logo-darklogo-darklogo-darklogo-dark
  • Tool Categories
    • 🎨Art & Creative Design505
    • 🏢Business Management644
    • 💻Coding & Development514
    • 👮Detection83
    • 🧠General Use728
    • 🏥Health & Wellness55
    • 📷Image & Photo Analysis100
    • 🖼️Image Generation & Editing618
    • 📐Interior & Architectural Design37
    • 🎓Learning & Education483
    • ⚖️Legal & Finance90
    • 🎭Lifestyle & Entertainment236
    • 📢Marketing & Advertising627
    • 🎧Music & Audio138
    • 👔Office & Workplace1,014
    • 🔬Research & Data Analysis373
    • 👥Social Media245
    • 🎥Video Generation & Editing426
    • 👧🏻Virtual Companion135
    • 🎤Voice Generation & Editing381
    • ✍️Writing & Editing808
    • All Categories
    • AI Use Cases
  • News
  • Events
    • Academic Conferences
    • Developer Conferences
    • Expos / Trade Shows
    • Industry Summits
    • Workshops / Training
    • All Events
    • Past Events
  • Saved Tools
  • Suggest a Tool
✕
Home › News › SpaceX signs $6.3 billion compute deal with open-source AI startup Reflection

SpaceX signs $6.3 billion compute deal with open-source AI startup Reflection

June 22, 2026
Hand holding a smartphone with a white Laravel logo on screen, SpaceX branding blurred in the background.

#image_title

SpaceX has signed another major compute deal, this time with Reflection AI, an open-source AI startup founded by two former Google DeepMind researchers. As reported by TechCrunch, the deal is worth up to $6.3 billion and gives Reflection immediate access to Nvidia’s latest GB300 AI chips at SpaceX’s Colossus 2 data center near Memphis, Tennessee.

Reflection will pay $150 million per month starting July 1, 2026, through 2029. Either side can walk away with 90 days’ notice after the first three months. This is Reflection’s first major compute deal, and the company says it’s one of the largest infrastructure commitments made by an open AI lab to date.

The deal is smaller than what SpaceX has already arranged with Anthropic and Google, which pay $1.25 billion and $920 million per month respectively for access to the same data center. Those contracts also run through July 2029. Elon Musk has downplayed the three-year length of these agreements in public, pointing to the cancellation clause as evidence they’re not as binding as they appear.

Reflection’s pitch is simple: it wants to be the open-source alternative to labs like Anthropic and OpenAI. Unlike those companies, Reflection publicly releases its model weights, meaning anyone can inspect, download, and build on top of its AI systems. That approach has gained more attention lately after the U.S. government banned Anthropic’s closed models, Fable and Mythos, pushing more developers and governments toward open alternatives.

In a statement, a Reflection spokesperson said: “Recent events highlight how important open source is to the AI ecosystem, with more nations and enterprises recognizing the risks and costs associated with exclusively depending on closed models. Our deal with SpaceXAI signals Reflection’s strategic importance within the frontier AI ecosystem, and more compute means more runway to build the world’s best open models at scale.”

The Colossus data center has an interesting backstory. It was originally built by xAI, Musk’s AI company, which is now folded into SpaceX. As xAI’s own AI efforts have lost momentum, SpaceX has turned the facility into a revenue source by renting out its chip capacity to outside labs. Anthropic and Google were the first major tenants. Reflection is now the third.

This pattern matters beyond the deal itself. SpaceX is quietly becoming a major player in AI infrastructure, not by building AI products, but by controlling access to the hardware that makes those products possible. With compute still scarce and Nvidia’s latest chips in high demand, whoever holds the chips holds real power in the AI market. That gives SpaceX a strong position regardless of how its own AI ambitions play out.

For Reflection, the access to GB300 chips is a significant step up. Training frontier AI models requires enormous amounts of compute, and startups typically struggle to get access to the latest hardware at scale. Locking in three years of capacity, even with an exit clause, gives the company the stability it needs to compete with much larger, better-funded labs.

Share

Related news

Wisp Flow branding logo in orange on a dark teal background. (Logo shows the text 'Wisp Flow' with an icon on the left)

#image_title

August 3, 2026

Wispr Flow is moving into meeting notes, and the timing makes sense


Read more
Collage showing Gemini branding with Seattle skyline, a dark 'Thinking it through' task list, a white airport status card, a floating 'Take over task' bubble, and a green 'Task done' banner.

#image_title

August 3, 2026

Gemini Spark can now browse Chrome on your behalf


Read more
Person typing on a laptop with an AI chat interface visible on screen.

#image_title

August 3, 2026

EU AI Act transparency rules are now live — here’s what they actually require


Read more

Recent Posts

  • Wispr Flow is moving into meeting notes, and the timing makes sense
  • Gemini Spark can now browse Chrome on your behalf
  • EU AI Act transparency rules are now live — here’s what they actually require
  • Alibaba’s Qwen3.8-Max is its biggest model yet, and it’s open source
  • June wants AI to fix AI deployment, and Marc Benioff just bet $20M on it
Best AI Tools

Discover the best AI tools for any use case

Explore
  • Tool Categories
  • AI Use Cases
  • AI Events
  • AI News
  • Saved Tools
Company
  • About Us
  • Contact Us
  • Media & Partnerships
  • Suggest a Tool
Legal
  • Privacy Policy
  • Terms of Service
Copyright © 2026 Best AI Tools 415 Mission Street, 37th Floor, San Francisco, CA 94105