Image Courtesy: Getty Images
Anthropic is set to pay rival AI company xAI roughly $1.25 billion per month for access to massive computing infrastructure, in one of the largest known AI compute agreements to date.
The deal gives Anthropic access to the full 300 megawatts of output from xAI’s Colossus 1 data center near Memphis, Tennessee, a facility designed to power large-scale AI training and inference workloads. The agreement runs through May 2029, though xAI will offer a discounted rate during the first two months while the system reaches full operational capacity, according to TechCrunch.
Financial details surfaced through SpaceX’s S-1 filing with the SEC, which described the arrangement as a way to monetize unused computing infrastructure. If maintained for the full term, the agreement could generate more than $40 billion in revenue for xAI.
The deal highlights how compute power has become one of the most valuable commodities in the AI industry. As companies race to train increasingly advanced models, access to GPUs, energy infrastructure, and data center capacity is becoming just as strategically important as the models themselves.
It also signals a growing shift toward the so-called “neocloud” model, where AI firms operate both as model developers and infrastructure providers. Traditionally, companies either built AI systems for internal use or rented infrastructure to external customers. xAI now appears to be doing both simultaneously, using excess capacity to generate revenue while continuing to develop its own products.
The arrangement may also reveal broader market dynamics behind the scenes. xAI reportedly expanded its infrastructure aggressively to support products like Grok, its flagship AI assistant, but demand appears to have softened in recent months. Leasing spare compute capacity to Anthropic allows the company to offset the enormous costs associated with maintaining hyperscale AI infrastructure ahead of a potential public offering.
For Anthropic, the agreement provides a direct path to securing long-term compute resources at a time when demand for high-end AI hardware continues to outpace global supply. AI companies are increasingly competing not just for talent and users, but for electricity, server capacity, and access to advanced chips capable of training frontier models.