The Information reveals Anthropic as the customer behind Rum Group's $13.7B GPU services contract
On September 14 The Information identified Anthropic as the previously unnamed customer in Rum Group's six-year, $13.7B GPU services deal, anchored by the Maysville, Georgia data center, with Anthropic also granted penny-exercise warrants.
On September 14, 2026, The Information identified Anthropic as the previously unnamed customer behind a six-year, $13.7 billion GPU services agreement signed with Rum Group. Compute capacity will come from Rum Group's data center under construction in Maysville, Georgia, with 120 megawatts of grid power and expansion headroom to 180 MW, scheduled to come online in early 2027. Rum Group first disclosed the agreement in an August 24 SEC filing without naming the customer, describing a six-year GPU access and services contract structured across three tranches with a combined value of about $13.7 billion, the third contingent on customer approval. The deal includes an equity sweetener: Rum Group issued Anthropic warrants for up to 50.81 million Class A shares at an exercise price of $0.01 per share, half vesting on completion of the three tranches at roughly 16.67 percent per tranche and the remaining half tied to any additional capacity commitments. Rum Group acknowledged in its August filing that it still needs to raise significant debt or equity to fund the buildout, which is expected to cost more than $10 billion for construction and GPU purchases, and warned of potential delivery discounts or contract liabilities if financing falls short. Best known previously for its video platform Rumble, Rum Group entered AI infrastructure last June through an all-stock acquisition of Northern Data valued at about $767 million, bringing roughly 22,000 Nvidia Hopper GPUs into the portfolio and rebranding its compute arm as Quake AI. Anthropic had earlier committed to giving embedded third-party evaluators employee-like access in response to a coordinated safety push with OpenAI, xAI, and SpaceX AI, with the lock-in of large-scale compute seen as the next leg of support for its frontier-model training and product expansion.