is burning faster. Mistral, Cohere, and the smaller foundation labs are all deeply unprofitable.
A 5.1% operating margin is thin, and the critics on financial AI Twitter have already pointed out that the calculation excludes equity compensation and probably some compute amortization that more conservative accounting would include. They are right. The bear case is that "profitable Anthropic" is a marketing construction designed to support a $900B valuation, not a GAAP truth.
The bull case is that even a marketing-construction profit number is a year earlier than anyone else in the category will get there. OpenAI has not given a profitability timeline at all. xAI is operating on Musk's "burn whatever it takes" philosophy. Mistral is racing to ARR not margin. If Anthropic prints $559M in operating profit in Q2 — even by aggressive accounting — they are the only frontier lab that can credibly look enterprise customers in the eye and say "we are not going away if the funding environment turns."
That credibility is worth tens of billions of dollars in enterprise contract pricing power.
Where the Profit Actually Comes From
Three structural shifts compounded to deliver Q2 profitability faster than anyone modeled:
- Enterprise mix at 80% of revenue. Anthropic now has 300,000+ business customers, with the over-$1M-per-year cohort growing from ~500 to 1,000+ in just two months. Enterprise contracts carry 70%+ gross margins versus consumer pricing closer to 40-50%. The mix shift alone explains roughly 600 basis points of margin improvement quarter over quarter.
- Claude Code became a multi-billion-dollar product line. Claude Code went GA in May 2025, hit $1B in annualized revenue by November 2025, $2.5B by February 2026, and is on track for $4B+ by mid-year. That is a developer-tools product line generated almost entirely from existing enterprise relationships, and the unit economics are dramatically better than chat-based usage because it bills directly into engineering budgets that are 10x larger than knowledge-worker SaaS budgets.
- Cost-side moves on inference. The rumored Microsoft Maia 200 deal I covered yesterday is not the first cost-side win Anthropic has booked. AWS Trainium volumes scaled aggressively through Q1 with reportedly 25-30% cost-per-token improvements over equivalent NVIDIA H200 inference. The Maia 200 deal, if it closes, adds another 20-30% on top. Cost structure improvements are what turn $10.9B of revenue into $559M of profit instead of $200M.
The 30x Multiple Versus Public Comps
A $900B valuation on $43B annualized revenue is 21x annualized — and if you instead use the forward 12-month run rate that VCs actually model against (probably $70-90B by Q3-Q4), the implied multiple compresses to roughly 12-15x forward revenue. That is not far from where public software companies with strong AI exposure trade.
I broke down how VCs are valuing AI companies in this cycle — the median Series A AI startup trades at 40-80x ARR, and frontier model labs have been pricing closer to 60-100x. Anthropic at 12-15x forward revenue is actually a discount to the cohort it sits in, because the growth rate is so extreme that any reasonable forward projection makes the trailing multiple look modest.
The bear case for the multiple is gross margin durability. If Google's aggressive Gemini 3.5 Flash pricing forces Anthropic to cut Claude API prices 25-30% over the next 12 months, the unit economics get tighter and the profitability story becomes a narrower window. The bull case is that enterprise contracts price on value delivered rather than per-token, so the Google API price war affects consumer plans and competitive deals but not the 1,000+ customers paying $1M+ annually for managed Claude deployments.
What This Does to OpenAI
OpenAI is now in the awkward position of having the largest revenue ($23B annualized roughly), the most users (900M+ weekly), and the most distribution (Apple Intelligence, Microsoft Copilot) — but the second-most-valuable valuation.
The pressure for OpenAI to demonstrate its own path to profitability just got severe. Sam Altman has been signaling "profitability by 2029" as a reasonable target. After Anthropic's Q2 print lands, that timeline looks weak. Expect OpenAI to either (a) accelerate its own profitability guidance to 2027-28, (b) announce IPO timing to give shareholders a different liquidity story, or (c) raise another massive round at a higher valuation to regain the most-valuable-private-AI-company crown.
The IPO clock matters here. I covered the OpenAI $13B+ ARR trajectory earlier. The market would absolutely buy an OpenAI IPO in 2026 at $1.2-1.5 trillion. But going public would force GAAP profitability disclosure, which would force the brutal comparison with Anthropic that the private market is now starting to make on its own.
What This Does to AI VCs
The Sequoia, Dragoneer, Altimeter, and Greenoaks $2B checks each represent the largest single-check growth-stage AI investments in history. The arithmetic of these LP returns is unusual: if Anthropic IPOs at $1.5T within 2 years (plausible) and goes to $2.5T within 5 years (also plausible), the $2B checks at $900B turn into $3-5B back inside 24 months.
For the limited partners in those funds — university endowments, sovereign wealth funds, pension funds — the Anthropic position alone covers fund returns for a vintage. Sequoia's growth fund probably 3-4x's its entire commitment from this single position if the bull case plays out.
I covered the alternative AI funding paths earlier — sovereign wealth, strategic corporate venture, retail offerings — because the traditional VC playbook gets squeezed when individual rounds get this large. The Anthropic round is interesting precisely because it is still being done by traditional growth VCs at a scale that two years ago would have required SoftBank-style mega-funds. The largest VCs are growing into the scale of capital that AI now requires.
The Macro Lens
The Anthropic round and the recent NVIDIA, OpenAI, and SpaceX-xAI milestones are individual data points in a single trend. The total private AI valuation as of this week:
- OpenAI: $852B
- Anthropic: $900B (pending close)
- SpaceX (including xAI): $1.25T (the AI portion is roughly $300B)
- xAI standalone equivalent: ~$250B pre-merger
- Mistral, Cohere, others: $20-40B combined
That is roughly $2.3 trillion in private AI valuation concentrated in a handful of foundation labs. The trillion-dollar AI race is no longer about who reaches $1T first. It is about which two or three labs survive the consolidation phase that comes when the public markets start absorbing this scale of valuation through IPOs.
The 2-Year Outlook
Three things determine whether Anthropic's $900B valuation looks brilliant or absurd by mid-2028:
- Whether Q3 2026 confirms the profitability. A single profitable quarter is a data point. Two in a row is a trend. Four in a row is a moat. Watch the Q3 disclosure carefully — if revenue keeps doubling but margin holds 5%+, the multiple expansion thesis is intact.
- Whether Claude Code keeps compounding. A multi-billion-dollar developer tools product line is the single most underpriced thing in the Anthropic story. If Claude Code hits $8-10B annualized by year-end, Anthropic becomes the dominant force in developer AI tooling and the valuation gets cheap.
- Whether the IPO window opens favorably. A 2027 Anthropic IPO at $1.5-2T puts the round investors in clean profit. A delayed IPO into 2028 in a weaker market is where things get interesting. Watch the rate environment and the public AI comps (Palantir, CrowdStrike) for the leading indicators.
The Honest Take
Anthropic just demonstrated that frontier AI labs can become profitable at scale on a 5-year timeline, not the 10-year timeline most investors had modeled. That single fact is what justifies the $900 billion valuation, not the revenue growth rate alone. Every AI investor and every AI competitor now has to rerun their financial models with "profitability in year 3-4" as a possible scenario instead of "profitability after IPO at some unclear future date."
The frontier-model game in 2026 is not about who builds the best model anymore. It is about who builds a model business that compounds without burning the capital base. Anthropic just punched its ticket. Everyone else has 12-18 months to follow or fall behind permanently.
