Anthropic has struck one of the largest infrastructure agreements yet in the artificial intelligence race, committing to pay more than $40 billion over four years for computing capacity from xAI, according to disclosures contained in SpaceX’s S-1 filing.
The agreement gives Anthropic access to 300 megawatts of compute capacity at xAI’s Colossus 1 data center near Memphis, effectively securing the facility’s entire output. Anthropic will pay approximately $1.25 billion per month through May 2029, although the first two months come at a discounted rate while xAI completes deployment ramp-ups.
The transaction indicates that the economics of artificial intelligence are rapidly shifting from purely model development toward control of power-intensive infrastructure. In the AI industry, access to electricity, advanced chips, and large-scale data centers has increasingly become as strategically important as the models themselves.
Register for Tekedia Mini-MBA edition 20 (June 8 – Sept 5, 2026).
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register for Tekedia AI Lab.
For xAI, the agreement offers a massive new revenue stream at a critical moment. Elon Musk’s AI company has spent aggressively building GPU clusters to compete with rivals, including Anthropic, OpenAI, and Google. But maintaining unused compute infrastructure is extraordinarily expensive, particularly as companies race to deploy hundreds of thousands of Nvidia AI chips.
SpaceX described the arrangement as a way to “monetize unused compute capacity in our infrastructure,” adding in the filing that it expects to pursue “additional similar services contracts.”
The language points to an emerging hybrid strategy in the AI sector. Rather than using infrastructure solely for internal AI development, xAI is increasingly acting like a commercial cloud provider, leasing excess capacity to outside companies when utilization drops below planned levels.
That model has become known in Silicon Valley as the “neocloud” approach, where AI firms attempt to offset the staggering cost of building hyperscale computing clusters by selling spare compute to rivals, startups, or enterprises.
The agreement also reveals how quickly competitive lines are blurring in the AI race. Anthropic and xAI are direct rivals in foundation models and AI assistants, yet Anthropic is now relying on Musk’s infrastructure to expand its own capabilities.
The deal may also signal that xAI built more infrastructure than it currently needs. The company has invested heavily in Colossus, which was promoted as one of the world’s largest AI supercomputing clusters. However, recent reports of slowing usage for xAI’s Grok chatbot suggest parts of that capacity may not have been fully utilized.
xAI can convert idle assets into recurring revenue while preserving its long-term expansion plans by leasing the infrastructure to Anthropic.
The arrangement also reflects mounting financial pressure across the AI industry. Training and operating frontier AI systems now require enormous capital expenditures, forcing companies to search for alternative monetization strategies.
Infrastructure spending across the AI ecosystem has surged dramatically over the past two years. Hyperscalers and AI labs are collectively projected to spend hundreds of billions of dollars annually on data centers, networking equipment, advanced cooling systems, and power procurement.
At the center of the spending boom is Nvidia, whose AI chips remain the industry standard for training large language models. But acquiring GPUs alone is no longer sufficient. Companies now need access to reliable electricity grids, land, fiber connectivity, and advanced cooling systems capable of handling increasingly dense AI clusters.
The scale of the Anthropic-xAI agreement highlights how compute itself is becoming a tradable commodity in the AI economy. The structure also gives both sides flexibility. According to the filing, either company can terminate the arrangement with 90 days’ notice. That clause could prove important in an industry where technological shifts, regulatory changes, and competitive dynamics move rapidly.
For Anthropic, securing long-duration compute access helps reduce dependence on traditional cloud providers while guaranteeing access to scarce infrastructure as demand for AI accelerates. The company has been aggressively expanding capacity in recent months amid rising adoption of its Claude models across enterprise and developer markets.
The deal is expected to strengthen investor confidence ahead of any future public market ambitions. Generating stable, long-term infrastructure revenue may help offset concerns about the profitability and adoption trajectory of Grok and other consumer-facing AI products.