Google will launch a satellite carrying its Tensor Processing Units aboard a SpaceX Falcon 9 on Oct. 1, testing whether AI computing hardware can survive radiation, extreme temperatures and the difficult cooling conditions of low Earth orbit.
By yourNEWS Media Newsroom.
Google is preparing to move part of its artificial intelligence infrastructure off the planet, with an Oct. 1 satellite launch that will test whether the company’s AI chips can operate under the demanding conditions of space.
The experimental flight is part of Project Suncatcher, Google’s long-term effort to determine whether large-scale AI computing could eventually be performed by networks of solar-powered satellites in orbit, according to The New York Times.
The satellite will carry Google’s Tensor Processing Units, or TPUs, aboard a SpaceX Falcon 9 rocket and operate in low Earth orbit using solar energy.
Unlike a conventional terrestrial data center, the mission is not intended to perform large-scale AI computing. Instead, Google wants to learn whether the underlying hardware can survive launch and continue operating after being exposed to conditions that cannot be fully reproduced on Earth.
The company said the mission will examine whether its chips can “handle the physical stress of spaceflight and the radiation and thermal extremes of space.”
“Some things can only be tested in space,” Google said.
Ground testing has already provided encouraging results in two areas.
Google subjected the hardware to substantial gravitational forces designed to simulate the stresses of a rocket launch and also exposed TPUs to radiation to determine whether they could continue processing AI workloads.
The chips performed well enough in those experiments for Google to proceed with orbital testing.
Managing heat may prove more difficult.
Data-center processors generate substantial amounts of heat on Earth, where buildings can rely on air circulation, liquid cooling and other systems to move that heat away from computing equipment.
Space presents a fundamentally different engineering problem because there is no surrounding atmosphere to carry heat away.
Project Suncatcher is experimenting with heat pipes and radiators designed to transfer thermal energy away from the chips and release it into space.
The Oct. 1 mission will allow engineers to test that cooling system under actual orbital conditions.
For the initial satellite, the limitation is significant.
Travis Beals, Project Suncatcher’s senior director of product management, told the Times that “the chips can only run for about 15 minutes before they’ll need to be turned off to cool down.”
That short operating period illustrates the distance between the upcoming experiment and Google’s eventual vision of an orbital AI data center capable of continuously handling major computing workloads.
Google is treating the limitation as an engineering problem to investigate rather than evidence that the concept cannot work.
“Exploring space as a viable location for scalable AI compute won’t happen all at once,” Beals wrote. “It takes methodical engineering, starting with proving our hardware can handle the physical and unpredictable realities of operating in orbit. This first launch is about seeing what works, identifying points of failure, and applying those findings to future missions.”
The company’s longer-term concept involves deploying groups of satellites carrying AI processors rather than relying on a single spacecraft.
Such satellites could theoretically take advantage of extensive solar energy while moving some of the massive electricity and land requirements associated with AI infrastructure away from communities on Earth.
Google has already planned additional orbital testing for next year as it studies how multiple spacecraft could eventually work together.
Connecting those satellites would present another technical challenge. Large AI systems depend on extremely fast communications between processors, meaning future orbital computing networks would need to transfer enormous volumes of data between satellites with minimal delay.
The Project Suncatcher effort comes as demand for AI computing power drives rapid construction of terrestrial data centers.
Those facilities require large supplies of electricity, significant physical infrastructure and, depending on their cooling technology, substantial amounts of water. Proposed projects have increasingly encountered resistance from communities concerned about electricity costs, water use, land development and other local effects.
That has helped fuel interest in space as a possible location for some future computing infrastructure.
Google is not alone in exploring the concept.
Elon Musk, Amazon founder Jeff Bezos and former Google CEO Eric Schmidt have all discussed the possibility of moving data-center computing into orbit as demand for AI processing continues to expand.
For Google, however, an operational orbital data center remains a long-term objective rather than an imminent product.
The immediate question is considerably narrower: whether the same specialized chips powering AI systems on Earth can withstand the vibration of launch, radiation exposure, severe temperature conditions and unusual cooling requirements encountered in space.
The Oct. 1 flight will give Project Suncatcher its first opportunity to answer those questions with hardware operating in orbit.
Comments