Anthropic will pay xAI $1.25 billion per month for compute | TechCrunch
Briefly

Anthropic will pay xAI $1.25 billion per month for compute | TechCrunch
Anthropic agreed to purchase 300 megawatts of compute from xAI, securing the full output of the Colossus 1 data center near Memphis, Tennessee. The payment is $1.25 billion per month through May 2029, with a discounted rate for the first two months while xAI ramps up. The arrangement could generate more than $40 billion in revenue for xAI. Terms allow either party to terminate with 90 days’ notice. The deal emerged from SpaceX’s SEC filing, which described monetizing unused compute capacity and indicated additional similar services contracts. The structure creates a hybrid approach where xAI both builds capacity and sells it when internal demand drops, offsetting infrastructure costs. Usage of Grok has declined, freeing servers that are now sold to a competitor.
"Anthropic will be paying xAI $1.25 billion per month through May 2029, with a discounted rate for the first two months as xAI completes its ramp-up. All told, the deal could bring xAI over $40 billion in revenue."
"The deal, the company said, "allows us to monetize unused compute capacity in our infrastructure." The terms of the deal allow either side to terminate the contract with 90 days' notice. "We expect to enter into additional similar services contracts," the filing stated."
"SpaceX argues the arrangement is a savvy use of resources. "We believe our dual monetization strategy provides multiple pathways to generate returns on invested capital," it wrote. But the subtext is hard to miss. xAI appears to have overbuilt its compute capacity and needed to find a way to monetize it ahead of a public offering."
"This emerging model, sometimes called a "neocloud," lets AI companies offset infrastructure costs by acting as a cloud provider when their own usage falls short of capacity. Most players either build data centers for themselves or build data centers for others to use - rarely both simultaneously."
Read at TechCrunch
Unable to calculate read time
[
|
]