Musk and Bezos ignite space race as orbital data centers emerge to solve AI's energy crisis

SpaceX and Blue Origin Enter the Orbital Data Centers Arena

The space race between tech billionaires has found a new battlefield. SpaceX is preparing a 2026 initial public offering that could raise over $30 billion toward a valuation of approximately $1.5 trillion, with significant capital earmarked for developing AI in space 3

3

. Meanwhile, Blue Origin has been developing technology for orbital data centers for more than a year, according to reports 5

5

. This Elon Musk vs Jeff Bezos competition reflects a broader industry shift as terrestrial data center limitations become increasingly apparent.

Elon Musk has expressed strong interest in leveraging Starlink for space-based data centers, stating that "simply scaling up Starlink V3 satellites, which have high speed laser links would work" 3

3

. During a recent Baron Capital event, Musk outlined a path to deploying 100 gigawatts of compute capacity from solar-powered AI satellites into orbit annually. Each Starlink V3 satellite is expected to deliver downlink speeds of 1 terabyte per second, a tenfold increase over current V2 mini satellites and comparable to terrestrial facilities 3

3

. SpaceX plans to begin launching these advanced satellites aboard Starship in 2026.

Jeff Bezos has predicted that gigawatt-scale orbital data centers will materialize within 10 to 20 years, arguing that constant solar energy available in space will eventually make off-planet AI computing infrastructure cheaper than Earth-based alternatives 5

5

. Speaking in October, Bezos emphasized that space offers "24/7" solar power without weather interruptions, making it ideal for large AI training in orbit clusters.

Starcloud Achieves Milestone with First AI Model in Space

While giants like SpaceX and Blue Origin plan their moves, smaller companies are already demonstrating viability. Nvidia-backed Starcloud trained the first AI model in space, successfully operating Google's Gemma on its Starcloud-1 satellite using Nvidia H100 chips 2

2

. The company also trained NanoGPT, an LLM created by OpenAI founding member Andrej Karpathy, on the complete works of Shakespeare, enabling the model to respond in Shakespearean English.

Starcloud CEO Philip Johnston told CNBC that orbital data centers will have 10 times lower energy costs than terrestrial facilities. "Anything you can do in a terrestrial data center, I'm expecting to be able to be done in space," Johnston said, adding that the motivation stems purely from energy constraints facing Earth-based operations 2

2

. The company plans to build a 5-gigawatt facility with solar and cooling panels measuring roughly 4 kilometers in both width and height, producing more power than the largest power plant in the U.S.

In a statement to CNBC, Google DeepMind product director Tris Warkentin noted that "seeing Gemma run in the harsh environment of space is a testament to the flexibility and robustness of open models" 2

2

. Starcloud's next satellite launch in October 2026 will feature Nvidia's Blackwell platform and a cloud infrastructure module from Crusoe, allowing customers to deploy AI workloads directly from space.

Aetherflux Targets 2027 Launch as Industry Momentum Builds

Aetherflux, founded by Robinhood co-founder Baiju Bhatt, plans to launch its first data center satellite in the first quarter of 2027 1

1

. The company views satellites as a time-saving alternative to terrestrial construction, which can take five or more years. "The race for artificial general intelligence is fundamentally a race for compute capacity, and by extension, energy," Bhatt stated, noting that current energy plans won't deliver fast enough 1

1

.

The economics are becoming increasingly favorable. Launch costs on SpaceX's Falcon Heavy currently run about $1,400 per kilogram. According to Google's calculations, if launch costs drop to around $200 per kilogram by 2030 as projected, the expense of establishing and operating space-based data centers would match ground-based operations 1

1

. Aetherflux anticipates sending approximately 30 satellites at a time on a Falcon 9, or potentially 100 or more if Starship becomes available.

The company, which secured $60 million in funding in 2024, initially focused on beaming energy from space to Earth via infrared laser. Its space data center plans represent an expansion of that vision. Aetherflux aims to provide "multi-gigabit level bandwidth with nearly constant uptime" for AI inference and general-purpose compute, though pricing remains undisclosed 1

1

.

Environmental Strain Drives Innovation Beyond Earth

The push toward orbital data centers stems from mounting environmental strain of data centers on Earth. Electricity consumption by data centers is projected to more than double by 2030, according to the International Energy Agency 2

2

. Current facilities strain power grids, consume billions of gallons of water annually for cooling, and produce substantial greenhouse gas emissions. As energy demand for AI training and operation surges, these limitations threaten to gate the industry's growth.

Proponents argue that space-based data centers can tap limitless solar power while avoiding public health and environmental consequences. Critics counter that the technological complexity and expense of operating in space's harsh environment make the concept impractical. Yet the number of companies pursuing this vision suggests confidence is building. Beyond SpaceX, Blue Origin, Starcloud, and Aetherflux, Google announced Project Suncatcher in November, while companies like Lonestar Data Holdings, Axiom Space, and Orbits Edge are developing related technologies 3

3

.

Starcloud's satellites are designed for a five-year lifespan based on the expected lifetime of Nvidia GPUs under orbital conditions 2

2

. Aetherflux plans to continuously launch new hardware and integrate the latest architectures, with older systems running lower priority tasks throughout their useful lifetime 1

1

. Real-world applications are already emerging, including real-time intelligence capabilities that could spot wildfire thermal signatures the moment they ignite and immediately alert first responders.