Key Takeaways
- • Apollo and Blackstone are jointly financing a $35 billion expansion of AI computing capacity for Anthropic, anchoring a new partnership with chipmaker Broadcom that launches the "AI XPV Platform."
- • The initial commitment will expand Anthropic's computing capacity by one gigawatt (sufficient to power approximately 750,000 homes), with deployment beginning mid-2026 at Fluidstack-operated data-center sites.
- • The broader partnership targets deployment of more than 20 gigawatts of computing capacity across leading AI labs, including OpenAI, through 2028, positioning Broadcom's custom chips as an alternative to Nvidia's AI infrastructure dominance.
- • The deal reflects private-equity firms emerging as critical funding sources for AI infrastructure—following Meta's $27 billion financing arrangement with Blue Owl Capital in October 2025—as companies face capacity constraints and supply limitations on specialized AI hardware.
- • Broadcom gains significant strategic momentum in reducing tech companies' reliance on Nvidia through custom AI chips and integrated computing systems designed to lower both cost and power consumption for AI model training.
Summary
Apollo and Blackstone have anchored a $35 billion financing commitment to fund Anthropic's AI computing capacity expansion through a structured partnership with Broadcom announced on June 9, 2026. The transaction establishes the AI XPV Platform, a dedicated infrastructure vehicle designed to finance and deploy custom AI chips and computing systems across leading artificial intelligence laboratories. The initial tranche covers a one-gigawatt capacity increase for Anthropic, equivalent to the power requirement for approximately 750,000 homes, with physical infrastructure provided by cloud computing operator Fluidstack beginning mid-2026.
The partnership extends significantly beyond Anthropic's immediate requirements. The joint venture targets deployment of more than 20 gigawatts of computing capacity across multiple leading AI research institutions, including OpenAI, with completion targeted by 2028. This scale positions the initiative as a material response to industry-wide constraints on AI infrastructure availability and the high capital intensity of deploying cutting-edge training and inference capacity.
The deal underscores the critical role private-equity and credit providers now play in funding AI infrastructure buildout. As major technology companies face supply constraints on specialized processors and rising capital requirements, institutional investors have shifted from traditional venture equity into direct infrastructure financing and long-duration private credit structures. Meta's October 2025 arrangement with Blue Owl Capital—a $27 billion facility for data-center development—established a comparable precedent for large-scale, PE-anchored infrastructure financing in the AI sector.
For Broadcom, the partnership advances its strategic objective of capturing share in the custom silicon market. By bundling Broadcom's proprietary AI chips and networking solutions with structured financing, the initiative positions the company's technology stack as a practical alternative to Nvidia's dominant GPU offerings. The emphasis on reducing both cost and power consumption in AI model training addresses a core pain point for AI labs operating at scale, where electricity and semiconductor costs represent substantial operational burdens.
The partnership extends significantly beyond Anthropic's immediate requirements. The joint venture targets deployment of more than 20 gigawatts of computing capacity across multiple leading AI research institutions, including OpenAI, with completion targeted by 2028. This scale positions the initiative as a material response to industry-wide constraints on AI infrastructure availability and the high capital intensity of deploying cutting-edge training and inference capacity.
The deal underscores the critical role private-equity and credit providers now play in funding AI infrastructure buildout. As major technology companies face supply constraints on specialized processors and rising capital requirements, institutional investors have shifted from traditional venture equity into direct infrastructure financing and long-duration private credit structures. Meta's October 2025 arrangement with Blue Owl Capital—a $27 billion facility for data-center development—established a comparable precedent for large-scale, PE-anchored infrastructure financing in the AI sector.
For Broadcom, the partnership advances its strategic objective of capturing share in the custom silicon market. By bundling Broadcom's proprietary AI chips and networking solutions with structured financing, the initiative positions the company's technology stack as a practical alternative to Nvidia's dominant GPU offerings. The emphasis on reducing both cost and power consumption in AI model training addresses a core pain point for AI labs operating at scale, where electricity and semiconductor costs represent substantial operational burdens.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.