344
Productivity & Workflow355
Automation & Workflow224
Software Development250
Marketing & Growth192
AI Infrastructure & MLOps174
Writing & Content Creation203
Data & Analytics141
Design & Creative169
Customer Support131
Photography & Imaging156
Sales & Outreach125
Voice & Speech135
Education & Learning131
Operations & Admin87
Reflection AI says it will pay SpaceX $150 million a month from July 2026 through 2029 to use Nvidia GB300 chips at a Tennessee data center.
In short: Reflection AI will pay SpaceX $150 million per month from July 1, 2026 through 2029 to use Nvidia’s latest AI chips in SpaceX’s data center near Memphis.
Reflection AI, an open-source focused AI startup, has signed a deal with SpaceX for access to computing power, meaning time on specialized computer chips used to train and run AI systems.
Reflection AI told TechCrunch it will pay $150 million a month starting July 1, 2026, and the contract runs through 2029. The agreement is worth up to $6.3 billion. After the first three months, either company can end the deal with 90 days’ notice.
The access is to Nvidia GB300 chips and related hardware inside SpaceX’s Colossus 2 data center near Memphis, Tennessee. You can think of these chips like high-end engines that make AI training go faster, and a data center like a warehouse full of computers.
SpaceX has made similar, larger compute rental deals recently. TechCrunch reports Anthropic agreed to pay $1.25 billion per month and Google agreed to pay $920 million per month, with both deals also running through July 2029.
Reflection AI was founded in 2024 by two former Google DeepMind researchers. The company says it is focused on “open-weight” models, meaning it publicly releases the model’s trained settings (like publishing the recipe, not just the finished dish).
AI companies increasingly need huge amounts of computing power, and the biggest bottleneck is often getting enough advanced chips. Deals like this show that large data centers and chip supplies are becoming something companies rent, not just something they build themselves.
Source: TechCrunch AI