SpaceX, orbital data centers, and the journey to Marsby Jeff Foust
|
| “We can’t go public until we’re flying regularly to Mars,” Shotwell said in 2018. |
In the interview, Shotwell was asked when the company might consider going public. Not for a long time, she responded, because the company was focused on company founder Elon Musk’s goal of Mars. “Getting people to Mars, it’s a big job. It’s going to take years,” she said, saying the company was “picky” about who it wanted to invest in it. “We want investors with patience as well as excitement about what we’re trying to do.”
“We actually don’t talk too much about going public right now,” she said. “We can’t go public until we’re flying regularly to Mars.”
A lot has changed since then: SpaceX has become the dominant launch provider, performing more orbital launches than the rest of the world combined. Starlink has become the revenue-generating machine hoped for years ago, with millions of customers worldwide. Crew Dragon has become a lifeline for the International Space Station as its competitor in the commercial crew program, Boeing’s Starliner, has stumbled.
But, the company hasn’t reached Mars yet, with people or with cargo (at the time of Shotwell’s interview, the company wanted to start sending people in 2024 in what is now known as Starship.) And yet, it is looking increasingly likely that SpaceX will go public as soon as next year before the company's goal of human Mars missions.
Reports earlier this month first indicated that SpaceX was considering an IPO, perhaps in the second half of next year. It was not the first time since Shotwell’s 2018 interview that the idea of SpaceX going public had come up, perhaps by spinning out Starlink into a standalone, public company while keeping the rest of SpaceX private. Those talks, though, never appeared to gain momentum.
This time, though, the discussions were about an IPO for the entire company, and gained momentum. On Friday, in a message to SpaceX employees, the company’s chief financial officer, Bret Johnsen, confirmed the company was laying the groundwork for an IPO.
Johnsen wrote that “we are preparing the company for a possible IPO in 2026. Whether it actually happens, when it happens, and at what valuation are still highly uncertain.” An IPO, he said, would provide SpaceX with “a significant amount of capital that enables us to ramp Starship to an insane flight rate, deploy AI data centers in space, build Moonbase Alpha and send uncrewed and crewed missions to Mars.”
It's clear something has changed significantly at SpaceX about an IPO, even as the company continued to do well privately. Johnsen’s comments about an IPO came in a message about the company’s latest “tender offer,” where it allows insiders like employees to sell stock, providing them with liquidity outside of the public market. The latest tender offer doubled the price of SpaceX shares, giving the company a valuation of about $800 billion.
| An IPO, Johnsen said, would provide SpaceX with “a significant amount of capital that enables us to ramp Starship to an insane flight rate, deploy AI data centers in space, build Moonbase Alpha and send uncrewed and crewed missions to Mars.” |
“For the last several years, there was this idea floated that SpaceX wouldn’t IPO until they make it to Mars or that potentially, upon hitting a certain customer milestone, maybe Starlink would spin off,” said Andrew Chanin, co-founder and CEO of ProcureAM, which manages the Procure Space ETF, an exchange-traded fund focused on space industry companies. “And it now seems like that story is possibly changing.”
He and others have speculated the change has to do less with Starship or Moon and Mars missions than with the other item Johnsen mentioned: orbital data centers. It is a concept that has gained public traction only in the last several months given the sharply growing demand for data centers to support AI companies as well as the challenges terrestrial data centers face, particularly with power.
Musk has been among those talking up orbital data centers, just in the last couple of months, linking them to the company’s Starlink constellation. “Satellites with localized AI compute, where just the results are beamed back from low-latency, sun-synchronous orbit, will be the lowest cost way to generate AI bitstreams in <3 years,” he wrote on X earlier this month.
“And by far the fastest way to scale within 4 years, because easy sources of electrical power are already hard to find on Earth,” he added. “1 megaton/year of satellites with 100kW per satellite yields 100GW of AI added per year with no operating or maintenance cost, connecting via high-bandwidth lasers to the Starlink constellation.”
The rush to build orbital data centers
Musk is not alone in his interest in orbital data centers. “These giant training clusters, those will be better built in space, because we have solar power there, 24/7,” Jeff Bezos said of AI computing in an onstage interview at Italian Tech Week in October. “We will be able to beat the cost of terrestrial data centers in space in the next couple of decades.” Blue Origin, his space company, has reportedly been looking into orbital data centers in the last year.
Meanwhile, Google recently announced it would work with Planet, which operates a constellation of imaging satellites, on a test to fly Google AI processors on a Planet satellite as a prototype of orbital data centers. Sundar Pichai, Google’s CEO, said in a TV interview described the Suncatcher project as one of the company’s “moonshots.”
“There’s no doubt to me that, a decade or so away, we’ll be viewing it as a more normal way to build data centers,” he said.
“Suncatcher is really exciting. I do think it's a very viable project long term,” Will Marshall, CEO of Planet, said in a company earnings call last week, noting that for now Suncatcher is only an R&D program. “I see a huge market opportunity here. I do in the long run.”
The concept had led others in the technology field to at least consider getting into the space industry. Orbital data centers may be one reason former Google CEO Eric Schmidt, a prominent voice on AI topics, elected to invest in and become CEO of launch startup Relativity Space earlier this year, although both he and the company have been quiet about their plans.
| “The race for artificial general intelligence is fundamentally a race for compute capacity, and by extension, energy,” said Bhatt. |
Sam Altman, the CEO of OpenAI, explored invested in Stoke Space, another launch startup, the Wall Street Journal reported earlier this month, as part of his interest orbital data centers. That could have involved billions of dollars of investment to take a majority stake in the company. The talks broke off, and when Stoke Space announced a $510 million funding round in October, Altman was not among the investors listed as participating in the round.
The concept of orbital data centers faces many challenges, such as the ability for advanced AI processors to operate in a radiation environment in low Earth orbit much higher than on the ground, as well as the need for large solar panels and radiators to both generate the power they will need and to eliminate the excess heat. Some think those challenges can be overcome, while others see them as being, at the very least, expensive and potentially uneconomical to do so.
The split is evident in an industry one might consider adjacent to orbital data centers: space-based solar power, which also requires large arrays to generate power. Last week, Aetherflux, one startup that raised $50 million earlier this year to advance work on space solar power satellites, said it planned to develop data center nodes called “Galactic Brain” that would start launching in 2027.
“The race for artificial general intelligence is fundamentally a race for compute capacity, and by extension, energy,” Baiju Bhatt, founder and CEO of Aetherflux, said in a statement. “Galactic Brain puts the sunlight next to the silicon and skips the power grid entirely.”
Another space-based solar power startup, Overview Energy, exited from stealth the next day to announce its plans, including a test of beaming energy from a moving aircraft to a ground receiver using a near-infrared laser. The company plans to launch a test satellite in 2028 and start commercial spacecraft as soon as 2030.
Overview, unlike Aetherflux, isn’t interested in orbital data centers. In an interview, company founder and CEO Marc Berte said it made more sense to leave the data centers on the ground and use the satellites to beam power to them, addressing the power grid concerns of terrestrial data centers.
“The real question should be not where you put the GPUs. The real question should be, where do you put the energy?” he said. Data centers in space will require frequent upgrades to update their processors, something straightforward to do on the ground but complex in space, short of simply launching a new data center spacecraft.
“Put the thing that doesn’t need the maintenance and can last for a long time in space,” he said. “Put the energy part into space as opposed to putting the processing part in space.”
Industry effects
If SpaceX does go forward with an IPO, it could raise by some estimates $20–30 billion dollars and value the company at $1.5 trillion. Analysts think that SpaceX should have no problem attracting buyers should it go public.
| “A SpaceX IPO in 2026 would be a seismic event for the space economy,” said Boggett. |
“We hear interest from investors on a regular basis looking for access to the stock,” ProcureAM’s Chainin said. “This is something many retail and institutional investors have been interested in for a while.”
Could that interest, though, take attention, and money, away from other companies in the space sector? Would investors prefer to put their money into SpaceX, a leader in many parts of the space industry, versus other companies in the sector, be they publicly traded or still raising capital privately?
“A SpaceX IPO in 2026 would be a seismic event for the space economy,” wrote Mark Boggett, CEO of Seraphim Space, a fund that invests in space companies. The funds would “give SpaceX the firepower to accelerate its operations in infrastructure, connectivity, and AI-powered services.”
He argued that a SpaceX IPO could jumpstart the industry, providing additional opportunities for others in the field. “That level of activity will create unprecedented opportunities across the broader space economy, from in-orbit servicing, satellite constellations, data analytics, to global security and sustainability solutions,” he stated. “It’s about backing the infrastructure and services that will define the next industrial revolution that we believe will result in many more giant space businesses like SpaceX being created over the forthcoming decade and beyond.”
Chanin held similar views. He noted that many companies in the industry have benefited from the increased space access SpaceX offers with its frequent launches, which would continue and likely even accelerate with Starship.
“There’s a lot of need for various technologies, some that SpaceX may have no interest in building out,” he added. “That will potentially create opportunities for other space companies.”
The strategy SpaceX appears to be pursuing offers significant risks and rewards. If SpaceX is able to tap into what today seems like an insatiable demand for AI computing, that could further turbocharge the company and its finances beyond what Starlink has done. Financing human Mars missions becomes much less of a challenge.
But, if the AI boom goes bust for any reason, that could drag down SpaceX if it does make a big bet on orbital data centers, and with it other parts of the space industry. Mars might recede into the distance.
SpaceX might no longer need to stay private to be able to go to Mars; instead, it could be necessary to go public to tap into the resources to enable that vision. However, it may still need, as Shotwell put it seven and a half years ago, “investors with patience as well as excitement about what we’re trying to do.”
Note: we are now moderating comments. There will be a delay in posting comments and no guarantee that all submitted comments will be posted.