Anthropic Secures $35B Private Credit Deal For Google TPUs

Anthropic has secured a $35 billion private-credit package backed by Apollo Global Management and Blackstone to finance Google-designed tensor processing units for its Claude models.

The structure puts another giant Wall Street funding channel behind AI infrastructure. A special-purpose vehicle will buy TPUs from Alphabet and lease the chips back to Anthropic, giving the company dedicated compute without funding the full hardware cost through equity alone.

The deal was arranged around Apollo’s Atlas SP Partners platform and includes several asset-backed tranches. Broadcom, which works with Google on custom TPU systems, is tied to the senior portions of the financing through residual-value support, giving investors more comfort if Anthropic fails to meet lease payments and the hardware later sells below the amount owed.

For Anthropic, the financing extends a compute buildout that is already pushing across Google, Broadcom, Amazon, SpaceX and other infrastructure partners. The company recently raised $65 billion at a $965 billion valuation, making compute access one of the biggest variables in whether that valuation can be supported by future Claude demand.

Google TPUs Move Deeper Into Claude’s Compute Stack

Anthropic has already expanded its Google and Broadcom TPU partnership, with multiple gigawatts of next-generation TPU capacity expected to come online starting in 2027. The chips are designed to support frontier Claude models and reduce dependence on one hardware path as AI demand keeps rising.

That diversification matters because Nvidia GPUs remain the most important scarce resource in frontier AI, but TPUs give Anthropic a custom-chip route tied closely to Google’s cloud infrastructure. The latest financing turns that strategic relationship into a financial structure, with debt investors now helping fund the hardware layer that supports model training and inference.

The move also connects to a broader AI chip reshuffle. Google and Nvidia are already exploring backup manufacturing options as demand stretches the supply chain, with Intel shares jumping on reports tied to future TPU and AI chip manufacturing demand. That pressure shows how quickly AI labs, cloud providers and chip designers are trying to secure capacity before the next wave of model scaling.

SpaceX GPU Rent Adds To The Circular Spending Debate

The private-credit package also lands as Anthropic is tied to a separate high-cost compute arrangement with SpaceX. SpaceX’s AI infrastructure business has drawn attention after Google agreed to pay $920 million per month for compute capacity, while Anthropic has also been linked to large-scale Nvidia GPU access through SpaceX’s Colossus infrastructure.

That makes the Anthropic-Google-SpaceX triangle a major focus for investors. Alphabet owns a large stake in Anthropic, sells or provides access to TPUs through Google infrastructure, and is also renting Nvidia GPU capacity from SpaceX to meet its own AI demand. Anthropic, meanwhile, is raising equity, using private credit and signing compute deals to keep Claude competitive.

Skeptics, including short sellers such as James Chanos, have questioned whether the AI infrastructure boom is turning into circular financing, where hyperscalers, AI labs and data-center operators keep buying, leasing and funding one another’s capacity at rapidly rising valuations.

The concern is not that the demand is fake. Claude, Gemini, ChatGPT and enterprise AI tools are clearly consuming more compute. The concern is whether the economics of that compute will justify the amount of equity, debt and long-term lease exposure now being created around it.

AI Compute Becomes A Financing Arms Race

Anthropic’s $35 billion package shows how AI infrastructure is moving beyond traditional cloud contracts. Private credit funds, chipmakers, cloud providers and AI labs are now part of the same funding chain, with hardware increasingly treated like a financeable asset class.

That shift also explains why companies are working harder to control model costs. Coinbase recently said AI costs are staying roughly flat by routing prompts to cheaper models, even as token usage grows sharply. Large AI labs face the same problem at a much bigger scale: usage keeps rising, and every model improvement requires more disciplined compute allocation.

The pressure is spreading into labor markets, chip supply chains and capital markets at the same time. Banks are preparing for AI-driven workforce cuts as automation changes staffing, while AI companies are raising tens of billions to secure the hardware that makes those productivity gains possible.

Anthropic now has another major financing lane for Claude’s compute needs. The harder test comes later: whether the revenue generated by Claude, enterprise AI agents and model subscriptions can support a stack of equity funding, TPU leases, GPU rental commitments and private-credit obligations that is growing almost as fast as the models themselves.

The post Anthropic Secures $35B Private Credit Deal For Google TPUs appeared first on Crypto Adventure.