Project Suncatcher (Photo: Google)

Google is joining forces with Blue Origin, founded by Jeff Bezos (제프 베조스), to begin experiments to build artificial intelligence (AI) data centres in space. The project aims to build AI computing infrastructure in orbit using solar power, drawing attention over whether it can reduce the power, land and cooling burdens of ground-based data centres.

On Nov. 19 (local time), IT outlet Engadget and others reported that Google plans to launch AI server satellites using Blue Origin’s heavy-lift New Glenn rocket as part of Project Suncatcher. The first dedicated launch is scheduled for the first quarter of 2029. Each satellite is to have computing capability equivalent to a single data centre rack and carry Google’s own AI chips, Tensor Processing Units (TPUs), and servers. The satellites are to be deployed in a sun-synchronous orbit where they can use solar power continuously.

Google described the initiative as a long-term research project to expand AI infrastructure in space using solar energy. It aims to test whether space can be an alternative as ground-based data centres require enormous power, land and cooling facilities.

Commercialisation still faces significant technical hurdles. Google succeeded last October in launching its first AI server satellite into low Earth orbit, operating its own TPU and transmitting to Earth a haiku generated by its onboard AI model, Gemma. But the satellite is experimental equipment that can operate only 15 minutes at a time.

In particular, releasing heat in space is difficult in the same way as on Earth. In a vacuum, cooling through convection is impossible, requiring heat to be discharged externally using pipes and panels. Radiation-driven “bit flip” phenomena are also a problem. Radiation can swap 0s and 1s in semiconductor binary data, causing computing errors.

Launch costs are also a key variable that will determine commercial viability. SpaceX’s Falcon 9 launch costs about $74 million per launch, and its payload capacity is also limited for transporting the equipment needed to build large-scale data centres. Building a large satellite network, beyond lofting a single satellite, would require huge amounts of money and time. Brandon Lucia (브랜던 루시아), a professor of electrical and computer engineering at Carnegie Mellon University, pointed out in The New York Times that building a large satellite data centre would take years and huge funding, and additional engineering problems would also need to be resolved.

Another challenge Google must solve over the long term is securing launch vehicles. The company expects launch costs to fall to about $200 per kg in the mid-2030s, but achieving that would require ultra-large reusable launchers. SpaceX’s Starship is cited as a leading candidate, but SpaceX is also moving to build its own space data centres, making it uncertain whether Google can secure the payload capacity and price it wants. The contract with Blue Origin is for a single launch for now, but the possibility of expanded cooperation remains. Still, Blue Origin is also pursuing its own space data centre project, Project Sunrise, which could lead to a competitive relationship over the long term.

Competition over space data centres is expanding to include SpaceX and Starcloud. Whether other launch companies such as Rocket Lab, Relativity Space and Stoke Space succeed in developing large rockets is another variable. Ultra-large launch vehicle development requires enormous capital and time, so it could take a long time before Google secures a stable launch option.

Space environment issues stemming from an increase in satellites must also be considered. The scientific community has warned of the “Kessler syndrome,” in which higher satellite density in low Earth orbit leads to cascading collisions and space debris. The number of objects in low Earth orbit is presented as having increased from fewer than 2,000 in the past to more than 47,000 now. With proposals also emerging to deploy millions more satellites, managing space debris and collision risks is becoming more important.

Ultimately, Google’s space data centre initiative is a long-term project that must address cooling technology, radiation 대응, launch costs, satellite operations and space environment issues, despite the advantage of abundant solar power. The first dedicated launch in 2029 is expected to be a technical test bed to verify whether AI computing infrastructure can be built in space.

Keyword

#Google #Blue Origin #Project Suncatcher #New Glenn #SpaceX
Copyright © DigitalToday. All rights reserved. Unauthorized reproduction and redistribution are prohibited.