Anthropic closed a $65 billion Series H round on 28 May 2026 at a $965 billion post-money valuation, the company confirmed in its own announcement, vaulting past OpenAI to become the most valuable private AI company in the world and bringing it within striking distance of a $1 trillion private mark. The round was led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital, and is widely expected to be Anthropic’s last fundraising before an IPO. What follows is less a recap of the figures than a reading of the cap table, which looks closer to an industrial policy document than a conventional venture deal.
The cap table reads like an industrial policy document
Beyond the lead investors, the round was co-led by a roster of crossover and sovereign-adjacent institutions, including Capital Group, Coatue, D1 Capital Partners, GIC, ICONIQ, and XN, with participation from the likes of Blackstone, Brookfield, Fidelity, and Temasek. More telling is the strategic layer. Micron, Samsung, and SK hynix — the three companies that together supply essentially all of the high-bandwidth memory used in AI accelerators — joined as strategic infrastructure partners rather than confining themselves to the usual commercial supply agreements.
What distinguishes this from a typical strategic investment is the alignment of equity with physical supply. In a conventional strategic round, a chip supplier might take a small minority stake to signal partnership while negotiating supply contracts separately. Here, the memory manufacturers, hyperscale cloud providers, and the model developer are bound into the same financing at the same moment that allocations for HBM3E and HBM4 are being contested across the industry. The structure resembles a vertically integrated supply agreement priced as equity: capital flows in, memory and compute allocation flow out, and the upside is shared rather than renegotiated annually.
Anthropic framed the memory makers’ role in supply terms, noting that the relationships will help it scale compute reliably as demand for Claude grows. The precise investment amounts, any board observer rights, and the multi-year supply commitments attached to their participation were not disclosed. But even at the level of disclosure available, the configuration is unusual: it is rare for all three dominant high-bandwidth memory suppliers to sit on the same private company’s register simultaneously.
The numbers behind the valuation
Anthropic’s enterprise adoption of Claude has accelerated sharply. The company said its run-rate revenue crossed $47 billion earlier in May, up from a reported annual run rate of roughly $1 billion at the start of 2025, driven largely by API usage from coding tools and enterprise deployments. It framed the raise not as cash for a speculative bet but as fuel for a business already expanding fast, with capital earmarked for safety and interpretability research, compute expansion, and product scaling. The round closed shortly after Anthropic released Claude Opus 4.8, which the company says improves performance on agentic tasks, advanced coding, and consistency on long-running work.
Investors familiar with the process described the allocation as oversubscribed, with existing holders requesting larger pro-rata participation than was available.
The frontier lab arms race in numbers
The Anthropic round sits inside a broader pattern. OpenAI closed $122 billion in committed capital on 31 March 2026 at an $852 billion post-money valuation. Read together, the two raises represent a concentration of private capital in frontier AI without real precedent.
Silicon Canals previously covered OpenAI’s earlier capital-intensive phase and Anthropic’s earlier funding trajectory. The trendline since then has bent sharply upward: Anthropic’s own valuation has more than doubled in three months.
What the structure rewards
Brad Gerstner, founder and chief executive of Altimeter Capital, which co-led the round, argued that Claude’s recent technical advances have driven large-scale adoption among the most demanding enterprise organisations and position Anthropic to lead the next phase of AI. Stripped of the framing, the institutional logic is straightforward: at this scale, frontier model development requires capital commitments only sovereign-adjacent pools can supply, and the only viable exit is a public market large enough to absorb the eventual float.
That dynamic compresses the field. The companies that can raise at these levels, principally Anthropic and OpenAI, increasingly resemble infrastructure utilities funded by the same handful of crossover funds and chip suppliers. Whether that compression hardens into something permanent, or whether a generation of smaller labs finds a way to operate beneath the ceiling these rounds have set, is the question the cap table cannot answer. What it does suggest is that by the time the IPO arrives, the most consequential decisions about who builds frontier AI, and on whose silicon, will already have been made in rooms like this one — and the rest of the ecosystem will be working out what to do about it.
