Blackstone has held early discussions with investors about a potential large-scale debt package to finance Anthropic’s use of Google AI accelerator chips, according to people familiar with the matter. An initial proposal for the new financing was for at least $36 billion, a figure that would exceed the $35 billion package that Apollo Global Management and Blackstone arranged roughly two months earlier for Anthropic’s lease of Google AI chips.

Key Facts

• Blackstone has held early talks with investors about a debt package for Anthropic’s access to Google AI chips.
• An initial proposal was for at least $36 billion.
• The potential deal would top the $35 billion private credit transaction arranged by Apollo and Blackstone about two months ago.
• The earlier transaction was one of the largest private credit deals in history.
• Google’s infrastructure financing program tied to Anthropic is estimated to cover more than $150 billion of TPU chips.
• An April arrangement gave Anthropic access to Google AI processing power and included commitments by Anthropic to pay Google about $200 billion over five years.
• Google said it would invest an additional $10 billion or more in Anthropic as part of the arrangement.

Blackstone’s Potential Debt Package

The discussions are at an early stage, and the final size, structure and terms of any transaction could change. Even so, the preliminary figure of at least $36 billion is significant because it would surpass the $35 billion financing that Apollo Global Management and Blackstone assembled only about two months ago. That earlier deal was also intended to fund Anthropic’s lease of Google AI chips and was described as one of the largest private credit transactions in history.

A new package of that scale would show how quickly private capital markets are being reshaped by the artificial intelligence boom. AI companies need vast amounts of computing power, and the cost of securing chips, data-center capacity and cloud services is rising faster than many companies can fund from operating cash flow. Debt financing allows those commitments to be converted into long-term obligations that investors can hold, while the technology company gains access to infrastructure it needs to train and run models.

Blackstone’s role is notable because the firm is not simply buying equity in an AI startup. It is helping to arrange financing tied to a commercial relationship between Anthropic and Google, two companies whose ties already run deep. Google was one of Anthropic’s earliest investors, and the relationship has expanded into a broader infrastructure partnership involving cloud capacity and custom chips.

The Scale of Private Credit Financing

Private credit has become a major source of financing for large, complex transactions that might once have been handled by banks or public bond markets. Asset managers including Apollo and Blackstone have raised enormous pools of capital and can write large checks quickly. That speed matters in AI infrastructure, where companies are racing to lock in computing capacity and where contracts can stretch for years.

The $35 billion transaction arranged earlier this year illustrated the appetite. It was structured around Anthropic’s lease of Google AI chips, giving investors exposure to a long-term payment stream linked to one of the fastest-growing AI developers. The new proposal suggests that demand for such financing has not faded. If anything, the reported $36 billion figure indicates that the market is prepared to go even larger.

For investors, the appeal is a yield-bearing asset tied to a strategic technology company. For Anthropic, the benefit is access to capital without relying solely on equity fundraising. For Google, the financing helps support demand for its cloud and chip capacity. The arrangement is therefore more than a simple loan. It sits at the intersection of venture investment, cloud computing, semiconductor supply and private credit.

Circular AI Financing Raises Questions

The proposed deal comes as technology companies strike increasingly complex and often circular deals involving AI chips, data centers and other infrastructure. In some cases, an investor backs a company, the company commits to buy cloud or computing services, and a financier provides debt against that commitment. Each part of the arrangement can make commercial sense on its own, but the links between them can make it harder to assess where risk ultimately sits.

Circularity is not automatically a problem. Long-term supply agreements are common in capital-intensive industries, and strategic investors often provide both capital and commercial support. But the AI sector is unusual because the amounts are enormous, the technology is evolving quickly and many of the key players are closely interconnected. When one company invests in another, leases capacity from a third and borrows against the resulting contract, investors need to understand the full chain of obligations.

The concern is that reported revenue and backlog can be inflated by deals among related parties. If Anthropic commits to pay Google for cloud services, and Google is also an investor in Anthropic, the transaction can look like both revenue for Google and a funding requirement for Anthropic. If private credit investors lend against that commitment, the quality of the debt depends on Anthropic’s ability to generate enough cash over time to meet the payments.

Supporters argue that the arrangements reflect real demand. AI models require massive compute, and companies are willing to sign long-term contracts to secure access. The financing simply matches long-lived infrastructure assets with long-term capital. Critics counter that the web of commitments can obscure leverage and concentration, especially if the same small group of companies appears on multiple sides of the transactions.

Google’s Expanding Role as Backstop

Google’s involvement extends beyond a simple equity stake. The company has been increasingly backstopping financing deals that involve buying chips and computing power through Google Cloud. Its infrastructure financing program, constructed to enable Anthropic to buy access to Google’s own chips and data-center capacity, is estimated to encompass more than $150 billion worth of Tensor Processing Unit chips.

That estimate underscores the scale of Google’s commitment to Anthropic and to the broader AI infrastructure race. Google designs TPUs as an alternative to general-purpose graphics processors for AI workloads. By tying Anthropic’s compute needs to Google Cloud and TPUs, Google can deepen a strategic relationship with a leading AI developer while also creating demand for its own hardware and cloud services.

The April arrangement giving Anthropic access to Google AI processing power included commitments by Anthropic to pay Google about $200 billion over the next five years, according to reports. The figures suggested that Anthropic accounted for more than 40 percent of the cloud revenue backlog Google disclosed at the time. Google also said it would invest an additional $10 billion or more in Anthropic as part of the arrangement.

Those numbers help explain why financing is needed. A commitment of that size is far beyond the scale of a typical cloud contract. It requires Anthropic to secure capital to support its payments and infrastructure buildout, and it gives financiers a long-term cash flow to lend against. It also gives Google a powerful incentive to ensure that Anthropic has the resources to honor its commitments.

Anthropic’s Compute Commitments

Anthropic is one of the leading developers of large language models and is the company behind Claude. Like its competitors, it depends on enormous amounts of computing power to train models and serve customers. Access to chips is a strategic bottleneck, and the ability to lock in supply through long-term agreements can be as important as raising equity.

The company has raised billions of dollars from strategic and financial investors, including Google. But equity alone may not be enough to fund the scale of infrastructure that AI developers now require. Debt can help bridge the gap, particularly when it is secured by long-term contracts with creditworthy counterparties. The proposed Blackstone package would be another example of Anthropic using private credit to convert future compute needs into present capacity.

The structure also reflects the changing economics of AI. Early-stage AI companies once focused mainly on research talent and model quality. Now they must also manage data-center leases, chip procurement, cloud contracts and financing arrangements. The balance sheet of an AI developer is beginning to look more like that of a capital-intensive infrastructure company, with long-term obligations and complex supplier relationships.

Why Chip Access Matters

AI accelerators are the engines of modern machine learning. Training large models requires thousands of chips working together for weeks or months, and serving those models to millions of users requires continuous access to compute. Supply constraints have made chips one of the most valuable resources in the technology industry.

Google’s TPUs are custom-designed for its own cloud infrastructure and AI workloads. For Anthropic, access to TPUs through Google Cloud can provide an alternative to buying Nvidia GPUs, which have been in extreme demand. The arrangement also ties Anthropic more closely to Google’s cloud platform, which can offer integrated tools, networking and support.

The reported $150 billion estimate for Google’s TPU-related financing program shows how large the chip commitment has become. It is not simply a purchase order. It is a multiyear infrastructure program that involves financing, cloud capacity and strategic investment. The new Blackstone discussions are another layer of that program, aimed at providing the debt capital needed to make the commitments work.

Market Implications

If completed, the new package would reinforce the role of private credit in financing the AI buildout. Banks have become more cautious about certain types of leveraged lending, and private credit funds have filled the gap. The largest asset managers can now arrange transactions that rival public bond deals in size and complexity.

The deal would also strengthen the ties between Blackstone, Apollo, Google and Anthropic. That concentration could be efficient, because the parties already understand the contracts and the technology. But it could also create interconnected exposures that are difficult for outside investors to assess. A problem at one company could ripple through the financing chain.

For the broader market, the transaction would be a test of investor appetite for AI-related debt. If investors are willing to commit at least $36 billion to a single AI infrastructure package, it would signal that private capital remains eager to gain exposure to the sector. If the deal struggles, it could suggest that concerns about circularity and concentration are beginning to limit demand.

Risks and Regulatory Scrutiny

Large financing arrangements involving strategic investors can attract scrutiny from regulators, rating agencies and investors. The key questions are likely to include the nature of the collateral, the enforceability of long-term contracts, the credit quality of the counterparties and the concentration of exposure.

Circular deals may also raise accounting and disclosure questions. If revenue depends on financing arranged by an investor, or if a cloud contract is backed by debt raised against the same contract, the economic substance can be complex. Investors will want to know whether the cash flows are truly independent and whether the obligations are properly reflected in financial statements.

There is also technology risk. AI hardware and model architectures are changing rapidly. A long-term commitment to a particular chip or cloud platform could become less valuable if the technology shifts. Google’s TPUs are powerful, but the competitive landscape in AI chips is evolving, and Anthropic may want flexibility in its compute supply over time.

Finally, there is execution risk. The reported discussions are early, and the final package may be smaller, differently structured or abandoned. Even a large debt financing must be syndicated or placed with investors, and market conditions can change quickly. The fact that a $35 billion deal was completed recently does not guarantee that a larger one will find the same reception.

The scale of the proposed package, the speed of the financing and the concentration of exposure across a small group of technology and asset-management firms make the transaction one of the most closely watched tests of how the AI buildout will be funded.

Source:Silicon UK News