June 9 - Two major private-equity firms, Apollo and Blackstone, are providing financing for a $35 billion expansion of AI computing capacity for Anthropic, the company behind the Claude Code family of models. The project will rely on Broadcom’s custom chips and networking technologies as part of a coordinated tie-up between the asset managers and the chipmaker.
Initial tranche and deployment plans
The partners said the initial investment will increase Anthropic’s AI computing capacity by one gigawatt. The companies noted that one gigawatt of compute infrastructure is roughly equivalent to the electricity needed to power about 750,000 homes. That initial capacity is planned to be deployed at sites operated by Fluidstack beginning in mid-2026, with Fluidstack supplying the physical data-center infrastructure to host Anthropic’s systems.
Longer-term platform ambitions
Beyond the initial tranche, the collaboration intends to scale significantly: the platform aims to enable more than 20 gigawatts of computing capacity for leading AI labs through 2028, including support for OpenAI. Apollo is leading the first investment tranche, alongside Blackstone’s Credit & Insurance business.
Private capital stepping into constrained infrastructure market
The deal highlights the growing role of private-equity funding for AI companies that face a shortage of costly, supply-constrained infrastructure needed to meet rising demand. The article cites a precedent in which Meta secured a separate $27 billion financing arrangement with Blue Owl Capital last October to support a major data-center project, illustrating how private credit has been used to underwrite large-scale compute investments.
Broadcom’s strategic position
The arrangement is also positioned as a boost for Broadcom’s push to expand its AI business. Broadcom said the partnership is intended to scale deployment of custom AI chips and integrated computing systems while reducing the cost and power required to train AI models. The move follows Broadcom’s April agreement with Alphabet’s Google to develop and supply future generations of custom AI chips for Google’s AI racks through 2031, and a separate deal giving Anthropic access to about 3.5 GW of computing capacity using Google processors starting next year.
Context and implications
From a physical infrastructure perspective, the initiative links capital providers, chip suppliers, and data-center operators. Fluidstack will provide the on-the-ground sites beginning mid-2026, Broadcom supplies chips and networking solutions, and private-equity backers deliver the capital to build out capacity at scale. The platform’s stated intent to reach more than 20 GW by 2028 frames this as an industry-scale build aimed at multiple leading AI labs.