Anthropic's $30B round at $900B valuation tops OpenAI

The day the runner-up became the most valuable AI company on earth

Anthropic just agreed terms that vault it ahead of OpenAI, and the entire AI capital stack has to be reread. According to Tech Funding News’ breakdown of the deal, Dragoneer, Greenoaks, Sequoia Capital, and Altimeter Capital have agreed to co-lead a $30 billion round at a $900 billion pre-money valuation, with each firm likely to write a check of $2 billion or more. The round could close this month. At post-money, Anthropic clears the trillion-dollar line — and on the pre-money figure alone, the maker of Claude already prices above OpenAI’s most recent $852 billion post-money mark from earlier this year, per MacObserver’s summary of the Financial Times reporting. The runner-up just stepped onto the top of the podium, and it did so on a single forty-eight-hour news cycle in mid-May.

The structure of the syndicate is half the story. Three of the four co-leads — Sequoia, Dragoneer, and Altimeter — also back OpenAI, which means the venture firms underwriting the duopoly are now publicly arbitraging against themselves. Cryptopolitan’s breakdown of the round framed it bluntly: OpenAI’s own backers are co-leading a deal that prices their other portfolio company down. The signal that sends to the next $30 billion of LP capital is not subtle. When the same partners are willing to mark Anthropic higher than OpenAI, the implicit thesis is that Claude’s revenue surface is now structurally faster-compounding than ChatGPT’s — and the marks are putting capital where the thesis lives.

The valuation jump itself is the kind of move that only happens in genuine paradigm shifts. Anthropic priced its Series G at $380 billion post-money in February 2026, per the company’s own funding announcement. Three months later, the pre-money is $900 billion. That’s a 2.4x mark in roughly a single quarter, which StartupHub’s breakdown of the talks flagged as among the fastest valuation accelerations on record for a late-stage private company at this scale. The pace matters because the round was telegraphed — the Cryptopolitan reporting on the co-lead syndicate gave the market two weeks to absorb the prospect, and the price not only held but tightened. There is no greater-fool dynamic visible in the bid here. The price is the price, and the syndicate is signing.

The stakes for the rest of the AI ecosystem are immediate and structural. Anthropic is reportedly considering an IPO as soon as October 2026, per the GovCon Wire coverage cited below, which would make this round the last private mark before a public-market test that will set comp prices for every other foundation-model company. The MacObserver summary underscored the IPO geometry: a fall listing at or above the $900 billion pre-money would let early Anthropic backers crystallize gains while letting late-stage entrants like Google — which previously committed $10 billion at a $350 billion mark — sit on roughly 2.6x of paper appreciation in three months. Google’s full commitment package, as the GovCon Wire reporting on the funding round details, includes up to $30 billion of follow-on capital contingent on Anthropic hitting performance targets. The hyperscaler’s option is now deeply in the money. So is every operator at the company, and so is every LP who funded the prior round.

The line that drew the line

Two numbers explain why the syndicate moved at this price: $40 billion of annualized revenue, and $2.5 billion of Claude Code. Both are running well ahead of any forecast public market analysts published twelve months ago.

The headline revenue figure is staggering on every axis. Anthropic’s annualized revenue run rate climbed from roughly $9 billion at the end of 2025 to more than $30 billion by April 2026, and is now closer to $40 billion per sources cited in the Tech Funding News breakdown. VentureBeat’s writeup of the $30 billion milestone quoted CEO Dario Amodei calling the trajectory “crazy” — a roughly 80x growth multiple from January 2024’s $87 million ARR, with the actual revenue path overshooting Anthropic’s own internal forecasts by a factor of eight. 24/7 Wall St.’s comparison piece put the number in historical context: Salesforce took 20 years to reach $30 billion in revenue. Anthropic took three. The pace is not just unprecedented in software — it is essentially unmatched in any sector of the American economy.

The product driver under that revenue is mostly one thing, and that one thing is Claude Code. Anthropic’s own Series F funding announcement from September 2025 tracked the agentic coding tool from roughly $500 million in run-rate revenue at that point, and the Crunchbase News coverage of the Series F added that usage was growing more than 10x in three months. Claude Code crossed $1 billion in annualized revenue within six months of public launch and reached over $2.5 billion by February 2026 — a pace the Tech Funding News piece on the new round flags as one of the fastest revenue ramps ever recorded for a software product. The 300,000-plus business customers Anthropic now serves, with large accounts up nearly 7x year-over-year, are the durable demand pool the round is being priced against. This is not consumer subscription froth. It is enterprise procurement.

The competitive geometry against OpenAI is where the price translates into a thesis. OpenAI closed a $122 billion raise at an $852 billion post-money valuation earlier this year, per the same MacObserver and Tech Funding News reporting cited above, and is widely reported to be running at a revenue rate in the $20 to $30 billion zone — a band confirmed by multiple syndicate sources, with the most cited figure landing around $24 billion. If Anthropic clears $40 billion of ARR this quarter and OpenAI sits at $24 billion, the revenue-multiple comparison is suddenly upside-down: Anthropic priced at 22.5x ARR pre-money, OpenAI at roughly 35x ARR post-money. The maker of Claude is cheaper per dollar of revenue than the maker of ChatGPT, even at $900 billion. That is the proprietary inference the syndicate is paying for, and it inverts the post-ChatGPT-launch narrative that has framed the entire 2023-2025 cycle. The enterprise-versus-consumer split between the two labs has reshaped which mix the market is willing to pay a premium for.

The Claude Code story specifically is where the multiple has the most room to compound. Coding agents are sticky in a way consumer chat is not — every customer that adopts Claude Code at scale signs up for hosted developer infrastructure that bills monthly, scales with seat count, and integrates into a long tail of CI/CD systems that are expensive to migrate off. The fork I covered in my May 9 piece on Claude Outcomes and OpenAI Codex Goals is showing up in the revenue mix exactly the way the labs telegraphed: Anthropic is building outcome-priced enterprise contracts that lock customers into multi-year commitments, while OpenAI is selling slot-based goal contracts that compete on price. Outcome pricing converts faster to enterprise procurement. That conversion is what the $900 billion is buying.

What could break a price this high

The first hole in the thesis is the cost structure underneath the revenue. Anthropic is running at $40 billion of ARR, but the gross margin on Claude inference and Claude Code is not yet public, and the dominant production model — Claude Opus 4.7 — is expensive to serve. The Cryptopolitan write-up on the round reported that several syndicate participants pushed Anthropic for more clarity on cost-of-revenue trajectory before signing, and the company has not yet disclosed a public gross margin. If the run rate is sustained on contribution margins that compress as Chinese open-weights coding models — the GLM-5.1, MiniMax M2.7, Kimi K2.6, and DeepSeek V4 wave shipped in the past month — push down the price-per-token customers are willing to pay, then the revenue multiple looks defensible only on the top line. The IPO will force disclosure. Operators should expect the gross-margin number to be the first line item public-market investors interrogate.

The second hole is the customer-concentration profile. Anthropic’s enterprise revenue is heavy in two segments — agentic coding (Claude Code) and regulated enterprise (financial services, life sciences, defense via the Gates Foundation partnership and similar deals) — and a sizable fraction of the run-rate ARR sits with a small number of large accounts. Anthropic’s own disclosure of 300,000 business customers tells you the long tail exists, but the 7x year-over-year growth in large accounts implies that the very largest contracts are growing materially faster than the small-customer base. If two or three of those large contracts churn — to a Chinese open-weights deployment, to a GPT-derived in-house stack, to a Gemini-based alternative — the implied revenue compression on a forward basis would meaningfully impair the multiple. The IPO prospectus will surface customer concentration. The current $900 billion mark prices it as a non-issue. That’s a bet.

The third hole is the compute pipeline. Anthropic needs roughly twice as much serving capacity over the next twelve months as it has currently contracted, which is why the company has been signing increasingly creative compute deals — the SpaceX Colossus orbital-compute arrangement I covered on May 8, the Google $40 billion compute commitment from April, and the multi-cloud build-out across AWS, Azure, and GCP. If any of those compute supply commitments slip materially in 2026 — because of power-grid constraints, GPU allocation politics, or hyperscaler capex re-prioritization — the demand Anthropic is booking gets bottlenecked before it converts to revenue. The $9 billion of compute capex the company has earmarked for the next four quarters, per the GovCon Wire coverage, is contingent on power and silicon availability that no contract can fully guarantee. Compute delivery risk is the most underappreciated downside in the bull case.

The fourth hole is the OpenAI counter-move. OpenAI is not standing still: the company has closed seven acquisitions year-to-date, including the Hiro acqui-hire announced in mid-April, and it remains the consumer-AI distribution leader by an order of magnitude. If OpenAI responds to the Anthropic mark by accelerating its own enterprise push — pricing Codex Goals more aggressively, signing larger anchor customers at lower per-seat rates, or building a comparable agentic platform — the Claude Code revenue moat could compress before Anthropic’s IPO clears. The duopoly is real, and the duopoly is competitive. A 35x ARR multiple at OpenAI and a 22.5x multiple at Anthropic is the kind of pricing gap that activates strategy. Expect OpenAI to act on it.

The fifth hole is the regulatory perimeter, which is widening faster than the syndicate is pricing in. The CAISI pre-deployment testing regime I covered on May 7 is one model-level intervention; the AI-built zero-day reporting I covered on May 13 sits inside a broader push toward mandatory testing, mandatory disclosure, and mandatory liability attribution for frontier labs. Anthropic has historically been the most aligned-by-policy of the major labs — the company has cooperated extensively with CAISI, the Gates Foundation, and the EU AI Office. That alignment is the basis of its premium with regulators today. But regulatory regimes shift in 24-month cycles, and the post-IPO Anthropic will be operating in a 2027 environment that may impose compute caps, training-data audits, and capability-class restrictions the current syndicate has not modeled. A 50-basis-point hit to revenue growth from regulatory friction compounds quickly at this multiple.

What operators should actually do with this print

The most likely scenario over the next two quarters is that the round closes near the announced terms, Anthropic files an S-1 in August or September, and the public-market test arrives in October at a market cap in the $1 to $1.2 trillion range. The harder scenario to handicap is the second-derivative effect — what happens to every other AI lab’s funding round once Anthropic clears the $900 billion pre-money. The expectation curve for what a “winning” foundation-model lab is worth has been re-anchored. Every Series B and Series C pitched after this round will reference the Anthropic comp implicitly. Mistral, Cohere, AI21, xAI, and the entire second-tier of labs now have to explain why their valuation isn’t a fraction of Anthropic’s — a conversation made harder by the Cohere-Aleph Alpha merger I covered on May 6, which already telegraphed consolidation pressure in the runner-up tier.

The implied market structure after this round is closer to a duopoly than the prior consensus. OpenAI and Anthropic together will represent roughly $1.8 trillion of foundation-model market cap on a post-money basis, against a second tier that probably sums to under $200 billion across all remaining labs. That is a 9:1 concentration ratio at the top of the AI stack, and the operating dynamics inside a 9:1 concentration market are different from a fragmented one: pricing power shifts to the top two, talent flows toward the top two, and customer procurement increasingly defaults to a single-vendor bet on either Claude or GPT. The third-tier labs become sovereign-AI plays, vertical specialists, or acquisition targets. The market is now structurally bifurcated.

Operators reading this should treat the Anthropic mark as a forcing function on three categories of decision: vendor strategy, talent strategy, and competitive positioning. The cleanest takeaways:

• Reprice your foundation-model vendor bet. If your enterprise AI roadmap assumed parity between Claude and ChatGPT on twelve-month enterprise pricing, this round tells you Anthropic is operating from a position of negotiating strength that will translate into less price flexibility in 2026 contracts. Lock in multi-year commitments on Claude before the IPO mark hits, or migrate to a multi-model architecture that lets you arbitrate vendor pricing as the duopoly hardens.
• Audit your AI-engineering hiring funnel against the new comp. Every senior researcher and platform engineer Anthropic recruits in the next two quarters is now negotiating on a $900 billion pre-money paper-equity base. If your offer letters reference $5-10 million in equity grants, the outsized Anthropic comp packages flagged in the StartupHub coverage will distort your talent pipeline within one cycle. Either match in scale or specialize aggressively on a domain Anthropic does not prioritize.
• Watch the Claude Code gross-margin disclosure when the S-1 lands. The Claude Code segment is the single most important line item to scrutinize in the IPO filing. A gross margin above 65% confirms the duopoly thesis. A gross margin in the 40-50% range suggests the open-weights wave is already eroding inference economics. The number will be the most-debated metric of the year.
• Map your AI-vendor counterparty risk to the IPO calendar. Pre-IPO Anthropic is a private company with maximum flexibility to renegotiate contracts. Post-IPO Anthropic will be a public company with quarterly earnings pressure, an SEC-disclosed customer concentration profile, and lockup expirations that drive employee turnover. Plan for two distinct procurement environments six months apart.
• Treat the $900 billion mark as a ceiling test, not a floor. This valuation will be either confirmed or repriced by the IPO. If it holds, the duopoly is real. If it cracks below $800 billion post-listing, the entire AI capital-stack thesis — including every second-derivative trade in chips, networking, and data-center buildout — gets re-examined. The next two quarters are the most consequential in the AI venture market since the original GPT-4 launch.

In other news

Anthropic forms $200 million partnership with the Gates Foundation — On May 14, Anthropic and the Gates Foundation announced a four-year commitment of grant funding and Claude usage credits targeting global health, education, and economic mobility in low- and middle-income countries, including specific workstreams on polio, HPV, and preeclampsia. The partnership signals Anthropic’s continued positioning as the policy-aligned lab in the duopoly.

Sierra raises $950M at $15B valuation — Bret Taylor’s enterprise AI agent startup Sierra closed a $950 million round led by Tiger Global and GV, post-money at over $15 billion, TechCrunch reported. Sierra’s software is now used by nearly half the Fortune 50 and generates $150 million in ARR — a sign that the agentic enterprise layer is consolidating around a few players with deep distribution.

Meta announces $125-145B in fresh AI capital spending — Meta’s latest capex guide pushes its 2026 AI infrastructure budget into the $125 to $145 billion range, funded almost entirely from ad revenue, alongside a new Muse Spark multimodal model rolling out across WhatsApp, Instagram, and Ray-Ban glasses, The AI Marketers summarized. The capex print continues to validate the hyperscaler-spending thesis that drove Cisco’s record print last week.

Google’s “Gemini Intelligence” reorganizes Android around AI agents — At Google’s Android Show, executive Sameer Samat declared the company is “transitioning from an operating system to an intelligence system,” with Gemini moving from chatbot to cross-app agent that can pull data from Gmail, build shopping carts, and book reservations across services, CNBC reported. The rollout starts on Samsung Galaxy and Pixel phones this summer.

Four Chinese open-weights coding models ship in 12 days — Z.ai’s GLM-5.1, MiniMax’s M2.7, Moonshot’s Kimi K2.6, and DeepSeek’s V4 (with Pro and Flash variants) all released between April 7 and April 24, with none costing more than a third of Claude Opus 4.7 on a per-token basis, per the DEV Community comparison of the late-April Chinese LLM stack. The cost-compression pressure on Western frontier-lab inference is the largest single risk to the Anthropic gross-margin thesis above.

Exaforce closes $125M Series B for agentic security — Cybersecurity AI startup Exaforce raised a $125 million Series B from HarbourVest, Peak XV, Mayfield, Khosla Ventures, and others to scale its agentic security-operations platform, YourStory reported. The round continues the broader pattern of AI-native security platforms attracting late-stage capital as enterprises automate SOC workloads.