Anthropic raises $65 billion (approx. ¥10.3 trillion) in Series H. Company valuation reaches $965 billion, overtaking OpenAI.

On May 28, 2026 (local time), the U.S.-based Anthropic announced that it had raised $65 billion (approximately ¥10.3 trillion) in a Series H round, bringing its post-money valuation to $965 billion (approximately ¥152 trillion). Having driven its valuation up 2.5-fold in roughly three months, the company surpassed OpenAI for the first time in terms of both valuation and annualized revenue, becoming the world's most valuable AI startup. While Silicon Valley VCs are buzzing that it "overtook OpenAI in 90 days," the lineup of backers—featuring HBM manufacturers and government-affiliated funds—has also drawn criticism that this is "industrial policy rather than a VC deal" and "circular trading." This article, written from a VC perspective, organizes the differences in how various outlets reported the news and the metrics worth watching next. (Unless otherwise noted, yen conversions assume roughly 1 do

In 3 Months, Valuation 2.5x──Surpassing OpenAI with Largest-Ever Capital Raise

On May 28, 2026, Anthropic officially announced that it had raised $65 billion (approximately ¥10.3 trillion) in its Series H round, bringing its post-money valuation to $965 billion (approximately ¥152 trillion). The round was led by four firms—Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital—and according to Bloomberg, each of the leading firms invested more than $2 billion. Co-leads included Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN, while participating investors included Baillie Gifford, Blackstone, Brookfield, D.E. Shaw, DST Global, Fidelity, General Catalyst, Lightspeed, MGX, Temasek, T. Rowe Price, Jane Street, and NTTVC.

The immediately preceding fundraising was the Series G on February 12, 2026 ($30 billion = approximately ¥4.7 trillion, at a valuation of $380 billion = approximately ¥60 trillion), meaning the valuation ballooned roughly 2.5-fold in just about three months. Cumulative funding since the company's founding in 2021 has reached approximately $144 billion (approximately ¥23 trillion, per Crunchbase tallies). As a result, Anthropic surpassed OpenAI—which had been valued at around $852 billion (approximately ¥135 trillion) after raising a record-breaking $122 billion (approximately ¥19 trillion) in late March 2026—in both valuation and annualized revenue for the first time. The company, founded in 2021 by the sibling duo Dario and Daniela Amodei, both former OpenAI employees, has thus overtaken their former employer in market capitalization.

Business figures backing the fundraising were also disclosed at the same time. Annualized revenue (run rate) surpassed $47 billion (approximately ¥7.4 trillion) in early May. CFO Krishna Rao commented, "This funding will help us meet the historic demand we are experiencing, stay at the frontier of research, and bring Claude to more places in the workplace." The stated uses of the funds are "advancing safety and interpretability research, expanding compute to meet growing demand for Claude, and expanding products and partnerships." Of the $65 billion, $1.5 billion (approximately ¥2.4 trillion) represents a confirmed portion of existing hyperscaler investments, including $5 billion (approximately ¥790 billion) from Amazon. On the same day as the announcement, the new model "Claude Opus 4.8" was also released, creating a picture in which the fundraising follows on the heels of the product's momentum.

Silicon Valley VCs React──"Surpassed OpenAI in 90 Days"

The lead VCs' bullishness is unmistakable. Altimeter's Brad Gerstner is reported to have stated publicly that, although Anthropic was viewed as having "dropped out of the game" last year, it has overwhelmed OpenAI over the past 90 days, and he has expressed the view that its annualized revenue will triple over the course of 2026 and could reach the $80 billion–$100 billion range by year-end. Dragoneer's Marc Stad described the technological progress as "breathtaking," emphasizing that commercialization is still at a "very early" stage. Greenoaks' Neil Mehta and Sequoia's Alfred Lin are reported to have cited the company's corporate culture and business momentum as reasons for their support. The very fact that each lead investor committed more than $2 billion is being taken as a sign of their conviction.

The core of the VCs' investment thesis lies not in the loftiness of the valuation but in the durability of demand. Its lead in the coding/agent domain—epitomized by Claude Code—together with company-wide adoption at large enterprises such as KPMG (with a workforce of about 276,000) and PwC, underpins its $47 billion in annualized revenue. According to reporting by The Wall Street Journal, the company is projected to reach its first operating profit on a 130% revenue increase, and the prevailing view is that its "AI Safety" brand—once seen as a hindrance—has instead turned into a procurement asset in regulated industries and the enterprise market.

The "Industrial Policy" Reflected in the Cap Table and the Circular-Trading Controversy

However, inside Silicon Valley, a different reading is gaining ground. The tech outlet Silicon Canals argued that this cap table (the composition of investors) is "less like a conventional venture deal and more like industrial policy." Its rationale lies in the fact that Micron, Samsung, and SK hynix—the three major suppliers of HBM (high-bandwidth memory)—all appeared simultaneously on the same investor roster. With supply of HBM3E/HBM4 running tight and turning into a scramble, memory makers are acquiring stakes tied to physical supply allocations on top of their ordinary supply contracts. This gives rise to a vertically integrated structure in which "capital flows in, allocations of memory and compute flow out, and the upside is shared rather than renegotiated every year." Layered on top of this are quasi-governmental capital pools such as GIC, Temasek, and MGX, along with infrastructure investors like Blackstone and Brookfield.

This structure is also the flip side of the "circular financing" critique. Anthropic buys components from the memory makers who are its investors, has supply contracts with Amazon for up to 5 gigawatts, with Google and Broadcom for 5 gigawatts of next-generation TPUs, and with SpaceX for GPU capacity on Colossus 1/2. The pattern in which supplier, customer, and investor merge into one is pointed out as being structurally identical to the circular arrangements among OpenAI and the cloud players centered on Nvidia. Some bears and short sellers also direct skepticism at the quality of the $47 billion run-rate (the possibility of double-counting or extrapolation from outliers), but at this point this is a company-disclosed figure, and public shareholders will scrutinize it after the IPO. The implication for VCs is clear: at this scale tier it is no longer pure "venture investing"—strategic capital and state-affiliated capital take the leading roles, and a "calcification of oligopoly" advances, making it harder for cash-poor emerging labs to remain at the frontier.

How Each Paper and Site Reported It──"Surpassing OpenAI" or "On the Verge of $1 Trillion"?

The emphasis of the coverage split along the lines of each outlet's character. Major U.S. business newspapers adopted a competitive, horse-race framing: Bloomberg ran headlines like "eclipses OpenAI" and "approaches $1 trillion," while in a separate piece it also posed the sober question of "whether the two companies' IPOs can live up to expectations." CNBC framed it as "overtaking OpenAI as the most valuable AI startup, with $1 trillion in sight," and TechCrunch positioned it as "likely the last private fundraising round ahead of an IPO." PitchBook described it as "beating OpenAI in the valuation race," and Crunchbase put it as "leaving OpenAI behind in the unicorn rankings." The Washington Post (via AP) put the demand side front and center, with "surging to $965 billion against a backdrop of soaring demand for Claude."

By contrast, Silicon Canals' structural angle—"more industrial policy than a venture deal"—drew attention as one of the few distinctive perspectives. Among Japanese outlets, ITmedia conveyed the facts succinctly with "a valuation of $965 billion, 2.5 times higher in three months, surpassing OpenAI," Forbes JAPAN billed it as "one of the largest fundraising rounds in human history" with a "valuation of ¥141 trillion," and Impress Watch reported the same $965 billion as a "valuation of ¥154 trillion." The fact that an identical figure swings from ¥141 trillion to ¥154 trillion stems from differences in each outlet's assumed exchange rate (roughly ¥146 to ¥160), and it is worth keeping in mind that yen conversions carry a considerable range. TradingKey, an investment-decision-oriented outlet, ran a piece questioning the probability that OpenAI—surpassed on both valuation and revenue—will actually realize its IPO.

When Will the Next Starting Gun Sound? ── IPOs and Indicators to Watch

This time, it was reported almost uniformly that this "could be the last private placement before the IPO." Some reports conveyed observations of a listing as early as the autumn of 2026 (October), with Goldman Sachs, JPMorgan, and Morgan Stanley floated as candidate lead underwriters and Wilson Sonsini mentioned for legal counsel. However, as of the end of April 2026, no S-1 (registration statement/prospectus) had been filed, nor had underwriters been formally appointed or a roadshow scheduled. Because it typically takes six to nine months from the engagement of banks to a listing, the view that an autumn listing is ambitious is strong, and the independent forecaster FutureSearch projects around March 2027, with a median valuation of roughly $560 billion (about ¥88 trillion). The fact that this projected figure falls below the current private placement valuation of $965 billion suggests the "down-round IPO" risk that later-stage investors and VCs are wary of, and it carries the tension that $965 billion is, after all, merely a private placement valuation that the public market could re-rate either upward or downward.

On OpenAI's side as well, caution within the company over the timing of its own IPO has been reported, and Anthropic's latest move could accelerate a countervailing capital raise or listing strategy. SpaceX, OpenAI, and Anthropic are all running side by side as candidates for some of the largest IPOs in history. From a VC's perspective, the next "measurable" milestones will be, first, the S-1 filing (this will be the real starting gun; the timing is likely roughly the second half of 2026 to the first half of 2027), and second, the revenue disclosures for the third and fourth quarters of 2026 (whether the $47 billion run rate converts into actual figures and whether the outlook for 130% revenue growth and a first-ever operating profit holds). In addition, the fate of HBM4 supply allocations, the progress in fulfilling 5-gigawatt-class compute contracts, OpenAI's reaction, and whether the indicative quotes in the private secondary market can hold near $965 billion heading into 2027 — these will be the empirical data for judging the sustainability of the enthusiasm.