The chipmakers who supply the memory inside artificial-intelligence data centers rarely write checks to software startups. This week, three of the world’s largest memory makers made an exception.
Micron Technology, Samsung Electronics and SK Hynix all participated in Anthropic’s latest funding round, according to the company and people familiar with the matter. Anthropic said it closed the round, a $65 billion Series H, at a post-money valuation of $965 billion, the first time the maker of the Claude chatbot has been worth more than rival OpenAI.
The round was led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital, with D.E. Shaw, Blackstone, Jane Street, Temasek and a fund called Situation Awareness among the followers, the company said. It included $15 billion that hyperscale cloud providers had previously committed, of which Amazon contributed $5 billion. OpenAI was last valued at roughly $852 billion, according to people familiar with its most recent fundraising.
The presence of the memory makers is the most unusual detail in a round full of unusual numbers. Micron, Samsung and SK Hynix dominate the market for the high-bandwidth memory that Nvidia’s most powerful AI accelerators require, and all three have been running at full capacity as data-center demand outstrips supply.
Their participation signals that the supply chain is placing a long-term bet on Claude workloads, analysts said. Component suppliers rarely take equity stakes in software companies; when they do, it is usually to lock in demand or hedge against a shift in the market. Here, the round effectively gives the memory makers a seat at the table as Anthropic plans its next generation of models and the data centers that will train them.
Anthropic’s ascent has been rapid even by the standards of the AI boom. In February, the company raised $30 billion at a $380 billion post-money valuation, in a round led by GIC and Coatue. Three months later, its valuation has more than doubled. Google, a major backer, committed up to $40 billion in a package reported in April, structured as $10 billion upfront with the remainder contingent on performance targets, plus as much as 5 gigawatts of compute over five years, according to people familiar with the terms.
OpenAI, for its part, raised $122 billion in a round that closed earlier this year, according to people familiar with the matter, a sum that dwarfed every venture round that preceded it. The two companies have spent the past two years trading the title of the world’s most valuable AI company, each round resetting the bar for the other.
Anthropic’s revenue has kept pace with its valuation. The company disclosed annualized run-rate revenue of about $14 billion in February; analysts now estimate the figure has climbed to roughly $47 billion, driven in part by Claude Code, its programming assistant, which passed $2.5 billion in annualized revenue earlier this year.
The money will go toward the capital-intensive work of frontier AI. Anthropic said it plans to expand compute capacity across AWS, Google Cloud and CoreWeave and to continue training its next-generation models, including the successor to Claude Opus 4.8, which the company released the same week. It is also building enterprise features for regulated industries, including health care and government customers.
Investors are betting the round sets up an initial public offering. Forge Global, a secondary-market platform, has flagged an October window as the earliest plausible IPO date, though Anthropic has not confirmed any plan to go public. An IPO would test whether public investors share the private market’s enthusiasm for valuations near the trillion-dollar mark.
The round’s size reflects how much capital frontier AI now consumes. Anthropic trains its models on clusters of tens of thousands of Nvidia accelerators, and each successive model generation demands more compute than the last. The company has said it expects the cost of training and running frontier models to keep climbing, which is why it has assembled a syndicate that spans venture funds, sovereign investors and now the semiconductor industry.
The timing is deliberate, according to people familiar with the strategy. With OpenAI and Anthropic both eyeing the public markets, the round resets the valuation conversation before any IPO filing. A higher mark gives existing investors a paper gain, raises the bar for OpenAI’s next round, and, not incidentally, keeps the startup’s employees holding options that look increasingly valuable.
For Amazon, the investment deepens a relationship that has already been lucrative on paper. The cloud giant disclosed earlier this year that it holds billions of dollars in Anthropic-related notes, a stake that has soared in value alongside the startup’s fortunes. Amazon also supplies much of the computing capacity on which Anthropic trains its models.
The memory makers are making a different kind of bet. Their chips are sold out for the foreseeable future, and the AI buildout shows no sign of slowing. By joining Anthropic’s cap table, they are buying insurance that the party lasts, and a front-row seat if it does.
Some investors caution that the round’s size carries its own risks. Funding at this scale requires ever-larger future rounds to mark up, and the compute bill for frontier models grows with each generation. A slowdown in enterprise AI spending would hit Anthropic’s valuation as hard as any downturn in the memory market.
For now, the mood is celebratory. Anthropic’s founders have said the company’s goal is to build AI systems that are safe enough to trust with consequential tasks, and investors are paying for that ambition at a price no AI company has commanded before.
