Anthropic closed a $65B Series H on May 28 at a $965B post-money valuation, then filed a confidential draft S-1 with the SEC four days later on June 1. The sequence is not subtle: a closed-room valuation set by crossover funds and hyperscaler partners is about to become the anchor price public-market investors scrutinize under mandatory disclosure.
What the Series H actually put on paper
The round was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with co-leads Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN, per Anthropic’s announcement. At $65B on a $965B post-money, the implied dilution for new capital was roughly 6.7%, according to Tech Insider’s round-structure analysis. The prior Series G in February valued the company at $380B, per Wikipedia. That is a 2.5x jump in roughly three months.
Run-rate revenue crossed $47B as of late May, up from $14B at the Series G in February and $9B at the end of 2025: a 3.36x increase in 105 days. On that run-rate, the Series H priced at roughly 20.5x revenue. For comparison, OpenAI raised $122B at an $852B post-money valuation in March 2026, per TechCrunch. Anthropic now commands the higher absolute valuation; how the revenue multiples compare depends on figures OpenAI has not publicly disclosed.
Claude Code is a key revenue driver. Anthropic’s CFO called out Claude Code and Cowork as central to serving the “historic demand” the company is experiencing. A developer tool scaling to prominence at a near-trillion-dollar company says something about where Anthropic’s growth is concentrated: enterprise and developer adoption, not consumer chat.
The S-1 filing: what changes when the curtain lifts
On June 1, Anthropic confidentially submitted a draft registration statement on Form S-1 to the SEC. The number of shares and price range have not been set; the filing explicitly conditions the offering on market conditions. SiliconReport notes that the filing leaves both IPO price and share count open.
The confidential submission is standard for large tech IPOs. It lets the company work through SEC comments privately before the public filing. What matters here is timing: the S-1 landed less than a week after the Series H close, in a season where SpaceX is targeting a $2T valuation and OpenAI is preparing its own IPO after raising $122B at $852B in March.
Revenue vs. compute burn: the obligations behind the multiple
Anthropic signed compute agreements for up to 5GW with Amazon, 5GW of TPU capacity with Google and Broadcom, and GPU capacity via SpaceX’s Colossus 1 and 2, making it the first frontier model deployed simultaneously on AWS, Google Cloud, and Azure. The multi-cloud posture is an operational advantage. It is also a balance-sheet commitment.
Google has separately committed substantial additional capital to Anthropic, per Tech Insider’s analysis. The total committed compute spend across Anthropic’s contracts is large enough to be a material disclosure item in the S-1.
The structural question for public-market investors: can $47B in run-rate revenue, which tripled in the 105 days between Series G and H, compound faster than the infrastructure burn required to train and serve frontier models? The compute commitments are multi-year obligations. They do not flex down if revenue growth slows.
How the three frontier labs diverged
The three companies competing for frontier-model leadership have taken structurally different paths to similar valuations:
| Anthropic | OpenAI | Google DeepMind | |
|---|---|---|---|
| Latest known valuation | $965B | $852B (March 2026) | N/A (Alphabet subsidiary) |
| Revenue run-rate | ~$47B | Not publicly disclosed | Not separately disclosed |
| Primary revenue driver | Claude Code and enterprise tools | ChatGPT and consumer products | Gemini (integrated into Google products) |
| Compute posture | Multi-cloud (AWS, GCP, Azure) | Primarily Microsoft Azure | Internal TPU fleet |
Anthropic’s valuation exceeds OpenAI’s even though both companies carry heavy compute commitments and neither has disclosed margin profiles. The B2B developer-tool concentration (Claude Code as a central revenue driver) is both a strength, given high retention and usage-based pricing, and a concentration risk that the S-1 risk factors will have to address.
What the private mark means for every AI investor’s portfolio
Every late-stage fund that bought into Anthropic’s Series E through H now holds paper priced between $380B and $965B. Until the S-1 becomes public, those marks exist in a closed system: the valuation is what the last round’s lead investors agreed to pay. Once the IPO prices, the private mark either validates or compresses.
If the IPO prices below $965B, every fund carrying Anthropic at the Series H price takes a markdown. If it prices at or above, the mark holds. The 6.7% dilution in the H round means Series H investors bought a narrow slice at a premium to the prior round; a flat or down IPO hits them first.
Risks the S-1 will have to disclose
Anthropic is a public benefit corporation, founded in 2021 by seven former OpenAI employees. The PBC structure imposes a statutory obligation to balance shareholder returns with public benefit, which means Anthropic’s safety commitments can override commercial opportunities. Public investors will have to weigh whether that governance stance is a moat (attracting enterprise customers who want enforceable safety commitments) or a liability (limiting the addressable market for government and defense contracts).
Anthropic also previewed its Mythos model in April 2026 but kept access restricted, warning developers it had discovered thousands of high-severity bugs. Bloomberg reported the EU cybersecurity agency may receive access. A frontier model with known high-severity vulnerabilities is a disclosure item. How Anthropic frames Mythos in the risk section will signal whether the company views the bugs as a normal pre-release finding or a structural problem with the model line.
The S-1 filing does not set a price or date. The private market has already set a number: $965B. Public investors now get to decide whether they agree.
Frequently Asked Questions
How does Anthropic’s revenue multiple actually compare to OpenAI’s?
OpenAI’s roughly $20B ARR against an $852B valuation implies a 42.6x revenue multiple. Anthropic carries a higher absolute valuation at $965B but trades at less than half that multiple (20.5x), reflecting its faster recent growth (3.36x in 105 days) and the heavier multi-year compute obligations baked into its contract structure.
Does Anthropic’s public benefit corporation status restrict its government work?
In 2026, Anthropic refused DoD demands to drop AI safeguards on domestic surveillance and autonomous weapons. The DoD responded by designating it a “supply chain risk.” Google, Amazon, Apple, and Microsoft publicly supported Anthropic’s lawsuit against the DoD over that designation. Expect this dispute to surface as a named risk factor in the public S-1.
How much of Anthropic’s revenue is concentrated in Claude Code?
Claude Code reached $2.5B annualized revenue by February 2026 and was estimated at roughly 33% of Anthropic’s total by May. Against the $47B run-rate, that implies over $15B annualized from a single developer tool. That concentration is a strength while enterprise coding budgets expand, but it means any slowdown in developer-tool spending compresses the top line disproportionately.
What compute spending obligations sit behind the $965B valuation?
Total committed compute spend across Anthropic’s contracts is estimated at $112B through 2029. Google separately committed up to $40B, structured as $10B upfront with the remainder milestone-gated. The $112B figure is 1.7x the entire Series H raise, so the $65B in fresh capital is already more than absorbed by infrastructure commitments that do not flex down if revenue growth slows.