Business Finance Media Technology Policy Wealth Insights Interviews

Art Art Fairs Art Market Art Reviews Auctions Galleries Museums Interviews

Lifestyle Nightlife & Dining Style Travel Interviews

Power Index Nightlife & Dining Art A.I. PR

About About Observer Advertise With Us Reprints

Starcloud CEO Philip Johnston on Putting the First A.I. Data Center in Space

Philip Johnston’s Starcloud is chasing limitless solar energy in orbit to fuel A.I.’s exponential growth. But the idea also faces technical and regulatory challenges.

In November, a 60-kilogram satellite the size of a small refrigerator called Starcloud-1 streaked into low Earth orbit aboard a SpaceX rocket carrying the first data-center-class GPU ever operated in space—an Nvidia H100 roughly 100 times more powerful than any prior orbital compute. Within weeks, Starcloud, the company making the satellite, announced it had trained a language model on the complete works of Shakespeare and had run Google’s Gemini from roughly 200 miles above Earth.

Sign Up For Our Daily Newsletter

Thank you for signing up!

By clicking submit, you agree to our terms of service and acknowledge we may use your information to send you emails, product samples, and promotions on this website and other properties. You can opt out anytime.

“The spacecraft is performing better than we could have hoped for,” Philip Johnston, co-founder and CEO of Starcloud, told Observer. Johnston founded the Redmond, Wash. company in early 2024, backed by a conviction that has since attracted $200 million from investors, including Nvidia, In-Q-Tel, Sequoia Capital and Y Combinator: Rather than build better infrastructure on Earth to satiate A.I.’s runaway energy appetite, move the infrastructure off of Earth entirely. Today (March 30), the company announced it raised a $170 million in Series A funding at a $1.1 billion valuation, helping it reach unicorn status in just 17 months (and the quickest to do so in Y Combinator’s history).

Johnston, 39, brings a distinctly financial pragmatism to the cosmos. Before pivoting to aerospace, Johnston served as an algorithmic trader at BNP Paribas, consulted for national space agencies at McKinsey and co-founded Opontia, an e-commerce aggregator that raised $46 million before being acquired.

A self-avowed space nerd, Johnston came upon the idea for Starcloud after a solo weekend trip to Starbase, the Texas city where SpaceX is headquartered, to see the Starship rocket. What struck him was across the road from the rocket: two gigafactories “similar to Tesla production lines,” he shared in a TED Talk in October, designed to each produce a new Starship every day. He went home, called his co-founders Ezra Feilden, a materials engineer he had grown up with in the U.K., and Adi Oltean, a former principal software engineer at SpaceX, and started running the numbers on what that launch capacity could enable. Data centers kept coming up as the answer.

Starcloud-1 is the first data center in space

The numbers driving his conviction are hard to argue with. A typical A.I.-focused data center consumes as much electricity annually as 100,000 households, according to the International Energy Agency. In space, Johnston says, one square meter of solar panel produces eight times the energy of an equivalent panel on Earth. 

Starcloud-1 is a 60-kilogram aluminum box packed with five Nvidia GPUs in total, with the H100 drawing the most attention, and other supporting hardware. It circles Earth every 90 minutes in a dawn-dusk orbit designed to keep its solar panels in near-constant sunlight. Starcloud is already running edge computing workloads in space for Earth-observation and military satellites: Running inference—say, real-time ChatGPT-like requests that don’t need as much computing power or data transfer as training LLMs—in orbit. “We can shorten the time to actionable insight from a few days to potentially a few seconds,” Johnston said. Longer term, he has filed with the Federal Communications Commission (FCC) for a constellation of up to 88,000 satellites and envisions a 5-gigawatt orbital hypercluster. Starcloud is positioned not as the next AWS, but as neutral infrastructure. “I would view us more like an Equinix,” the CEO said, referring to the global data center developer and operator giant.

Starcloud plans to launch its second satellite, Starcloud-2, in October this year, and a third, Starcloud-2.1, in 2027. The commercial inflection point will happen with Starcloud-3: a 200-kilowatt, three-ton satellite. It’s designed so that 50 satellites, each weighing 3 tons, can be launched with SpaceX’s Starship rocket, which is still in development. Johnston projects mid-2028 as the moment when orbital data centers become cost-competitive with terrestrial facilities, so Starcloud-3 would presumably launch sometime between now and then, though it’s an ambitious bet on a launch vehicle that has repeatedly slipped its own schedules. Further along in the future is Starcloud-4: a Starship-deployed satellite with a four-kilometer solar array powering a 5-gigawatt data center,  a structure that would exceed the generating capacity of the largest power plant in the U.S. For now, however, it’s more concept video than engineering blueprint. 

“By moving A.I. compute to space, we unlock access ot unlimited solar power and completely remove the energy bottleneck,” Johnston said in regards to the latest fundraise, which will go mainly toward development of Starcloud-3, increased headcount, more manufacturing capacity and business development. “This funding allows us to rapidly scale our orbital infrastructure and meet the massive commercial demand for sustainable A.I. compute.” 

Orbital data centers have a cooling problem

Many experts consider thermodynamics the most severe bottleneck for orbital data centers. Since space is essentially a vacuum, containing almost no matter to dissipate heat, it’s inherently cold. It might seem ideal for chilling high-performance electronics, but it’s also an insulator.

“How do you dissipate heat into a vacuum? It’s like your thermos,” Josep Miquel Jornet, a distinguished professor of electrical and computer engineering at Northeastern University, told Observer. “The outside is cold, great. The inside is hot–great if you want a tea, but not if you want to operate a data center in space.”

Without air or water for convective cooling, spacecraft rely on thermal radiation. Cooling a modern A.I. training cluster requires massive radiator panels. Pushing gigawatts of infrared radiation into deep space has consequences, many of which aren’t yet understood. Jornet noted that the astronomical community relies on those same electromagnetic signatures to observe the formation of stars and galaxies. At the constellation scale, radiating infrared outward could interfere with NASA and other observatories scanning those frequencies. “If we start shooting infrared everywhere, it’s going to piss off many people,” he said. Even taking Starcloud’s dawn-dusk sun-synchronous orbit window into account, Jornet says that twilight is a significant window for wide-field surveys that track near-Earth objects, and that space-based infrared telescopes like NASA’s SPHEREx operate entirely above the atmosphere, unaffected by Earth’s day-night cycle.

In a recent episode of Sequoia Capital’s Training Data podcast, Johnson said that cooling consumes roughly 70 percent of his engineering team’s attention. Starcloud-1 is equipped with a proof-of-concept cooling system that can’t run continuously. The production version—a liquid-loop radiator with custom heat sinks feeding a large deployable radiator panel aimed out at space—will go on Starcloud-2 in October. “We’ve tested it in thermal and vacuum chambers, and it works,” he said. Johnston added that chip reliability is another concern: even a 10 percent higher failure rate in space than on Earth “would basically wipe out all of the savings from the energy,” he said. Luckily, the H100 on Starcloud-1 has had no chip-level restart failures to date.

Space data centers are gaining steam amid regulatory and political risks

Starcloud, which currently employs 13 people, is hardly alone in the race. In January, SpaceX filed plans with the FCC for a constellation of up to one million orbital data center satellites. The company cites orbital data centers as a primary growth driver as it prepares for a historic IPO. In the same months, Jeff Bezos’s Blue Origin unveiled its own data center effort, Project Sunrise, which plans to launch up to 51,600 satellites (per its FCC filing) in orbit. Google’s Project Suncatcher plans to launch two test satellites with its A.I. processing chips in 2027 in partnership with Planet Labs. Aetherflux, founded by Robinhood co-founder Baiju Bhatt and backed by Andreessen Horowitz and Breakthrough Energy Ventures, is also targeting a 2027 launch for its first orbital data center node. Axiom Space has already deployed a data center node to the International Space Station. Against that backdrop, Starcloud’s distinction is concrete: It’s the only company to have already demonstrated working A.I. compute hardware in orbit.

Ian Christensen, senior director of private-sector programs at the Secure World Foundation, which filed formal comments with the FCC on SpaceX’s data center application, noted that Starcloud’s filing for 88,000 satellites represents close to an order of magnitude increase over Starlink’s current fleet of 10,000 satellites, and that existing safety standards were established for a far smaller orbital population. Currently, most satellites simply burn up upon re-entry to the Earth’s atmosphere. Space policy experts at the SWF warn that burning up large numbers of aluminum satellites at scale could have unknown effects on the ozone layer. The FCC began reviewing its 1980s-era categorical environmental exemptions roughly a year ago without producing change. 

“It’s an area where I think there might be some downstream regulatory risk in certain jurisdictions that may take a more forward-leaning approach,” Christensen told Observer, noting that environmental law might eventually apply in ways that go far beyond standard space regulation. “Space is a unique domain,” he said. “There are unique regulatory risks that are going to affect your business plan in ways you might not think about for a terrestrial plan.”

Johnston has been building against skepticism since before Starcloud-1 left the ground. Now he has a satellite in orbit running real workloads for paying customers, a production cooling system that hasn’t yet been tested in space and a chip reliability risk he identifies as potentially existential to the business case. And given how much oil and, subsequently, energy prices have risen as a result of increased global conflicts over the past few weeks, the appetite for risk that could solve energy scarcity has grown significantly almost overnight. 

While he may not have had specific current events in mind, it’s worth noting something that Johnston told a TED audience in San Francisco last October: “The most effective way that we can save our own children and grandchildren from the scourge of war will be to stop competing over the fundamentally finite resources of Earth and to start utilizing the near limitless energy of our solar system and eventually of our galaxy.” 

SEE ALSO: Why Ford’s Electric F-150 Never Took Off

We noticed you're using an ad blocker.

We get it: you like to have control of your own internet experience. But advertising revenue helps support our journalism. To read our full stories, please turn off your ad blocker.We'd really appreciate it.

How Do I Whitelist Observer?

Below are steps you can take in order to whitelist Observer.com on your browser:

Click the AdBlock button on your browser and select Don't run on pages on this domain.

For Adblock Plus on Google Chrome:

Click the AdBlock Plus button on your browser and select Enabled on this site.

For Adblock Plus on Firefox:

Click the AdBlock Plus button on your browser and select Disable on Observer.com.

© Observer