The SpaceX IPO exploded onto markets and then blinked. Masayoshi Son sat on stage in Tokyo and dismissed a dream that just made Elon Musk a trillionaire overnight. You could feel a question hanging in the room: who will actually win the AI war, Earth or orbit?
I’ve watched enough tech spectacles to know how quickly hype hardens into strategy. You should be skeptical when a company sells centuries of future profits in a single filing. I’m going to cut through the shine and tell you why Son thinks the fight for AI will be decided on terra firma.
At the investor event, a single sentence landed like a verdict. ‘He who strikes first wins’
At that Tokyo event, Son didn’t give a lecture—he gave a margin call on fantasy. He said the next few years on Earth will define the AI race, not ambitions a decade out in orbit.
Son’s logic is blunt and financial: electricity savings from orbital solar farms are one lever, but power is only a sliver of an AI data center’s bill. The rest is chips, servers, cooling, maintenance and the human expertise that keeps all of it humming. Launching racks into space adds big new costs—rockets, redundancy, latency headaches—so the math doesn’t yet favor orbit.
He called Elon Musk a remarkable agent of change, then pivoted to SoftBank’s plan: build massive, terrestrial capacity and move fast. That’s a bet on execution and speed. As Son put it, “He who strikes first wins.”
Why would companies put data centers in space?
Think lower electricity bills and abundant solar. Musk has suggested satellites could host solar-powered GPUs at scale, and Jeff Bezos has said humans might build orbital data centers within two decades.
But here’s where reality intrudes: sending hardware to orbit is not just an engineering puzzle—it’s a recurring cost center. You’re trading familiar operational expenses for launch manifests, replacement cycles measured in launches, and complex network architecture to bridge space and ground. For investors, that trade needs a credible line from launch to margin.
On trading floors and analyst reports, the IPO’s sparkle met practical scrutiny. Son isn’t the only one bearish on orbital data centers
Minutes after SpaceX hit the market, the stock wobbled below its opening price and then climbed back—proof that enthusiasm and doubt were fighting in real time.
SpaceX’s IPO filing claims a $28.5 trillion total addressable market (about €26 trillion), with roughly $26.5 trillion (€24 trillion) expected to come from AI. That scale is arresting; it’s also a sales pitch for imagination. Morningstar and other analysts have warned the valuation is optimistic. Morningstar gives a best-case scenario—where reusable rockets and orbital compute capture meaningful market share—a single-digit probability (7%).
The IPO made headlines and made Musk a trillionaire on paper, but shares wobbling under their opening-day price of $150 (€138) reminded everyone that markets prize proof over promise.
Are orbital data centers economically viable?
Short answer: maybe one day, probably not yet.
There are technically conceivable pathways—reusable rockets, modular space infrastructure, and economies of scale—but each step requires capital and time. For companies already spending tens of billions on data center hardware and chips, the immediate return from orbit is thin. In the meantime, terrestrial strategies—fast buildouts, partnerships like SoftBank’s with OpenAI and the Stargate data center initiative—offer clearer near-term leverage.
What did Masayoshi Son say about SpaceX’s plan?
He said orbital data centers would likely cost more than they save and that the AI race will be decided closer to home in the coming years. Son framed the debate in practical cost buckets: power is cheap relative to silicon and server infrastructure, and space introduces new recurring costs. He acknowledged Musk’s visionary role but doubled down on building formidable Earth-based capacity.
I’ll give you two images to keep this straight: the IPO’s valuation looked like a mirage in the desert—tempting but unstable—and the AI race now resembles a high-stakes chess game where speed, positioning, and reliable pieces matter more than dramatic gambits. You can admire the ambition and still bet on the side that delivers results this quarter, not ten years from now.
SoftBank is moving money into OpenAI and practical data center projects; SpaceX is selling a long, lofty runway full of possibility. You need to decide whether you’re buying tickets for a launch or shares in the machines that will actually run AI tomorrow—where do you put your stake?