$35 billion. That’s the check Apollo and Blackstone are writing to build out Anthropic’s AI compute capacity, and it buys a lot of megawatts. The first phase alone adds one gigawatt of capacity—enough to power a mid-sized city, not just a datacenter.
1 GW Live by Mid-2026
Fluidstack-operates the sites where this hardware lands, and the initial deploy runs from mid-2026. One gigawatt is a statement: that’s roughly the electrical draw of 750,000 U.S. homes. And it’s only the appetizer.
Broadcom’s Custom Chips Underpin the Buildout
This isn’t a generic cloud rental. The expansion uses Broadcom’s custom silicon and networking gear, purpose-built for AI training and inference. That means Anthropic gets tighter integration between compute and memory, not just more GPUs bolted onto a rack. Apollo and Blackstone are betting on hardware specificity over commodity density.
Why This Scale Matters
The partnership targets over 20 GW for leading AI labs by 2028. For perspective, the entire state of Virginia, home to the world’s largest datacenter corridor, currently draws about 2.5 GW for its data centers. Reaching 20 GW means building the equivalent of eight Northern Virginia data-center clusters from scratch—per lab.
Apollo and Blackstone have the balance sheets to do it. The question is whether Fluidstack can get the sites online while the rest of the industry fights over every megawatt of interconnection capacity. Mid-2026 delivery on a greenfield gigawatt buildout is aggressive; that timeline will test the supply chain as hard as the hardware.
Source: Apollo, Blackstone back Anthropic's $35 billion capacity expansion in new Broadcom tie-up
Domain: economictimes.indiatimes.com
Comments load interactively on the live page.