Elon Musk believes one of the simplest ways to resolve the difficulties of constructing AI knowledge facilities on earth is to transfer them into outer house. His merger this week of his rocket firm SpaceX along with his synthetic intelligence firm xAI may assist get them there.
And he isn’t the one one considering that means.
“The only logical solution…is to transport these resource-intensive efforts to a location with vast power and space. I mean space is called ‘space’ for a reason,’” Musk wrote Monday when announcing the merger.
Musk has efficiently launched bold tasks earlier than, like creating mass market electrical automobiles and creating reusable rocket engines to carry individuals and cargo into house. This time, Google, OpenAI and others are additionally wanting to create knowledge facilities in house
AI knowledge facilities’ large demand for each energy and water means rising AI know-how would require options.
“We are tending to exceed the ability to generate the power (needed),” mentioned David Bader, distinguished professor of information science on the New Jersey Institute of Technology. “I think it is a necessity for looking not on terrestrial ground but looking to space to provide some of these solutions.”
Space presents higher entry to photo voltaic vitality, and the atmosphere additionally retains the tech cool and bypasses the necessity for land.
“There are clearly technical challenges to making this a viable endeavor, but these seem to be engineering constraints as opposed to physics,” mentioned a Deutsche Bank Research analyst word on orbital AI knowledge facilities final month, noting that many firms had been exploring methods to make it work.
Google introduced plans in November to take a look at orbital AI knowledge facilities by launching two take a look at satellites as early as subsequent 12 months.
“In the right orbit, a solar panel can be up to eight times more productive than on earth, and produce power nearly continuously,” Google mentioned in a statement on the time. “In the future, space may be the best place to scale AI compute.”
Sam Altman, CEO of OpenAI, final summer time thought-about shopping for rocket firm Stoke Space to put knowledge facilities in orbit, in accordance to the Wall Street Journal.
And in November, Washington-based AI start-up Starcloud launched a take a look at satellite tv for pc with an AI server aboard a SpaceX rocket.
“In space, you get almost unlimited, low-cost renewable energy,” mentioned Philip Johnston, cofounder and CEO of Starcloud.
Johnston predicted that, inside 10 years, all new AI knowledge facilities will likely be in orbit, which may additionally resolve the rising backlash to AI knowledge heart development.
The price of offering knowledge facilities with the electrical energy wanted is probably going driving up electrical payments for shoppers. A Bloomberg News analysis discovered that areas close to knowledge facilities noticed a rise in electrical energy prices of as a lot as 267% in contrast to 5 years in the past.
Bader mentioned a tough quantity for the elevated electrical price to shoppers is hard to decide due to the dearth of public info on knowledge heart utilization, however it is evident that customers are paying extra due to AI vitality calls for.
“As the demand goes up with the limited supply, the cost to consumers in those markets also tends to increase,” he mentioned.
Data facilities may also create water shortages for neighboring communities. A big knowledge heart can burn up to 5 million gallons of water per day, the equal to the water use of a city of 10,000 to 50,000 individuals, in accordance to the Environmental and Energy Study Institute.
“The Earth may be becoming a complicated place for Big Tech’s data center development,” mentioned Mark Muro, a senior fellow at Brookings Metro. Political backlash in many communities is making it tougher to get approvals for extra development, he added.
No matter the technological hurdles of transferring knowledge facilities to obit, Big Tech wants to discover new methods to ship the wanted gigawatt energy, Muro mentioned.
“It’s not just consumers’ electric bills going up,” he mentioned. “It’s Big Tech’s power bills going up immensely too. They’re already paying top dollar.”
Musk predicted Monday that orbital knowledge facilities will likely be less expensive than earth-bound ones “within two to three years.” Experts disagree: Deutsche Bank estimates it will likely be properly into the 2030s earlier than orbital knowledge facilities “reach close to parity.”
Musk has a historical past of overpromising technological advance timelines. But knowledge facilities in house may finally occur. The price of launching satellites to house is coming down simply as the prices of constructing and working AI knowledge facilities on earth are rising.
“Two or three years may be a stretch,” mentioned Bader. “But I would believe in three to five years, it would more comfortably be a regular deployment for AI data centers to be able to do that processing in space.”