A massive financing deal tied to Anthropic is reshaping how artificial intelligence infrastructure is funded, as private credit giants Apollo Global Management and Blackstone assemble what could become one of the largest private debt transactions ever linked to the AI industry.

The roughly $36 billion structure, first reported by Bloomberg, is designed to finance huge volumes of computing hardware for Anthropic without placing the debt directly on the company’s balance sheet. Instead of borrowing conventionally, Anthropic would lease AI chips through a special-purpose financing vehicle created specifically for the transaction.

The structure is seen as another piece of evidence that the economics of frontier AI are rapidly converging with large-scale infrastructure finance, turning computing power into an asset class increasingly funded like aircraft fleets, pipelines, or telecom towers.

Google’s custom tensor processing units, or TPUs, which have become an alternative to Nvidia’s AI accelerators for companies building massive language models, lead the arrangement.

Under the proposed transaction, borrowed funds would be used to acquire TPUs that would then be leased back to Anthropic for deployment across data centers in New York, Texas, Louisiana, and Indiana. The financing mechanism offers Anthropic a crucial advantage: access to enormous amounts of compute capacity without immediately burdening its own balance sheet with tens of billions of dollars in debt obligations.

That matters because AI companies are facing a new reality in which compute availability has become just as strategically important as model quality. Training and deploying advanced AI systems requires infrastructure spending measured not in millions, but in tens of billions of dollars annually.

The deal also reveals how deeply interconnected the AI supply chain has become. Broadcom, which works with Google on TPU development, is reportedly providing a residual value support agreement on the senior portions of the debt.

That effectively means Broadcom would absorb losses for top-tier lenders if Anthropic defaulted and resale values of the chips failed to cover repayment obligations. The arrangement gives investors an additional layer of protection in what would otherwise be a highly specialized and technologically volatile asset-backed financing structure.

The debt itself is reportedly divided into several tranches, including roughly $6 billion of A1 notes, $25 billion of A2 notes, and $4.5 billion of riskier B notes, though the figures may still change before closing.

Rather than retaining all the exposure internally, Apollo and Blackstone are syndicating portions of the debt to outside investors, a model more commonly associated with leveraged buyouts and structured credit markets.

That approach denotes growing institutional appetite for AI-linked infrastructure exposure as pension funds, insurers, and asset managers search for higher-yielding investments tied to the global AI boom.

The structure is seen as another example of private capital markets stepping into roles traditionally occupied by banks. Regulatory constraints and the sheer scale of AI infrastructure spending are pushing more financing activity toward private credit firms capable of assembling complex, multi-billion-dollar funding packages quickly.

Another notable aspect of the deal is its staged funding model. Instead of releasing all capital upfront, financing draws will reportedly occur gradually as chips are delivered and lease agreements begin. That reduces idle capital costs for investors while aligning funding schedules with the physical rollout of infrastructure.

The transaction arrives during an extraordinary escalation in AI spending globally.

Anthropic recently announced a new funding round valuing the company at approximately $965 billion post-money, surpassing the valuation of rival OpenAI. Both firms are reportedly exploring potential IPOs as soon as this year, amid investor demand for exposure to the AI sector.

Analysts are seeing the financing deal as a signal of the emergence of a more mature AI infrastructure economy. In the early phase of the generative AI boom, companies largely relied on direct equity funding from venture capital firms and hyperscalers. Now, the industry is evolving toward highly engineered financing structures involving leasing, securitization, structured debt, and infrastructure-style capital deployment.

That transition could have profound implications for the sector. Financiers may help accelerate expansion while distributing risk across broader capital markets by separating ownership of compute infrastructure from AI model companies themselves. At the same time, it introduces new vulnerabilities tied to hardware depreciation, technological obsolescence, and long-term demand assumptions for AI services.

However, the deal has also revealed something else.

While Nvidia remains dominant in AI accelerators, Google’s TPUs are becoming important for large-scale model training and inference, particularly for companies seeking more diversified supply chains amid persistent chip shortages and soaring GPU costs.