In a groundbreaking leap for computing, Nvidia is partnering with Redmond-based startup Starcloud to send an H100 GPU into orbit. Scheduled for November 2025, SpaceX will launch this first-of-its-kind orbital data center via its Falcon 9 rocket, marking a historic milestone where a data center soars above antennas and cables.
The Starcloud Project: A Refrigerator-Sized Supercomputer in Space
.jpg)
Last week, Nvidia unveiled its collaboration with Starcloud, a startup set to launch a 60-kilogram satellite - roughly the size of a refrigerator - equipped with an Nvidia H100 GPU. This compact powerhouse will process data directly in orbit, boasting computational capabilities 100 times greater than any previous orbital system.
The concept draws inspiration from terrestrial analogs like China’s solar farms on the Tibetan Plateau, where high altitudes offer abundant sunlight and cooler temperatures for efficient operations.
In space, sunlight is limitless, and the ambient temperature hovers just a few degrees above absolute zero, creating an ideal environment for high-performance computing.
Starcloud’s CEO, Philip Johnston, claims that energy costs in orbit could be ten times cheaper than on Earth, even factoring in launch expenses. However, this ambitious venture comes with its share of challenges.
Orbital Data Centers: Latency? Not a Problem
At first glance, one might assume that orbital data centers would struggle with latency. However, Johnston dismisses these concerns:
“From a latency perspective, operating on low Earth orbit is no different from a terrestrial data center. At 400 kilometers altitude, it’s not much different from a data center in Alaska or Seattle.”
Starcloud’s long-term vision is to support the same workloads as ground-based data centers. “With high bandwidth and low latency, we expect to handle real-time applications, large-scale inference, and training on orbital GPU clusters just as we do on Earth,” Johnston explains.
For now, Starcloud is focusing on space-native use cases, such as processing Earth observation data. Synthetic Aperture Radar (SAR) imaging generates up to 10 gigabytes of data per second. Transmitting this raw data to Earth consumes significant bandwidth and time.
By running inference in orbit - building 3D terrain maps, identifying crop types, or tracking environmental changes - Starcloud’s satellite can send only the processed results back to Earth, streamlining operations.
In essence, satellites will process data where it’s collected, reducing dependency on Earth-based infrastructure.
SpaceX + Nvidia: A Cosmic Partnership
Starcloud-1 will hitch a ride on SpaceX’s Falcon 9 as part of the Bandwagon 4 rideshare mission, designed for small satellites. The cost? A modest $300,000 for a 50-kilogram slot, with Starcloud’s slightly heavier 60-kilogram payload still being a bargain in space terms.
The relationship between SpaceX’s Elon Musk and Nvidia’s CEO Jensen Huang adds an intriguing layer to this project. In summer 2024, Musk and Oracle co-founder Larry Ellison reportedly pleaded with Huang for more GPUs during a dinner at Nobu. “Take our money. Please, take more,” Ellison quipped, recounting the meeting.
This camaraderie extends beyond business. Huang personally delivered a DGX Spark system to Musk at SpaceX’s Starbase in Texas, echoing a 2016 moment when he handed the first DGX-1 to OpenAI. Nvidia’s investment in Musk’s xAI further cements this alliance, with Huang later expressing regret for not investing more.
Why Nvidia Is Betting on Space
Starcloud is part of Nvidia’s Inception program, a free initiative that supports startups without taking equity or fees. While Nvidia doesn’t own a stake in Starcloud, the partnership aligns with its broader strategy.
By powering the “first H100 in space,” Nvidia positions itself at the forefront of humanity’s most ambitious projects - a branding win that transcends mere financial support. This move also creates a lock-in effect: Starcloud’s reliance on Nvidia’s hardware could make switching to competitors like AMD a daunting prospect in the future.
Moreover, this venture serves as market reconnaissance. If orbital data centers take off - both literally and figuratively - Nvidia will have a first-mover advantage, deeply understanding the challenges of space-based computing, from radiation to zero gravity and extreme cold. Proving that its chips can withstand these conditions underscores Nvidia’s reputation for unmatched reliability.
Also read:
- Something Strange Is Happening with the Series 'The Savant'—It Looks Like Apple TV+ Isn't Planning to Release It Anytime Soon
- x402 Transaction Volume Surges 10,000% in a Single Month
- Is the AI Sector a Bubble? How API Fakes Are Killing the Industry
A New Ecosystem in Orbit
When Starcloud launches its satellite via SpaceX using Nvidia’s chips, it’s more than a transaction—it’s the birth of an ecosystem.
This interconnected network of innovators, from Musk’s SpaceX to Huang’s Nvidia, is laying the foundation for a future where data centers in space are as commonplace as those on Earth.
With giant radiators replacing air conditioners, the cosmos is becoming the next frontier for high-performance computing.
