Skip to main content

SpaceX inks compute deal with Reflection AI, an open source AI lab

Reflection AI will lock in a $150 million‑per‑month lease for SpaceX’s Colossus 2 data centre, gaining immediate access to Nvidia’s cutting‑edge GB300 AI chips from July 1 2026 through 2029. The deal not only fuels Reflection AI’s open‑source model training but also turns SpaceX’s aerospace‑grade compute assets into a commercial AI‑cloud offering.

Published

22 Jun 2026

Reading Time

2 min read

Share this article:

Contents

Deal Overview

Reflection AI will pay $150 million a month beginning July 1 2026 through 2029 for immediate access to Nvidia's latest GB300 AI chips and supporting hardware across SpaceX's Colossus 2 data center near Memphis, Tennessee.

  • Monthly spend: $150 M

  • Contract period: July 1 2026 – 2029

  • Location: SpaceX Colossus 2, Memphis, TN

  • Hardware: Nvidia GB300 AI chips + supporting gear

Why It Matters

  • The size of the monthly payment indicates a substantial allocation of compute resources, likely sufficient for training large‑scale AI models.

  • Immediate access to Nvidia’s latest GB300 chips gives Reflection AI a high‑performance foundation in the competitive open‑source AI landscape.

  • By leasing its Colossus 2 facility for AI workloads, SpaceX is positioning its data‑center assets as a commercial compute platform for external developers and startups.

Who Is Affected

  • Reflection AI – an open‑source AI lab that gains direct, high‑end GPU capacity.

  • SpaceX – monetizes its Colossus 2 infrastructure beyond aerospace operations.

  • Nvidia – sees increased demand for its GB300 AI silicon through a high‑visibility partnership.

What to Watch

  • Utilization trends at Colossus 2 over the three‑year term, which could signal demand for similar AI‑compute leasing agreements.

  • Model output and releases from Reflection AI that may leverage the GB300 hardware, offering insight into the performance impact of such compute access.

  • Potential follow‑on deals from other AI startups seeking comparable access to high‑end GPU clusters.

Source: TechCrunch, 22 June 2026

13

views

0

shares

0

likes

Related Articles