Table of Contents
AWS CEO Matt Garman said that transportation remains the main bottleneck for future space data centers
In sum – what to know:
Physical build limits AI scale today – Permitting, power provisioning and construction timelines, not capital, remain the main bottlenecks for hyperscale AI infrastructure.
Space data centers face major hurdles – Launch costs, payload limits and the lack of permanent space structures keep orbital facilities economically unviable.
Transport is the key constraint – The cost and availability of rockets remain the primary bottleneck for any future space-based data center concepts.
At Cisco’s AI Summit, AWS CEO Matt Garman said the biggest constraints in AI infrastructure are physical, not financial, and that space-based data centers remain far from practical deployment.
Speaking in a session hosted by Cisco president and chief product officer Jeetu Patel, Garman said the industry’s main challenge is the time and complexity involved in building data centers on Earth. Permitting, securing power and constructing facilities still follow traditional timelines.
“A big chunk of this is just it’s hard,” Garman said. “People are very excited that they can build software in a fraction of the time, but pouring concrete takes the same amount of time and building buildings takes the same amount of time. We haven’t yet built AI agents that can do that. Maybe the robots will eventually.”
Patel noted that hyperscalers face constraints beyond capital, including the time needed to secure land, power and permits. He asked whether space-based data centers could ease those pressures, given theoretical advantages such as constant solar power and natural cooling.
Garman said: “There’s a lot of compelling ideas about a data center that’s in space… Infinite amount of power that’s always available. Easy cooling. That’s great,” he said.
However, he stressed that the practical barriers remain significant. “I don’t know if you’ve seen a rack of servers recently. They’re heavy,” he said. “And the last I checked, humanity has yet to ever build a structure, a permanent structure in space, on the moon or anywhere like that.”
Garman also pointed to the limits of current launch capacity. “There are not enough rockets to launch a million satellites yet,” he said, referring to plans discussed by space industry players. He added that the cost of sending payloads into orbit remains prohibitive, describing it as “massive” and “not economical today.”
Transportation is the central bottleneck, he said. “That is the bottleneck today, is the cost and availability of just getting things into space.”
While he said improvements in launch economics could change the equation over time, Garman emphasized that the industry is still “a ways off” from making space-based data centers viable.
China is planning to deploy artificial intelligence data centers in space over the next five years, according to recent Chinese media reports, signaling Beijing’s intent to move energy-intensive computing workloads into orbit as part of its long-term technology and space strategy.
China Aerospace Science and Technology Corporation (CASC), the country’s main space contractor, has committed to building what it describes as “gigawatt-class space digital-intelligence infrastructure,” according to a five-year development plan cited by state broadcaster CCTV. The initiative aims to create space-based computing platforms capable of processing data directly in orbit rather than transmitting it back to Earth.
According to the report, the proposed space data centers would integrate cloud, edge and terminal capabilities, combining computing power, storage and transmission bandwidth in orbit.
“When it comes to AI data centers in space, there are varying stages of development and ambition. In terms of feasibility, space is a promising approach to address the upcoming energy crunch that threatens to hinder AI development. Space is very cold, which simplifies the notorious cooling issues; the unlimited solar power available brings down energy costs; and the decreased rocket launch costs achieved in recent years expand space accessibility. However, operating in space brings its own challenges, including extreme temperature swings, radiation, and space debris,” Martina Raveni, senior analyst at research firm GlobalData, told RCR Wireless News.
