Anthropic signs $13.7B AI compute deal with RUM Group
What's the story
Anthropic has signed a six-year compute agreement worth $13.7 billion with RUM Group, as reported by The Information. The deal is one of Anthropic's biggest compute commitments to date and strengthens its relationship with a company that has strong ties to US President Donald Trump's political circle.
Deal details
Details of the deal
As part of the agreement, Anthropic will leverage GPU-based computing power from RUM Group's data center in Maysville, Georgia.
The deal was revealed in an SEC filing by RUM Group last month, which described it as a $13.7 billion GPU services agreement over six years without revealing the customer's name at that time.
Warrant details
Warrant issued by RUM Group
As part of the deal, RUM Group issued a 10-year warrant to the customer for as many as 50.8 million shares at $0.01 each.
Half of the warrant is tied to GPU purchases under the initial Maysville contract, while the other half will only be triggered if the customer agrees to an additional 400-500 megawatts of capacity at another RUM Group facility currently under construction.
Company evolution
What is RUM Group?
Founded in 2013 as Rumble, RUM Group started as a video-sharing platform similar to YouTube. It gained popularity among conservative users after the 2020 US presidential election.
The company has hosted Trump's Truth Social and White House livestreams, and was backed by early investors Peter Thiel and JD Vance (now US vice-president).
Earlier this year, it acquired Northern Data in a bid to expand into AI infrastructure.
Strategy expansion
Anthropic's compute footprint expands with the RUM Group deal
The deal with RUM Group expands Anthropic's compute footprint beyond its existing partners, which include Google, Amazon, SpaceX, and Nscale.
It reportedly takes the company's total secured computing capacity to at least 14.8 gigawatts.
Despite a sharp increase in its annualized revenue run rate through 2025 and into 2026, Anthropic has pursued a diversified multi-cloud strategy to secure infrastructure for training and running its Claude models.