Record $6.3 Billion AI Compute Agreement
SpaceX has signed a landmark $6.3 billion compute deal with Reflection AI, one of the largest such agreements in the AI industry, to provide access to Nvidia's latest GB300 AI chips at the Colossus 2 data centre near Memphis, Tennessee. The deal, structured as a monthly commitment of $150 million starting July 1, 2026 through 2029, further cements SpaceX's position as a dominant force in the AI infrastructure market.
This agreement brings SpaceX's total committed compute revenue to over $80 billion since its record-breaking IPO earlier this year, demonstrating the extraordinary demand for AI computing power from leading technology companies and AI labs.
Understanding the Colossus 2 Infrastructure
Colossus 2 is SpaceX's latest and most powerful AI data centre, located in Memphis, Tennessee. The facility is designed to house tens of thousands of Nvidia GB300 AI chips, which represent the cutting edge of AI acceleration hardware. The GB300, Nvidia's next-generation GPU architecture, offers significantly improved performance for both AI training and inference workloads compared to its predecessors.
The Colossus campus represents SpaceX's ambitious pivot from aerospace to AI infrastructure, leveraging the company's expertise in managing complex, high-power systems and its access to capital markets following its IPO. The facility is designed with advanced liquid cooling systems and dedicated power infrastructure to handle the enormous energy demands of AI supercomputing.
Reflection AI: A Major AI Player Emerges
Reflection AI has emerged as one of the fastest-growing AI companies, and this compute deal signals its ambitious plans for model training and deployment. The $150 million monthly commitment is among the largest single compute contracts in the industry, suggesting that Reflection AI is preparing to train models at a scale that rivals or exceeds the largest known AI training runs.
Analysts estimate that the compute capacity secured through this deal could support training of frontier-scale AI models requiring tens of thousands of GPUs over extended periods. Reflection AI has not disclosed its model architecture or training plans, but the scale of this commitment indicates confidence in a substantial return on compute investment.
SpaceX's AI Infrastructure Business Post-IPO
Since its record-breaking IPO, SpaceX has aggressively expanded its AI infrastructure business. The company has now signed over $80 billion in committed compute revenue, with customers including major AI labs, cloud providers, and enterprise customers. SpaceX's entry into the AI compute market has been one of the most significant developments in the industry, challenging established players like AWS, Azure, and Google Cloud.
The company's competitive advantages include access to competitive power pricing in Tennessee, experience building and managing large-scale infrastructure projects, and the ability to secure favourable terms for Nvidia's latest hardware. SpaceX has also been exploring dedicated AI chip development, though no formal announcement has been made.
Impact on India's AI Ecosystem
While the Colossus 2 facility is located in the United States, the deal has implications for India's growing AI ecosystem. Indian AI startups and enterprises that rely on cloud compute services may face increased competition for GPU access as major players like SpaceX lock in long-term capacity commitments. However, the growing AI infrastructure market is also driving investment in data centre capacity in India, with companies like RMZ and Reliance expanding their data centre operations.
For Indian AI researchers and developers, the long-term trend of expanding AI compute capacity globally should eventually translate into more accessible and affordable cloud AI services. Indian companies building AI applications will need to plan their compute strategies carefully in this constrained market.
Sources
Sources: AIToolsRecap, Reuters, Teslarati, Bloomberg, SpaceX Investor Relations
