It’s hard not to think about SpaceX and AI every minute of every day these days. I remember in the late 1990s., we were all abuzz about the Internet. We thought it would change everything. The comparison of the Internet and AI is valid in that the buzz is similar in feel, but AI may be more fundamental. The hype was genuinely euphoric and widespread with every business needing a “.com” strategy overnight, just as every business now needs an “AI strategy”. Valuations detached completely from earnings with infamous examples like Pets.com, Webvan, eToys raising hundreds of millions on pure narrative. The press coverage was relentless and breathless. Smart people made the case it would change everything, and skeptics were dismissed as dinosaurs or Luddites. I started an New Media incubator, which was the current vernacular for venture capital fund. We were caught up in the buzz as much as anyone and called it B2B Hive in honor of the new lingo for business to business dealings.
There was a genuine underlying technology revolution beneath the froth, as we all know very well now, but there was also plenty of exaggeration. We thought brick and mortar retail was dead and that certainly didn’t happen… right away. It’s hard not to wonder whether the collapse of the mall business of late isn’t the fulfillment of those early e-commerce dreams. Well, I would hardly consider myself a typical retail consumer (I’ve always said that I am a buyer, not a shopper), I would say that 90 to 95% of my retail purchases are done online. Once in a while, I want to see and touch what I’m going to be buying, but those instances are getting fewer and further in between. It’s been 30 years in the making, but I think it’s fair to say that the Internet has pretty much changed everything.
Going back to AI, let’s think about ways that the AI buzz may be bigger or different than it was for the Internet. To begin with, the capital deployment is vastly larger. The dot-com bubble peaked at roughly $5–6 trillion in market cap destruction. AI capex commitments from Microsoft, Google, Meta, and Amazon alone are running $300–400 billion annually already even before the bubble gets big enough for us to worry about it popping. AI is penetrating every industry simultaneously and almost immediately, where the internet took years to reach most businesses. I seem to recall everyone dipping a toe into the Internet game, but now it feels like with AI. everyone is jumping in with both feet andjust hoping they don’t drown.
The technology of AI is more general-purpose and ubiquitous wherethe internet was a communication/distribution layer. It’s important to note that AI tries to, and potentially will touch cognition itself and therefore, dare I say it, REALLY change everything. Back in the day, people all around the world wanted to tap into the Internet, and we had a vision that it would be a great democratization of prosperity. The AI phenomenon feels quite a bit different. Nation-state competition (most notably betweenUS vs. China) adds a geopolitical dimension the the internet boom lacked. The ecological impact is another big difference. I don’t recall anybody being worried about what the Internet would do to pollution or our access to clean air and clean water. By contrast the data center requirements for AI have become such a big issue, the government, jurisdiction and countries are already up in arms and en acting legislation to restrict or at least regulate the building of data centers, mostly do their extreme use of power and water. AI is bringing a whole new meaning to not in my backyard (NIMBY).
The notion of putting AI data centers in outer space has officially moved from sci-fi to a funded, crowded race over the past year, though everything flying so far is demonstration-scale. Given the prominence of that strategy in the SpaceX IPO, it’s fair to say we’ve officially moved to stage two of that proposition. Both SpaceX and Blue Origin have announced plans for orbital data centers, networks of thousands or even millions of spacecraft computing on board, powered by constant solar energy. SpaceX has filed an FCC application for a million satellites, with a test launch targeted by 2028 (reportedly telling IPO-roadshow investors it would be more like late 2027), while Blue Origin’s Project Sunrise filing covers a more modest50,000 satellites. SpaceX just unveiled an 11-million-sq-ft “Gigastat” factory, aiming for 1 GW/year of orbital AI compute by late 2027, roughly 6,000+ satellites annually at 150 kW each, with Musk talking about 100 GW/year by 2030. On actual hardware deployed andin orbit, Starcloud (Nvidia/YC-backed) launched a satellite with an Nvidia H100 in November 2025, and the company recently raised $170M at a $1.1B valuation. Starcloud-2, planned for October 2026, will carry H100s plus Blackwell hardware and the first AWS Outposts deployment in space; the company has filed for an 88,000-satellite constellation toward a 5 GW “hypercluster” with a 4 sq km solar array . Axiom launched its first two orbital data center nodes to LEO in January 2026 , and Google’s Project Suncatcher will launch two TPU-equipped prototype satellites with Planet Labs by early 2027, envisioning 81-satellite clusters.
The economics of all of this rests on two pillars: solar panels in the right orbit are up to 8x more productive than on Earth with nearly continuous power, and you sidestep the permitting, grid, and water fights now plaguing terrestrial builds. But the skepticism is substantive. The GAO notes that large-scale versions need solar arrays bigger than anything ever assembled in space, and radiative cooling at that scale is unproven since heat doesn’t disperse easily in vacuum… and last I looked, spaces of vacuum. There’s also a brutal obsolescence mismatch (thank you, Moore‘s Law) GPUs advance every 1–2 years while satellites last 5–7, plus orbital debris/congestion concerns. The emerging consensus is specialization rather than replacement with orbit handling asynchronous, energy-heavy workloads like model training, while terrestrial centers keep real-time computing going. Bezos’s own framing is telling. He says it will be cheaper than terrestrial data centers “in the next couple of decades”… a long runway for something attracting this much capital today.
Should all of this give us pause? You’re damn right it should. I remember the year 2000 when the NASDAQ died. The dot-com bust was brutal. The Nasdaq fell 78% peak to trough, but the underlying technology was eventually completely vindicated. Amazon, Google, and eventually Facebook emerged from the rubble and became the most valuable companies in history. The skeptics were right about the valuations (at least for those few companies…. I still have shares of some of the ones that weren’t so lucky) and wrong about the technology. Most serious observers think AI will follow the same pattern. There’s genuine bubble dynamics in valuations and narrative, but the technology itself is not going away and will likely be more transformative than the internet, not less. That doesn’t make me watch the shenanigans at the NASDAQ launch of SpaceX without the Cassandra in me coming out.
We’ve all heard the expression FOMO (fear of missing out) and it’s always a fun likeness to paint of someone they are guided by FOMO. Today I saw in one of Scott Gallaway’s comments, a reference to the latest variation of that. It was sad in reference to the SpaceX IPO, the impending OpenAI IPO and probably even they Anthropic IPO the term was JOMO, the joy of missing out. I must admit I’ve never been driven much by FOMO, and I have certainly not been a big proponent of NIMBY thinking. I’m now beginning to think that I’m sympathetic to AI NIMBY thinkers and increasingly giddy about being a JOMO player on the SpaceX and AI investment bandwagons. And here I thought those feelings were only reserved for crypto.