Blackstone and Apollo Arrange $36B TPU Financing for Anthropic

Blackstone and Apollo are lining up what could be one of the largest private credit deals in history — roughly $36 billion to finance Anthropic’s next wave of AI compute infrastructure. The financing, reported by Reuters and Bloomberg, will let Anthropic buy custom tensor processing units (TPUs) from Google and lease them to power its Claude family of models as the company races to scale. Deal mechanics and timeline - Apollo Global Management and Blackstone are syndicating the $36 billion package while planning to retain “significant portions” on their own books. - Broadcom — a co-designer of Google’s TPUs — is backstopping the largest tranches, effectively de‑risking parts of the financing by guaranteeing payments. - Investors have reportedly been asked to submit orders this week, and the transaction could close as soon as next week, though final terms may still shift as books close. Why Anthropic needs it Anthropic plans to use the chips to expand TPU capacity for Claude and related models. This financing sits atop an already complex capital structure: in April, Anthropic expanded its partnership with Google and Broadcom to secure access to roughly 3.5 gigawatts of TPU compute, with deployments slated to begin scaling from 2027 as part of a broader $50 billion push into domestic U.S. compute capacity. The company also announced a $6.5 billion equity raise at a $965 billion post-money valuation in April. Growth and scrutiny Anthropic’s commercial momentum helps explain investor appetite: the company’s revenue run rate recently topped $30 billion — up from about $9 billion at the end of 2025 — as its enterprise API market share reportedly grew from roughly 12% in 2023 to 32% by mid‑2025. Large financial and industrial clients have integrated Claude into production workflows. At the same time, regulators have taken notice: U.S. Treasury officials convened major bank CEOs in April over cyber risks tied to Anthropic’s upcoming Claude Mythos model after internal tests and a code leak revealed the model’s ability to surface large numbers of software vulnerabilities. Broadcom and the hardware-finance convergence Broadcom’s decision to guarantee major slices of the loan highlights a shift: AI hardware suppliers are increasingly acting like structured finance counterparties rather than pure vendors. Broadcom already sits at the center of Google’s TPU roadmap and is expected to support future TPU iterations — including next‑generation designs where partners such as Marvell are exploring improvements in memory bandwidth and model efficiency. Private equity isn’t standing on the sidelines The Blackstone–Apollo transaction is also part of a broader push by private capital into AI infrastructure. Earlier this month, Anthropic and a group including Blackstone, Goldman Sachs, Apollo and Hellman & Friedman launched a $1.5 billion venture to accelerate Claude adoption across sectors from healthcare to manufacturing. What this means for crypto and tokenized AI projects For crypto markets, the deal reinforces a familiar dynamic: massive pools of institutional capital are clustering around a very small number of AI platforms that command outsized shares of cloud, chip and power budgets. That concentration raises two related implications for the on‑chain and tokenized AI ecosystem: - Competitive pressure: Tokenized AI and decentralized compute projects will face tougher economics competing with well‑capitalized, vertically integrated incumbents that can secure long‑term hardware agreements and favorable financing. - Opportunity for differentiation: Decentralized providers can still carve out niches — for example by offering lower‑cost, geographically distributed compute, specialized privacy-preserving services, or token-driven incentives — but they’ll need to demonstrate clear cost, latency, or security advantages. Bottom line The $36 billion chip financing — underwritten and held in part by private equity and partially guaranteed by a major semiconductor supplier — is a major vote of confidence in Anthropic’s growth trajectory and in the centrality of bespoke AI hardware. It also underscores how off‑chain institutional capital is shaping the competitive landscape for AI compute, creating both headwinds and strategic openings for crypto-native infrastructure and tokenized AI projects.