The demand for electrical power seems unstoppable as I’ve written about over the past two years.
The combination of demand for new data centers to power artificial intelligence plus general growth in other commercial and residential needs suggest U.S. peak load could grow by roughly 166 gigawatts by 2029, making Doc Brown’s “Great Scott” reaction to needing 1.21 jigawatts seem like a drop in the bucket.
Electric utilities and construction companies are working overtime to meet this projected demand.
But what if actual demand fell short of forecasts? What if there was another way to power AI technology — and do it way less expensively?
Enter Elon Musk, who recently merged xAI into SpaceX in a deal valued at about $1.25 trillion, making it one of the biggest mergers ever.
So, why the merger, and what does that have to do with AI and its insatiable thirst for electrical power? Well, a lot.
Always a dreamer, Musk envisions launching tens of thousands of satellites and orbital AI data centers to use free solar energy and the cold of space as a long-term solution to the massive energy and cooling demands of advanced AI that are hard to meet on Earth.
Think about it: free electrical power from the sun and no need for massive buildings to provide cooling.
Musk claims that within two or three years, orbital AI data centers running on constant solar energy could become cheaper and more scalable than earthbound ones.
Betting against Musk has been a bad idea over the years. The combination of control of rockets, launches, Starlink’s global satellite network (providing last-mile connectivity) and Grok, plus planned orbital data centers, give Musk’s company an end-to-end stack that no other AI company has.
And going public provides SpaceX access to more capital as needed.
But there’s a lot to accomplish if this is to work. First is the anticipated initial public offering of SpaceX (which now includes xAI) for an expected $1.25 trillion.
What if actual demand fell short of forecasts? What if there was another way to power AI technology — and do it way less expensively?
And then comes the technology challenges of building systems and getting them launched in these kinds of numbers, getting the AI models to work, and finally getting the data back down to Earth so it can be used.
These are no small tasks, but folks dispelled earlier Musk ideas like self-driving cars, too.
I know power utilities have been cautious in negotiating with tech giants on meeting expected power demands.
These power generation plants and data centers are incredibly expensive, and providers are demanding long-term commitments from tech firms — whether it turns out they need the power or not.
There are also cries from all the other power users to protect against AI power costs being passed on to them.
If — and when — space can provide more abundant and lower cost power for AI, the demand for Earth-based power and data centers could fall.
This will be fascinating to watch, and potentially transformative. Stay tuned.