Apollo expects AI startups to seek debt financing earlier than previous tech companies

Photo: Tima Miroshnichenko / Pexels

Apollo expects AI startups to seek debt financing earlier than previous tech companies

The private credit giant is structuring off-balance-sheet deals worth tens of billions to fund AI compute, reshaping how the industry gets built

Apollo executives are now openly predicting that AI companies will turn to debt markets far sooner in their lifecycles than software predecessors ever did. The reason is straightforward: building AI requires physical infrastructure on a scale that equity financing alone cannot comfortably absorb.

On June 5, 2026, Apollo and Blackstone closed a roughly $35 billion debt package for Anthropic, structured to help the AI company procure and lease Google Tensor Processing Units through a special-purpose vehicle called Apollo Atlas SP Partners. The deal is tiered: senior portions yield around 5.75%, backed by Broadcom’s credit support, while junior tranches carry a higher yield of 8.5% to compensate for additional risk.

The structure is designed so that Anthropic’s balance sheet stays relatively clean. Hardware acts as collateral, lease payments provide cash flow, and the SPV absorbs the liability.

Advertisement

Earlier in 2026, Apollo had already put roughly $7 billion to work in Elon Musk’s xAI, specifically to fund access to Nvidia GPUs. That deal carried a 10% coupon, and Apollo has since reported approximately $250 million in paper gains on the position.

The firm is also exploring a potential increase to an OpenAI-linked loan tied to SoftBank, possibly lifting that facility to $9 billion.

AI companies face a different constraint than classic enterprise software predecessors. Training large models and running inference at scale requires enormous quantities of specialized chips, the kind that cost hundreds of thousands of dollars per unit and need to be reserved, sometimes years in advance. A startup in the large language model space can credibly need billions in hardware before it has generated meaningful revenue.

Apollo originated a record $309 billion in loans in 2025, and its executives have framed AI infrastructure as a central pillar of where that volume is headed. The firm projects that AI-driven capital expenditures could reach into the trillions by 2030.

For private credit investors, the appeal is yields between 5.75% and 10% on deals collateralized by hardware that major AI companies need to operate, with senior tranches carrying additional credit support from names like Broadcom. Blackstone’s involvement in the Anthropic deal suggests Apollo is not alone in recognizing the opportunity.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
Apollo expects AI startups to seek debt financing earlier than previous tech companies
Apollo expects AI startups to seek debt financing earlier than previous tech companies

The private credit giant is structuring off-balance-sheet deals worth tens of billions to fund AI compute, reshaping how the industry gets built

Photo: Tima Miroshnichenko / Pexels

Apollo executives are now openly predicting that AI companies will turn to debt markets far sooner in their lifecycles than software predecessors ever did. The reason is straightforward: building AI requires physical infrastructure on a scale that equity financing alone cannot comfortably absorb.

On June 5, 2026, Apollo and Blackstone closed a roughly $35 billion debt package for Anthropic, structured to help the AI company procure and lease Google Tensor Processing Units through a special-purpose vehicle called Apollo Atlas SP Partners. The deal is tiered: senior portions yield around 5.75%, backed by Broadcom’s credit support, while junior tranches carry a higher yield of 8.5% to compensate for additional risk.

The structure is designed so that Anthropic’s balance sheet stays relatively clean. Hardware acts as collateral, lease payments provide cash flow, and the SPV absorbs the liability.

Advertisement

Earlier in 2026, Apollo had already put roughly $7 billion to work in Elon Musk’s xAI, specifically to fund access to Nvidia GPUs. That deal carried a 10% coupon, and Apollo has since reported approximately $250 million in paper gains on the position.

The firm is also exploring a potential increase to an OpenAI-linked loan tied to SoftBank, possibly lifting that facility to $9 billion.

AI companies face a different constraint than classic enterprise software predecessors. Training large models and running inference at scale requires enormous quantities of specialized chips, the kind that cost hundreds of thousands of dollars per unit and need to be reserved, sometimes years in advance. A startup in the large language model space can credibly need billions in hardware before it has generated meaningful revenue.

Apollo originated a record $309 billion in loans in 2025, and its executives have framed AI infrastructure as a central pillar of where that volume is headed. The firm projects that AI-driven capital expenditures could reach into the trillions by 2030.

For private credit investors, the appeal is yields between 5.75% and 10% on deals collateralized by hardware that major AI companies need to operate, with senior tranches carrying additional credit support from names like Broadcom. Blackstone’s involvement in the Anthropic deal suggests Apollo is not alone in recognizing the opportunity.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.