§ News
By AI Blog Editor
May 29, 2026 · 14 min read
The leapfrog — Anthropic Series H closes at $65B and passes OpenAI on the way to $965B
On May 28 Anthropic closed a $65B Series H at a $965B post-money valuation, lifting it past OpenAI to become the world's most valuable AI startup. Three months ago the same company was worth $380B with $14B of run-rate. Both numbers have since roughly tripled.
On May 28, 2026, Anthropic closed a $65 billion Series H at a $965 billion post-money valuation, co-led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, with Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN writing in alongside them. Strategic infrastructure partners — Samsung, SK hynix, and Micron — joined as well, and $15 billion of the headline number is previously committed hyperscaler money, including $5 billion from Amazon announced in April. The round is documented in Anthropic's own press release, TechCrunch's writeup, Fortune's explainer, The Decoder's report, and CFO.com's piece pairing it against OpenAI's recent commentary.
$965 billion is what the round priced. The number it was clearly built to imply is a different one. Anthropic stopped $35 billion short of a trillion-dollar private valuation, which is the sort of round geometry that gets designed in a room. A trillion would have been the headline. $965 billion is the headline plus a small piece of unfinished business that the IPO can collect later.
The leapfrog
The cleanest news fact in the announcement is the one Anthropic's release does not mention by name. At $965 billion, Anthropic has passed OpenAI to become the most valuable AI startup in the world. Fortune put it directly, calling it "eclipsing its rival OpenAI." Three months ago that ordering ran the other way. At the Series G in February, Anthropic closed $30 billion at a $380 billion post-money — well behind the OpenAI valuation circling around the half-trillion mark in the same window. Between then and now, OpenAI moved by some amount, and Anthropic moved by 2.5x. The order changed.
The trajectory is the part worth pausing on. Three months ago: $380 billion, $14 billion run-rate revenue. Now: $965 billion, $47 billion run-rate. Valuation roughly 2.5x. Revenue roughly 3.4x. The two numbers moved in the same direction, in close to the same multiple, in the same quarter. That is unusual. Most fundraises of this size at this growth rate involve a valuation that races ahead of the underlying business. Anthropic's growth, on the metric the company chose to publish, is keeping pace with what investors are pricing.
The CFO, Krishna Rao, said the relevant thing in the press release: "This funding will help us serve the historic demand we are experiencing, stay at the research frontier, and bring Claude to more of the places where work happens." The sentence is bland because the numbers it sits next to are not. From $7 billion of run-rate at the start of 2025 to $47 billion in mid-May 2026 is six-and-a-half-fold growth across roughly seventeen months. The investor pitch did not have to argue that frontier AI is a real business. It only had to argue that this particular vendor is winning the part of the market that pays.
What the $65 billion is actually paying for
If you strip the press release back to verbs, the money has one job: buy compute. The Decoder's reporting lists more than ten gigawatts of capacity tied up alongside the round — a multi-gigawatt Google/Broadcom TPU arrangement, the Colossus 1 deal that the SpaceX S-1 surfaced for 220,000 GPUs, and the $5 billion Amazon contribution that anchors the hyperscaler portion of the round.
Read it back the long way. Anthropic raises $65 billion. Anthropic pays Google, Amazon, and SpaceX something close to that, over a multi-year window, for GPU and TPU hours. Google, Amazon, and SpaceX use the receipts to keep ordering from Nvidia, Samsung, SK hynix, and Micron. Two of those four memory and HBM vendors are also direct investors in the round. The capital does a loop. It enters Anthropic as equity, leaves as compute spend, returns to the vendors who supplied the hardware, and a non-trivial slice of those vendors are sitting on the cap table.
This is the through-line on every frontier-AI round at this scale right now. The Series H is the financing layer that converts limited-partner appetite into Nvidia revenue. Anthropic gets compute; the LPs get exposure to compute demand without having to pick a vendor; and the GPU food chain gets a customer that has been paid in advance to keep ordering. Calling it a circle is a description, not a complaint. A circle that produces $47 billion of customer revenue at the end of seventeen months is what the entire industry is hoping the picture looks like at scale.

The Series H club, and what it usually means
Series H rounds are rare and the company that hits one is usually telling you something. Fortune's piece flagged the lineage: Facebook, Lyft, Discord, Slack. Databricks got to H and kept going through I, J, and K. The historical reading of a Series H is the company that meant to be public by now and isn't, raising at progressively flatter terms while the IPO window opens and closes.
That is not the read this time. Fortune's reporting and TechCrunch's both note that Anthropic and OpenAI are now both reportedly preparing IPO paperwork, with OpenAI expected to file confidentially within weeks. The Series H here is the opposite of the Databricks pattern. It is not a stall round; it is a bridge round designed to get one last enormous private check on the books before the public-market price-setting machine takes over. The participating investors — Baillie Gifford, Blackstone, Brookfield, Fidelity, DST Global — are crossover funds that buy in private and then continue holding through the IPO. They are not in the round to flip on the next preferred share. They are in the round to be there at the open.
The $965 billion mark, for the same reason, reads less like a private mark-to-market and more like a public-market reference price. The roadshow has not started; the comp has been set.
Where this leaves OpenAI
OpenAI's CFO, Sarah Friar, gave a parallel set of public comments recently that read very differently now that Anthropic has gone first. Friar described 2026 as a market where there is "not a lot of compute" and her sales team is "running ragged" against enterprise demand. The framing is correct on its own terms, but it lands awkwardly next to a competitor that just secured ten gigawatts of forward capacity and bought it with equity that priced at a near-trillion mark.
Brad Gerstner of Altimeter, who led on the Series H, was on record earlier in the year questioning OpenAI's $1.4 trillion in compute commitments as a leverage problem. He is now writing into the company that has spent the last quarter winning the other half of the same compute table. That is not subtle. The Altimeter position is a thesis trade about which of the two frontier labs is going to be paying for fewer obligations and producing more revenue from them.
What this means
Three takeaways.
- The most valuable AI startup is no longer OpenAI. This is the first time the private-market ordering between the two frontier labs has flipped. The change is recent — three months ago Anthropic was a clear second — and it was driven by the rate at which Claude's enterprise revenue compounded, not by the OpenAI number going down. The market has decided, at least for this quarter, that the company with PwC training 30,000 professionals and Claude on AWS, Google Cloud, and Azure is the safer bet on AI-as-a-product. Whether that survives the next major OpenAI model release is the open question.
- The Series H is the IPO comp, not a stall round. Crossover funds — Baillie Gifford, Blackstone, Brookfield, Fidelity, DST — are the LPs you put into the last round before the S-1. The structural read of this round is that Anthropic and OpenAI will both file within months, that the public market reference for Anthropic is now $965 billion, and that the company is using the H to put long-duration holders on the cap table before any retail price discovery happens. The number $965 billion was not chosen lightly. It is, to within a rounding error, the most you can be worth privately while letting the IPO put a "trillion-dollar" headline on the open.
- The capital is the compute, and the loop is now visible. Ten gigawatts of forward GPU and TPU commitments, two memory vendors on the cap table, $15 billion of hyperscaler money baked into the round itself. The check is a financing instrument for a compute pipeline. The next question, for both Anthropic and OpenAI, is whether the demand at the end of the pipeline grows fast enough to clear the obligations at the front. Anthropic's $47 billion run-rate, up from $7 billion seventeen months ago, is the company's current answer. The IPO will ask the same question with a different set of buyers in the room.
The Series H closes a private fundraising cycle that started at the $200 million seed in 2021 and ran through eight letters of the alphabet in five years. The next round Anthropic raises will not be lettered. It will have a ticker.
* * *
Thanks for reading. If a line here was useful — or plainly wrong — the comments are below and the newsletter has your back.
Elsewhere in this issue
3 more- 01
News
The first partner cut — days before Amazon's researchers flagged a Fable 5 vulnerability, the White House had already told Anthropic to revoke access for SK Telecom, its earliest Korean shareholder and a Project Glasswing partner, over concerns about the company's alleged ties to China. Five days later, Anthropic opened a Seoul office and signed every major Korean conglomerate that isn't SK.
Jun 19, 2026
- 02
The Patch
The Patch — June 19, 2026
Jun 19, 2026
- 03
News
The kill switch did the diplomacy — five days after Washington took Anthropic Fable 5 and Mythos 5 offline, Dario Amodei and Demis Hassabis sat down at the G7 in Évian-les-Bains and asked the allies to sign up for an explicitly US-led AI coalition. Canada said yes; France brought a list.
Jun 18, 2026
Letters
Arguments, corrections, questions. Anonymous comments allowed; be kind, be specific.