SpaceX and Google Are in Talks to Launch Data Centers in Orbit, Here’s Why It Matters
Imagine a data center that never touches the ground. No diesel generators. No cooling towers gulping millions of gallons of water. No planning board hearings where neighbors complain about the noise. Just hundreds of satellites, each packed with AI chips, silently orbiting 400 miles above the Earth, drinking unfiltered sunlight and beaming compute power back down.
Sounds like science fiction, right?
Except, as of this week, it’s not.
On May 12, 2026, the Wall Street Journal dropped a report that stopped the tech world mid-scroll: Alphabet’s Google and Elon Musk’s SpaceX are in advanced talks to put data centers into orbit.
Let that sink in for a moment. Two companies that have spent years circling each other, sometimes as allies, sometimes as rivals, may be about to partner on the most ambitious infrastructure project since the internet itself.
What’s Actually Happening? The SpaceX-Google Talks Explained
The WSJ Scoop and What We Know So Far
According to the Journal, Google is negotiating a rocket-launch deal with SpaceX that would help the search giant place orbital data centers in space. The talks are still in progress, and Google is reportedly speaking with other launch providers too. But SpaceX, as the dominant private rocket company on the planet, is the obvious frontrunner.
This isn’t entirely out of the blue. Google already owns a 6.1% stake in SpaceX, and a Google executive holds a seat on SpaceX’s board. The relationship, as they say in Silicon Valley, has history.
But the timing? That’s where it gets fascinating. SpaceX is preparing for what could be the largest IPO in history, a $1.75 trillion market debut expected as soon as June 2026, and orbital data centers have become a centerpiece of the company’s pitch to prospective investors. Elon Musk has publicly declared space the cheapest place to run AI within three years. A partnership with Google would add serious credibility to that claim, right when SpaceX needs it most.
Side note: there's a quiet irony here. Musk helped launch OpenAI in 2015 partly to counter Google’s AI ambitions. A decade later, his rocket company may become the launch provider for Google’s orbital AI infrastructure. Tech grudge matches, it turns out, have expiration dates, especially when IPOs are involved.
Project Suncatcher: Google’s Moonshot for Orbital AI
Google’s side of this story has a name: Project Suncatcher.
Unveiled quietly in November 2025, Project Suncatcher is Google’s "moonshot" initiative to build data centers in space. The plan is ambitious but concrete: launch two prototype satellites by early 2027, each equipped with Google’s custom-designed TPU chips (Tensor Processing Units, the same silicon that powers Gemini, Google’s flagship AI model).
Google has partnered with Planet Labs, the satellite-imaging company, to build and operate those prototypes. The test mission will validate whether TPUs can survive the harsh radiation environment of space and whether two satellites can coordinate compute tasks across optical laser links.
Google CEO Sundar Pichai framed it like this: "We'll send tiny racks of machines and have them in satellites, test them out, and then start scaling from there. 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."
If the tests succeed, the vision scales fast: a constellation of 80-plus satellites in sun-synchronous low-Earth orbit, connected by terabit-capable laser links, forming a distributed AI cloud that never sleeps.
Why Space? The Physics That Makes Orbital Data Centers Tempting
Before we get to the obstacles, and there are many, let’s talk about why smart people like Sundar Pichai and Elon Musk are betting billions on this idea. The physics is actually pretty compelling.
Solar Power Without the Atmosphere
Here’s a fact that stuck with me: a solar panel in orbit can produce up to eight times more energy than the same panel sitting on Earth’s surface.
Why? No atmosphere. No clouds. No night, at least not in a sun-synchronous orbit, where satellites can be positioned to stay in continuous sunlight. The solar constant in low-Earth orbit delivers roughly 1,361 watts per square meter of pure, unfiltered radiant energy.
Google’s research blog put it bluntly: the sun produces one hundred trillion times more energy than all of human civilization generates today. Tapping even a fraction of that, without atmospheric losses, changes the energy economics entirely.
No Cooling Towers, No Water Wars
Data centers on Earth generate ferocious heat. Cooling them is a massive engineering challenge, and a growing environmental headache. U.S. data centers are on track to consume up to 9% of the nation's electricity by 2030, with cooling alone accounting for a significant chunk.
In space, though, the surrounding temperature drops to roughly -270°C. That’s cold enough to make liquid nitrogen look warm. Theoretically, you can just radiate excess heat away into the void. No water required. No cooling towers. No fights with local governments over aquifer rights.
Well... theoretically. We’ll get to why that’s trickier than it sounds in a moment.
No NIMBYs, No Permitting Gridlock
Even if you solve the power problem on Earth, you still have to build the data center somewhere. In places like Northern Virginia (the global data center capital), grid interconnection queues now stretch 7 to 12 years. You read that correctly, you might wait a decade just to plug in.
In space? No zoning board. No neighbors. No environmental review that takes longer than building the actual facility. As one space industry analyst put it: "The biggest problems are cooling, security, power transmission, all those things can be solved if you just move it into space."
That’s the dream, anyway.
The Numbers That Make This a Truly Massive Bet
Let’s put some scale on this, because the numbers are staggering:
- SpaceX has filed with the FCC to launch up to 1 million satellites to support its orbital data center ambitions.
- The company aims to shift up to 100 gigawatts of compute off the planet , that’s more than the current total global data center capacity.
- The orbital data center market is projected to grow from roughly $1.77 billion in 2029 to $39.09 billion by 2035, a 67.4% compound annual growth rate.
- Goldman Sachs estimates current global data center capacity at about 59 GW. Musk claims Starship could deliver 300-500 GW per year of solar-powered AI satellites to orbit.
- Hyperscalers (Amazon, Microsoft, Google, Meta, Oracle) are expected to spend roughly $400 billion on terrestrial data centers in 2026 alone.
Those are the kinds of numbers that make investors lean forward in their chairs. And with SpaceX’s IPO looming, they’re exactly the narrative the company wants in the air.
The Brutal Reality Check, Cooling, Costs, and Catch-22s
Okay. Time for the other side of the story, the part that makes engineers wince.
Jensen Huang’s Blunt Warning About Radiative Cooling
Nvidia CEO Jensen Huang was asked about orbital data centers on a podcast in March 2026. His response was... measured, in the way a surgeon is measured when describing a complicated operation.
"In space, there’s no conduction, there’s no convection, it’s just radiation," Huang said. And radiative cooling, where heat leaves a surface as infrared light, requires enormous radiator panels. "This makes the system extremely complex and expensive," he continued. He estimated solving the cooling challenge could take years.
Think of it this way: on Earth, you can cool a server by blowing air over it (convection) or running cold water through a plate (conduction). In the vacuum of space, neither works. You have to build giant radiator wings that glow in infrared, like the ones on the International Space Station, and those wings are heavy, which means higher launch costs, which means the economic case gets shakier with every kilogram.
The Economic Equation That Hasn’t Closed Yet
Andrew McCalip, a space engineer, built a calculator comparing terrestrial and orbital data center costs. His baseline estimate: a 1 GW orbital data center would cost roughly $42.4 billion, nearly three times more than a ground-based equivalent , once you factor in satellite construction and launch costs.
And launch costs are the linchpin of the entire model. Falcon 9 currently delivers payloads at roughly $3,600 per kilogram. To make orbital data centers economically competitive, experts say you need something closer to $200 per kilogram , an 18-fold improvement that depends entirely on Starship, SpaceX’s still-in-development mega-rocket, achieving rapid, reliable reusability.
If Starship works as advertised? The math starts to make sense. If it doesn't? The entire business case collapses.
SpaceX’s Own S-1 Filing: “May Not Be Commercially Viable”
Here’s the twist that didn’t get enough attention when the Google-SpaceX talks hit the headlines.
In its pre-IPO S-1 filing with the SEC, SpaceX included a risk disclosure stating that its orbital AI data center plans "involve significant technical complexity and unproven technology that may not achieve commercial viability" and would operate in "extreme and unpredictable space environments."
In other words: SpaceX is telling investors something unusually honest, this might not work. Meanwhile, on the same roadshow, executives are positioning orbital data centers as the company’s next major growth engine.
That’s not dishonesty. That’s the reality of frontier technology. But it’s a tension worth keeping in mind as the headlines get breathless.
The Anthropic Connection, and Why It Changes the Story
While the world focused on the Google-SpaceX talks, a parallel deal may end up being just as significant.
In early May 2026, SpaceX announced it was leasing the entirety of its Colossus 1 data center in Memphis, more than 300 megawatts of compute, backed by over 220,000 Nvidia GPUs , to Anthropic, the AI company behind the Claude chatbot.
But here’s the kicker: Anthropic also expressed interest in working with SpaceX to develop multiple gigawatts of space-based orbital data centers in the future.
So the picture looks like this: SpaceX is building the rocket capacity (Starship), securing terrestrial compute revenue (Colossus 1 + Anthropic), merging with its own AI company (xAI, now "SpaceXAI"), and lining up a launch customer (Google), all while preparing for the largest IPO ever.
It’s a vertical integration play of staggering scope. Whether it’s brilliant or overextended depends on which analyst you ask. But you can’t deny it’s coherent.
What Happens Next? Timeline Out to 2035
Here’s a realistic, hype-free timeline:
What This Means for AI, Energy, and Investors
Let’s zoom out.
The AI industry has a problem that no amount of clever prompt engineering can solve: it's running into a physical wall. The electrical grid wasn’t built for models that consume the energy of small towns. Permitting takes a decade. Water for cooling is increasingly contested. Land is finite.
Orbital data centers aren’t a silver bullet. They may not even work at scale. But they represent something important: the tech industry finally acknowledging that terrestrial infrastructure has ceilings, and those ceilings are closer than we thought.
If SpaceX and Google can make this work, even partially, it rewrites the rules for where and how AI compute lives. It separates compute from geography. From politics. From local grid capacity.
That’s worth paying attention to, whether you’re an investor, a data center operator, or just someone curious about where the cloud actually lives.
A Bet Worth Watching
I’ll be honest: when I first read about orbital data centers, I rolled my eyes a little. It sounded like the kind of thing a billionaire says on a podcast to sound visionary. And honestly? Some of it still does.
But the more I dug into this story, the more I realized something: the engineering is brutal, the economics are unproven, and the timeline is optimistic at best, but the physics actually checks out. Solar power in space really is eight times more efficient. The vacuum really does solve the cooling-water problem. Launch costs really are falling, even if they’re not where they need to be yet.
Whether SpaceX and Google succeed won’t be clear for years. But one thing is certain: a conversation that started as a science-fiction thought experiment has now become a negotiation between two of the most powerful companies on Earth, about building infrastructure beyond it.
That’s a bet worth watching. And if you’re involved in AI infrastructure, energy planning, or tech investment, it’s one you should probably be tracking.
Got thoughts on orbital data centers? Think it's visionary, or vaporware? Drop a comment below. I read every one, and I'd love to hear where you land on this.
And if you found this breakdown useful, subscribe to the newsletter. I cover the intersection of AI, infrastructure, and the future of computing every week, no jargon, no breathless hype, just honest analysis.
Comments
Post a Comment