By Rodger Morrow, Editor & Publisher, Beaver County Business
Listen to a podcast discussion about this article.
From Clouds to Orbit
There was a time when “the cloud” meant a puff of weather drifting over the Ohio River, possibly dropping rain on a Little League doubleheader. Then it meant someone else’s computer in Northern Virginia. Now, according to Elon Musk, it may soon mean a rack of humming servers orbiting the Earth at 17,500 miles an hour, blissfully sunburned and beyond the reach of zoning boards.
Musk’s Orbital Vision
On a recent appearance on the Dwarkesh Podcast, Musk predicted that within 36 months the most economically compelling place to put artificial intelligence would be space. Not cheaper, mind you—just possible. On Earth, he argues, we are running out of the essentials: power, land, patience. In space, by contrast, “it’s always sunny,” the cooling is a vacuum, and no one shows up at a township meeting to complain about the view.
It’s classic Musk: audacious, elegant, faintly alarming, and delivered with the casual confidence of a man who has already launched his car toward Mars. The logic goes like this. AI needs electricity the way a steel mill needs iron ore. Terrestrial data centers are gulping power so fast they’re starting to look less like warehouses and more like small nations. Grids are straining, regulators are frowning, and neighbors are discovering that a data center is a lot louder than a cornfield.
Why Space Looks Tempting
Space, on the other hand, offers solar energy in industrial quantities—five times more efficient without the atmosphere getting in the way—plus natural cooling and unlimited elbow room. If humanity hopes to grow into a Kardashev Type II civilization, harnessing the full output of the sun, this is apparently the on-ramp. It also helps explain why SpaceX has been busy filing FCC paperwork for up to a million satellites, and why Musk recently folded xAI into the SpaceX ecosystem. Visionaries call it destiny. Cynics note that it also sounds like a very effective pitch deck.
The Practical Objections
To be fair, Musk is not alone in gazing skyward. Engineers at Google, Amazon, and Open AI have all kicked the tires on orbital computing. Google’s Project Suncatcher plans test launches later this decade. The European Union has studies with names that sound like prog-rock albums. A handful of startups insist that by the 2030s, once launch costs fall and lasers start flinging data between satellites at obscene speeds, the heavens will be full of GPUs.
Then come the wet blankets, armed with calculators. Launch costs would need to drop from roughly $3,000 a kilogram to something closer to $200 for the numbers to make sense, and even Musk admits that’s a stretch before the mid-2030s. Space is a fine place for solar panels, but a harsh one for delicate chips, which do not appreciate being sandblasted by radiation. Cooling works differently in a vacuum, requiring radiators the size of football fields. Latency— those tiny but maddening delays—makes space a lousy spot for anything that needs instant responses. And all of it adds to a growing halo of orbital debris, which is not exactly the environmental solution it sounds like at first blush.
Back to Beaver County
In other words, space data centers are probably coming someday. Just not by 2029, unless the laws of physics agree to work overtime.
Which brings us back down to Earth, and specifically to Beaver County.
If you stand in Shippingport and squint hard enough, you can almost see the future rising from the footprint of the old Bruce Mansfield coal plant. Aligned Data Centers wants to build a $10 billion campus there—Project Phoenix, naturally—spreading more than two million square feet of servers across three buildings, powered by gas and humming away at the edge of the Ohio River. Nearby, the Beaver Valley Nuclear Power Station is in talks to feed electrons to AI workloads, including a high-profile deal with Meta. Old steel sites in Midland and Aliquippa are being eyed with the same speculative gleam once reserved for beachfront property.
Complement, Not Replacement
Should anyone here be worried that all of this will be obsolete in 36 months, replaced by server farms circling the globe like a technological halo? Probably not.
Even if orbital AI becomes a reality, it’s far more likely to complement what we’re building than replace it. Space is good for brute-force training—long, power-hungry computations that can tolerate a little delay. Earth is better for everything else: applications that need to answer now, hardware that needs a wrench now, and facilities that benefit from being near people rather than micrometeorites. Terrestrial data centers still win on maintenance, speed of deployment, and the simple virtue of gravity.
Where the Power Is
If anything, Musk’s celestial musings underline how valuable places like Beaver County have become. We have land with industrial bones, power sources that don’t blink, and communities that know the difference between a passing fad and a long-term employer. Global AI demand is projected to double data-center electricity use by the end of the decade. That doesn’t vanish because someone put a server in orbit.
So yes, someday the cloud may drift beyond the clouds, glowing serenely in permanent sunshine. Until then, the future of AI will look a lot less like science fiction and a lot more like Shippingport—concrete, cables, cooling towers, and the quiet satisfaction of being right where the power is.
