Even as Google dramatically expands its portfolio of clean energy infrastructure, the hyperscaler expects to need to find more energy to power its operations.
One possible solution is space-based machine learning.
Last November, Google announced the launch of Project Suncatcher, an effort to one day extend machine learning into outer space. As a first step, Project Suncatcher will send two solar-powered satellites equipped with tensor processing units (TPUs) into orbit next year. In orbit, exponentially more solar energy is available than on Earth.
About ” [one ten billionth] The energy that the sun emits falls on the Earth,” explained Blaise Agüera y Arcas, CTO of Technology and Society at Google and core member of the Project Suncatcher team, in a recent episode. where the internet is.
Agüera y Arcas brings a unique perspective to the conversation about extraterrestrial data centers. Author of the 2025 book What is Intelligence?, his approach to artificial intelligence centers on the provocative idea that artificial intelligence is not artificial at all, but instead that computer parallelism mirrors the structure of biological evolution. As a result, he believes technology will expand dramatically.
“I think our thirst for intelligence is limitless,” Aguera y Arcas explained on the podcast. “In nature, brains grow until they can’t get any bigger. Cities grow until they can’t get any bigger. Even assuming we get a factor of 1,000. [energy] When you think about efficiency, you need to think about where the energy comes from. The answer is obvious. There are places in the universe where large amounts of energy are available. ”
Agüera y Arcas helped design Project Suncatcher and its initial satellites through Google’s Paradigms of Intelligence team, which he founded. These satellites need to be as flat as possible to maximize their absorption of the sun. “Any volume, any three dimensions, is just a mass that has to be launched,” Aguera y Arcas explained. “In other words, the challenge is how much you can make it two-dimensional.”
The project also presents other significant engineering challenges. For example, there are issues with temperature control and how to dissipate excess heat from the TPU. “Liquid cooling is not possible in space. The only way to dissipate heat in space is through infrared radiation from objects,” Aguera y Arcas said. “So you need a large surface to radiate the heat.”
He cited communication as another hurdle. Data centers typically use fiber optics to transmit data, but that physical infrastructure is impossible in space. As an alternative, Aguera’s y Arcas is pushing for a “free-space optical link,” or communicating with satellites via lasers. “In other words, you’re communicating with lasers within the satellite constellation. That requires the development of new kinds of optical communications that are still very experimental on Earth,” he explained.
Aguera y Arcas expects hyperscalers to continue to embrace terrestrial energy options like traditional solar power, but argues that the sheer scale of solar power available in space means off-planet options are a real possibility.
For the complete conversation with Blaise Agüera y Arcas, listen to his interview at: “Where the Internet is” Season 5:
Listen to the episode:
This is partner content provided by Google. This is taken from an interview published in. where the internet isThis is a podcast produced in partnership with. latitude studio And Google.
Where the Internet Lives is an award-winning podcast about the invisible world of the data center. continue apple, spotifyor wherever you get your podcasts.
