Skip to content
AI Usage4 min read

SpaceX Signs $6.3 Billion Compute Deal With Reflection AI

SpaceX signs a $6.3 billion compute lease with Reflection AI to provide Nvidia GB300 capacity at its Memphis Colossus 2 facility.

AB

Author

AUG Bot

Published

Representation of a high-density AI supercomputer cluster at the Colossus 2 facility

SpaceX Signs $6.3 Billion Compute Deal With Reflection AI

New agreement secures Nvidia GB300 capacity at Memphis Colossus 2 facility

SpaceX has signed a compute lease with Reflection AI for $150 million per month starting July 1, 2026, through the end of 2029, totaling approximately $6.3 billion. This deal highlights the growing trend of GPU compute as a standalone business and the massive capital requirements for training frontier-scale AI models.

Key details

Under the agreement, Reflection AI gains immediate access to Nvidia GB300 chips at SpaceX's Colossus 2 facility in Memphis, Tennessee. The contract includes a flexible exit clause allowing either party to terminate with 90 days' notice after an initial three-month period.

The deal makes Reflection AI the fourth major external tenant at the Colossus complex, following Anthropic, Google, and Cursor. SpaceX's committed compute revenues from outside clients now exceed $80 billion through 2029. The Colossus facility, originally built for training xAI's Grok, has been transformed into one of the largest third-party compute platforms in the world.

Why this matters

This deal underscores the massive infrastructure costs associated with AI scaling. For a startup like Reflection AI, securing 42 months of guaranteed high-end GPU access is an existential necessity to compete with deep-pocketed incumbents. It also demonstrates how specialized data center infrastructure—originally built for internal use—is becoming a critical commercial resource in the AI economy.

Context

SpaceX has rapidly pivoted into an AI infrastructure provider. The Colossus complex in Memphis is planned to reach 2 gigawatts of total power capacity, housing over 555,000 Nvidia GPUs purchased at an estimated cost of $18 billion. This scale allows SpaceX to host multiple frontier AI labs, positioning it alongside traditional hyperscalers like AWS and Microsoft Azure despite having no prior cloud computing background.

What happens next

The $6.3 billion commitment suggests that Reflection AI is preparing for an imminent large-scale training run for its frontier open-weight models. Meanwhile, SpaceX is expected to continue expanding its Memphis infrastructure to meet the surging demand for high-density compute capacity, potentially generating more revenue from its data center business than its launch services over the next three years.


Source: MLQ News Published on AI Usage Global, author: AUG Bot

Related

Read more

More posts that expand on the topics, companies, and AI trends covered in this story.

Digital rendering of a massive natural gas power generation facility supplying electricity to a hyperscale data center
AI Usage

Planned Amazon Data Center Could Become the Biggest Climate Polluter in the U.S.

Amazon invests in a planned natural gas power plant for a Pecos County, Texas, data center permitted to release up to 33 million tons of carbon dioxide annually.

Digital rendering of a modern AI data center with on-site substation power infrastructure in McGregor, Texas
AI Usage

Galaxy Digital's Texas AI Data Center Reflects New Grid Funding Model

Galaxy Digital acquires 500 acres in McGregor, Texas, for a 74 MW AI campus under a new 'beneficiary proves' model requiring self-funded grid and water upgrades.

Digital rendering of a hyperscale data center powered by Bloom Energy fuel cells on-site
AI Usage

Vineland AI Data Center Eyes Bloom Energy Fuel Cells for 300 MW

Developers of a 300-megawatt AI data center in Vineland, New Jersey, propose on-site Bloom Energy fuel cells to bypass public grid capacity bottlenecks.