Private equity and alternative asset management behemoth Blackstone Group LP has initiated preliminary discussions with investors to gauge interest in a second massive AI-related debt financing package, aimed at funding AI application leader Anthropic's use of Alphabet's Google TPU computing chips. According to sources familiar with the matter, one initial proposal seeks to raise at least $36 billion in debt. The sources, who spoke on condition of anonymity as they were not authorized to discuss the matter publicly, noted that details including the financing size, structure, and whether Blackstone will ultimately lead the deal are still under discussion and subject to change.
These large-scale AI financings primarily strengthen the order certainty of the computing power supply chain, with secondary concerns around AI bubble risks and liquidity issues. The proposed second round of at least $36 billion from Blackstone, combined with the already completed $35 billion AI XPV first tranche deal from Apollo and Blackstone, essentially monetizes a portion of Anthropic's expected future Claude revenue. This capital is converted into current orders for massive AI infrastructure, including Google TPU computing clusters, optical communication/interconnect, data center CPUs, high-performance Ethernet switching, HBM/DRAM/NAND storage components, and data center power systems.
The AI XPV platform's first tranche funds over 1 gigawatt of computing power and plans to support deployments exceeding 20 gigawatts by 2028. This means chip orders no longer depend solely on AI labs' current cash flow; they can be "capital marketized" through special purpose vehicles, long-term leases, and supplier credit support. Consequently, near-term order visibility for Broadcom's AI chip supply chain, Google's TPU ecosystem, network equipment, power, and data center hardware suppliers is significantly extended.
The risk lies in the fact that financial engineering does not eliminate repayment risk; it merely shifts it from AI lab balance sheets to SPV creditors, chip suppliers, cloud platforms, and their guarantors. This AI boom is not necessarily doomed to burst, but its pricing core is shifting from technology adoption rates to whether the credit chain can be sustained by genuine cash flows. Citadel Securities predicts that debt financing for AI chips could exceed $500 billion by 2028, making "credit leverage risk brewing" a reasonable analyst warning.
The AI Financing Race Shatters Records
If the proposed transaction closes near the reported size, it would surpass the $35 billion debt financing arranged by Apollo Global Management and Blackstone approximately two months ago for Anthropic to lease Google's custom AI chips. That deal was one of the largest credit transactions in private credit market history. Representatives for Blackstone, Apollo, Anthropic, and Google declined to comment.
As Silicon Valley races to build AI infrastructure, leading companies have entered into complex, often circular financing transactions to secure computing resources. Google, an early investor in Anthropic that has repeatedly purchased equity stakes, is now increasingly providing guarantee support for the data center construction underpinning this AI startup's massive expansion.
This potential new financing follows Anthropic's confidential filing for an initial public offering in the U.S. stock market. The company is attempting to beat competitor OpenAI to the public markets. As the developer of the Claude AI model series, Anthropic plans to lease high-performance computing chips across five data centers with Google's assistance, supported by the previous AI debt financing arrangement.
Global technology companies are utilizing all corners of the credit market to meet AI's unprecedented capital demands, forcing Wall Street asset managers to collectively design new debt structures to keep pace with the industry's expansion. Due to concerns that AI investments may not yield expected returns, some companies have recently been forced to pay higher benchmark yields on new debt issuances.
Broadcom, Apollo, and Blackstone established a cooperation platform called "AI XPV Platform" this year to help leading AI technology developers, including Anthropic, finance their computing infrastructure. The $35 billion AI-related debt financing completed about two months ago was the platform's first transaction. In that deal, Broadcom provided backstop financing support for repaying the largest senior debt tranche. Morgan Stanley reportedly advised Broadcom and assisted in arranging the deal. Broadcom's representative did not respond to requests for comment.
Wall Street Begins 'Securitizing Computing Power'
The $35 billion AI XPV financing supports Anthropic's expansion of over 1 gigawatt of computing power, while Blackstone's discussed second round is at least $36 billion. These funds are converting Anthropic's future strong revenue expectations into current TPU and data center orders. The platform's first tranche supports over 1GW and aims to drive global AI infrastructure deployment exceeding 20GW.
This means chip orders no longer depend entirely on AI labs' current cash flow; they can be "capital marketized" by Wall Street asset managers through SPVs, long-term leases, and supplier credit support. Consequently, near-term order visibility for Broadcom's AI chip chain, Google's TPU ecosystem, network equipment, power, and data center hardware suppliers is significantly extended. The risk is that financial engineering merely shifts repayment risk from AI lab balance sheets to SPV creditors, chip suppliers, cloud platforms, and their guarantors.
Anthropic's financing structure involves an SPV purchasing Google TPUs and leasing them to Anthropic. The largest senior debt tranche relies on Broadcom providing shortfall or residual value support, allowing the approximately $25 billion core debt to achieve a yield of around 5.75%. Subordinated debt, lacking equivalent guarantees, has a yield of about 8.5%. This forms a typical "supplier financing loop": chip manufacturers facilitate customer purchases through credit backing, which in turn becomes revenue and valuation justification. If Anthropic's revenue growth, computing utilization, or IPO proceeds fall short, risk propagates backward through the chain of lease payments, SPV debt servicing, chip residual value, and supplier guarantees.
Facebook parent company Meta demonstrates that such off-balance-sheet capital commitments have escalated from project-level arrangements to core financing models for mega-cap tech firms. As of end-June 2026, Meta had $278.99 billion in commitments that had not yet commenced lease, thus not recorded as balance sheet lease liabilities. This figure significantly expanded from the previous quarter. In July, it signed approximately $68 billion in data center leases with 18-20 year terms. There are also $349.31 billion in irrevocable contract commitments and up to $14.72 billion in contingent cloud capacity purchase obligations. Meanwhile, Meta has raised its 2026 capital expenditure guidance to $130-$145 billion.
Strictly speaking, these large AI financings and debt projects are not traditional "hidden debt," but their economic substance is long-term fixed payment obligations. Once facilities become operational, lease liabilities gradually appear on balance sheets, but cash flow pressure is already locked in by contracts. Citadel Securities forecasts that by 2028, new debt issuance solely for AI chip purchases could exceed $500 billion, representing over 5% of the U.S. investment-grade bond index at that time. Single-year issuance in 2028 alone could surpass $250 billion, with most debt maturities of only three to five years to match chips' shorter economic lives.
The real danger is not simply "too much debt," but the maturity mismatch between short-duration debt, rapidly depreciating assets, and long-term, highly uncertain AI revenue. If interest rates remain high, model inference prices decline, chip architecture updates cause old equipment residual values to plummet, or credit markets demand higher issuance premiums, many projects could face a simultaneous refinancing wall around 2028. This could crowd out technology, media, and telecom bond allocations, triggering credit spread widening and private credit liquidity contraction.
Thus, some Wall Street analysts argue that investment strategy should separately value "financed orders" and "terminal cash demand." Priority should go to diversified, net-cash-rich chip and infrastructure leaders with prepayments or irrevocable order commitments and sustainable free cash flow. For projects dependent on SPVs, supplier guarantees, circular investments, and rent repayment from a single AI lab, risk premiums must be increased. Key metrics to monitor include credit spreads, CDS, lease-adjusted leverage ratios, chip residual values, and computing utilization rates.
Overall, this AI boom is not inevitably destined to burst, but its pricing core is shifting from technology penetration rates to whether the credit chain can be sustained by genuine cash flows.
Comments