Even Sam Altman Thinks Elon Musk's Space Data Center Pitch Is Overselling It And He's Not Wrong

Sam Altman and Elon Musk spent the weekend trading insults on social media, which is not, in itself, particularly newsworthy. These two have a long and well-documented history of public sniping.

But buried inside the back-and-forth was something genuinely substantive: Altman accidentally articulated the exact skepticism that space industry experts, satellite entrepreneurs, and Google's own orbital compute engineers have been quietly voicing for months, just with fewer curse words and more technical caveats.

After Musk accused Altman of being a scammer, Altman fired back: "Homeboy, you're the one selling public market investors on short-term space datacenters."

Strip away the typo and the "homeboy," and what's left is a genuinely important claim: space data centers the concept currently underpinning a meaningful chunk of SpaceX's astronomical valuation are not going to become a real business anytime soon. And Altman isn't the only one saying it. He's just saying it loudest, and with the most direct motive to needle Musk publicly.

The $2 Trillion Idea Getting Questioned

SpaceX's plan to launch a fleet of orbital data center satellites designed to perform AI inference tasks directly in space, taking advantage of unlimited solar power and the vacuum's natural cooling properties, is one of the central pillars behind the company's roughly $2 trillion valuation heading into its recent IPO.

Bullish analysts have framed the concept as genuinely unprecedented: orbital compute that could either fuel SpaceX's own AI ambitions or function as a kind of "orbital neocloud," selling processing power the way terrestrial cloud providers sell server capacity today. It's a compelling story, and it's clearly one that public market investors have been willing to price in as a real, near-term value driver.

The problem, according to nearly every subject-matter expert who's actually run the numbers, is that the story is significantly ahead of the physics and economics required to make it real.

What The Experts Who Actually Do This Keep Saying

This isn't just skeptics on the sidelines. The doubt is coming directly from people building competing versions of the same idea.

Entrepreneurs behind other space data center startups, like Starcloud, which raised $170 million in a Series A specifically to build orbital data centers, and Cowboy Space, which raised $275 million pursuing a similar concept, have themselves acknowledged a brutal underlying constraint: there simply aren't enough rockets available right now to launch space data centers at any meaningful scale.

Google's own team, working on the company's orbital compute project (detailed in a technical paper the company published, sometimes referred to as the "Suncatcher" project), has arrived at similarly grounded conclusions about the timeline required.

And engineers who've independently run the economics, including some who published detailed breakdowns simply because the math itself is compelling to work through, keep landing in the same place: orbital AI computing doesn't meaningfully move the needle until two things happen simultaneously: rockets become dramatically cheaper to launch, and high-powered satellites become possible to manufacture cheaply and at real mass-production scale.

Right now, neither of those conditions exists. And critically, this isn't one skeptical voice, it's a converging assessment from Google's engineers, competing startup founders, and independent analysts, none of whom have an obvious incentive to specifically undermine SpaceX's narrative the way Altman does.

Why Starship Is The Actual Crux Of The Argument

Musk's implicit answer to all of this skepticism is Starship SpaceX's enormous new rocket, which is expected to attempt its 13th test flight as soon as July 16 (a launch we covered in detail as SpaceX works through the specific booster failure that grounded the program back in May).

The logic here is straightforward: if SpaceX can get Starship to the point where it's genuinely and routinely reusable flying, landing, refueling, and flying again, repeatedly, without extensive rebuild between launches, the fundamental cost structure behind orbital data centers could finally close. Cheap, frequent, high-capacity launches are the missing ingredient that every expert keeps pointing to.

But "if" is doing enormous work in that sentence. Even in the best-case scenario where Thursday's test flight succeeds completely, both stages recovered, booster re-ignition working properly, no repeat of May's Gulf of Mexico failure genuinely operational, routine reusable Starship flight is still likely years away, not months.

And even once that operational reusability exists, space data center launches would likely have to compete for launch capacity against SpaceX's existing, contractually binding commitments, specifically its NASA obligations and the ongoing buildout of its own Starlink satellite constellation, which we've covered as the company's actual current profit engine. Orbital data centers would be queuing up behind those priorities, not launching alongside them from day one.

The Admission Buried In SpaceX's Own IPO Filing

Here's the detail that makes Altman's criticism land harder than a typical rival's Twitter jab: SpaceX itself has already conceded significant uncertainty about Starship's reusability timeline.

During its IPO roadshow, SpaceX acknowledged that Starship may not achieve full reusability in the near term, and that the company will likely need to expend essentially throw away each second stage during every launch for the foreseeable future, rather than recovering and reflying it the way the booster stage is designed to be recovered.

That's a significant qualifier buried inside the company's own public disclosures, and it directly undercuts the economic case for space data centers. If every launch requires manufacturing and discarding an entire new second stage, the cost-per-launch math that makes orbital compute theoretically viable simply doesn't work; you're back to something closer to traditional, expensive, one-time-use rocket economics, just with a reusable first stage bolted on.

Musk's Response Doesn't Actually Answer The Question

Musk's public rejoinder to all of this, "We start flying them next year," is technically true in the narrowest possible sense, and functionally beside the point.

Nobody seriously disputes that SpaceX could launch a single satellite equipped for high-speed data processing sometime next year. That's not the hard part, and it's not what the skepticism is actually about.

The real question that every expert cited above keeps circling back to isn't whether SpaceX can launch one demonstration satellite. It's when SpaceX can launch and manufacture these satellites at a genuine commercial scale: dozens or hundreds of high-powered orbital data center satellites, launched frequently and cheaply enough to constitute an actual business, rather than an impressive one-off proof of concept.

And based on the current state of Starship's reusability, the expendable second-stage issue disclosed in SpaceX's own IPO materials, and the competing demand for launch capacity from NASA and Starlink commitments, that scale question is realistically pointing toward the 2030s, not next year, and probably not for several years after that.

Why This Matters Beyond A Twitter Spat

This dispute connects directly to broader questions we've been tracking about how public markets are pricing ambitious, unproven technology narratives right now, from SpaceX's own IPO valuation, to General Fusion's 40% debut-day stock surge despite a delayed breakeven timeline and recent layoffs, to the general pattern of public investors rewarding compelling technical stories well ahead of commercial or even technical proof.

SpaceX's roughly $2 trillion valuation is, by the company's own framing during its IPO process, substantially tied to the space data center opportunity not just its existing, profitable Starlink business or its established launch services. If that specific piece of the valuation story is running years ahead of the actual physics and economics required to realize it, as Altman and a genuinely broad coalition of space industry experts are suggesting, that's a meaningful gap between narrative and reality sitting inside one of the most valuable companies on the planet.

The fact that this particular critique is coming from Sam Altman, someone with obvious personal and competitive motives to undermine Musk publicly, doesn't actually make the underlying technical argument wrong. It just makes it easy to dismiss as personal sniping rather than engaging with the substance, which happens to align closely with what Google's engineers, competing space startup founders, and independent analysts have all separately concluded.

Sam Altman and Elon Musk's weekend social media fight produced, almost incidentally, a genuinely useful public airing of a real industry disagreement: whether space data centers are a near-term business opportunity or a 2030s-and-beyond proposition still waiting on foundational breakthroughs in rocket reusability and satellite manufacturing costs.

The expert consensus spanning competing startups, Google's own orbital compute team, and independent engineers who've run the numbers lines up much closer to Altman's skepticism than to Musk's "we start flying them next year" optimism.

Starship's upcoming test flight, expected as soon as July 16, matters enormously here, but even a fully successful flight doesn't resolve the deeper issues: SpaceX's own acknowledged uncertainty about full reusability, its second-stage expendability admission from its IPO filing, and the queue of existing NASA and Starlink commitments that will likely take priority over speculative data center launches.

None of that means space data centers are a bad idea, or that SpaceX won't eventually get there. It means the timeline currently priced into SpaceX's valuation may be running well ahead of what the physics, the manufacturing capacity, and the company's own disclosed roadmap actually support.

Sometimes, the most useful business analysis comes from two billionaires calling each other scammers on social media. This was apparently one of those times.