Elon Musk's Grok heads to space

SpaceX’s brand is physics: unforgiving, audited, checklist-deep — the kind that punishes improvisation. Grok’s brand is the internet: reactive, unregulated, frequently allergic to consequences — the kind that rewards it. Now, SpaceX is buying xAI, the company behind Grok, and folding their reputations together is less a moonshot than a wager that the most regulated kind of company can safely adopt the least regulated kind of product.

CEO Elon Musk, never one to make a deal announcement without a little cosmic flair, called it the “next book” in a mission that involves “scaling to make a sentient sun.”

The mechanics are clean, which is exactly why the implications aren’t. SpaceX is taking xAI in a transaction that pegs SpaceX at $1 trillion and xAI at $250 billion, with xAI holders getting 0.1433 share of SpaceX per share (or, for some, cash at $75.46). Yes, that’s as enormous as it sounds; it’s a record-setting deal, eclipsing the long-standing “largest M&A” crown jewel. And that’s another consolidation move by Musk, who has already shown a fondness for corporate nesting dolls; last year, xAI absorbed the social platform X $TWTR via a share swap.

This isn’t a normal tech tie-up where the biggest risk is a Slack $WORK migration. SpaceX is a launch provider, a satellite operator, and a defense contractor with billions in federal contracts — the sort of company that gets measured in clearances, audit trails, and “please don’t make us brief this to Congress.” Those government relationships give agencies authority to review M&A transactions for national security and other risks, and the transaction invites extra scrutiny around governance and conflicts, given Musk’s overlapping leadership across his companies.

Tesla $TSLA is already in the blast radius. The company disclosed a $2 billion investment in xAI even as it’s ramping spending for its own AI ambitions — planned capex above $20 billion this year, per its chief financial officer. Longtime Tesla bull Dan Ives put the empire logic bluntly in a Tuesday note from Wedbush: “If you’re trying to build robots, and build autonomous cars, and build rockets, these things all fit together.” The fit may be real. The financing tab will be, too.

And one ugly Grok scandal can travel faster than a rocket and land in exactly the wrong place — a procurement office, a regulator’s docket, an oversight hearing. A space-and-defense contractor can survive a product controversy, but it survives by being boringly compliant, not by improvising safety policies while the headline churns.

SpaceX has been preparing for a public offering that could land this year and aim north of a $1.5 trillion valuation, with fundraising chatter reaching into the tens of billions. xAI, for its part, has been guzzling capital the way frontier-model companies do, including a $20 billion Series E round at a $230 billion valuation — one report put its burn at around $1 billion a month — while trying to chase rivals with bigger distribution and deeper cash engines such as OpenAI and Alphabet $GOOGL.

SpaceX brings Starlink, global connectivity, and a hardware backbone that makes “AI everywhere” feel less like a product strategy. And for an IPO-bound company, adding the word “AI” to the story can turn a launch business into a platform business, and a platform business into a religion. A rocket company selling launches and satellite internet has financial gravity; an AI company selling subscriptions and dreams has narrative gravity. Put them together and you’ve built a valuation engine that can, in theory, run on belief for a very long time.

If SpaceX’s core value proposition is “we can do hard things reliably,” then importing Grok’s current baggage is a choice to be less of that. Maybe the market will pay for it anyway. The market has certainly paid for worse. And in the meantime, the best (maybe) rocket company in the U.S. just adopted the hungriest (definitely) hobby in Silicon Valley.

Musk’s roadshow heads into orbit

The strategic pitch writes itself: AI needs power; power is scarce; space has sunlight; therefore, data centers… in orbit. 

In an FCC filing late last week, SpaceX asked for permission to build an “orbital data center” system of up to 1 million satellites designed to “harness the sun” to power AI data centers, arguing that solar-powered computing in space could scale faster and cheaper than terrestrial data centers choking on land fights, permitting timelines, cooling, and local politics. 

On paper, that looks like a neat little Muskian inversion: If the bottleneck is Earth, then quit Earth. Musk has said he expects the most cost-effective AI compute to be in space within two to three years.

But in practice, the filing reads less like a well-thought-out construction plan than a starting bid.

SpaceX has a history of asking regulators for headline numbers that create room to negotiate down later, and orbital real estate is already crowded enough to make “one million” feel like a dare. There are about 15,000 satellites currently in orbit, and SpaceX’s existing network is roughly 9,500 satellites. SpaceX has asked for big numbers before; it sought approval for 42,000 Starlink satellites ahead of deployment.

Back down on Earth, analysts at MoffettNathanson did the math and concluded that sustaining a million-satellite constellation with five-year lifespans would require launching about 200,000 satellites annually — a pace that turns “ambitious” into “logistics problem,” and “logistics problem” into “capital markets event.” MoffettNathanson also flagged that the capital demands could be so large that SpaceX might have to tap public equity markets. The orbital compute dream is being stapled to the IPO because the orbital compute dream needs IPO money.

Deepwater Asset Management investor Gene Munster, who has cheered the “datacenters in the sky” idea, pointed to two advantages in a video posted to X: better solar access and less need for cooling. But that tidy little sentence skips past the part where space hardware has to survive radiation, operate without convenient human maintenance, and get replaced on schedules dictated by physics, not product roadmaps. If this is the future, it’s a future with a lot of rockets in it.

Still, the point isn’t whether SpaceX can launch one million satellites. The point is that SpaceX wants investors picturing AI capacity as something it can manufacture the way it manufactures launches — industrially, repetitively, at scale.

Wall Street, predictably, can’t resist a grand narrative when it comes with rockets. AJ Bell’s Russ Mould wrote that “others may buy into Musk’s grand vision of data centers in the cosmos,” and that may “whet the appetite” ahead of what could be the largest IPO ever. And Seraphim Space CEO Mark Boggett called the merger “the strongest validation yet that space will be the backbone of the next wave of AI.”

To be fair, this merger does glue together two things that actually do fit. SpaceX is an infrastructure beast with a distribution system (Starlink) and a rocket roadmap built around getting mass to orbit at scale. XAI is in the business of turning compute into models and models into… more compute. 

There’s also a more practical pitch, the one analysts like because it has nouns you can model. Ali Javaheri, a senior emerging-spaces analyst at PitchBook, wrote, “Starlink was already a cash flow engine and now it adds an AI revenue layer on top while also becoming a distribution surface for AI services and data.” That’s the cleanest argument for bundling the businesses: a built-in pipe to deliver AI services, plus a torrent of user data, plus the ability to launch infrastructure without negotiating with a hundred local utilities and zoning boards.

So yes, “orbital compute” is a pitch. It’s also a permission structure. It makes xAI’s burn rate feel like an investment in a platform rather than a bonfire in a basement. And it recasts a merger that looks, financially, like a bailout into something grander: an infrastructure strategy with rockets as the ultimate workaround. But folding a consumer chatbot into a rocket contractor means SpaceX inherits the chatbot’s problems — not abstract, philosophical “AI ethics” problems, but active legal and regulatory ones with names, deadlines, and prosecutors who don’t care how cool Starship’s latest and greatest rockets look.

A rocket company inherits internet problems

The orbital storyline can’t laser-link away the fact that Grok isn’t a neutral asset. Grok is a consumer-facing AI product plugged into a social platform that has spent the last few years making regulators learn new headache symptoms.

In January, a bipartisan coalition of 35 state attorneys general urged xAI to do more to prevent Grok from generating nonconsensual intimate images and child sexual abuse material, according to a release from Pennsylvania Attorney General Dave Sunday’s office and a separate release from New York Attorney General Letitia James. In California, Attorney General Rob Bonta sent a cease-and-desist letter demanding xAI halt the generation and distribution of nonconsensual sexual images via Grok; xAI said it had implemented new restrictions, while scrutiny was spreading internationally. 

On Tuesday, Britain’s Information Commissioner’s Office opened a formal investigation into Grok over personal-data processing and its potential to generate harmful sexualized content. The regulator said the reported creation and circulation of non-consensual sexual imagery — including involving children — raises serious concerns under UK data-protection law. Separately, Ofcom has its own investigation into X under the Online Safety Act framework. In France, prosecutors raided X’s offices as part of an investigation into those same behaviors and requested “voluntary interviews” with Musk and former CEO Linda Yaccarino. There’s growing pressure from EU regulators as well. 

This is what an “active controversy pipeline” looks like when it’s one million statues in a trench coat.

Now, zoom out and put that controversy inside SpaceX’s world. 

SpaceX isn’t a normal consumer-tech company that can treat moderation blowups as reputational potholes. It’s a major government contractor with billions in federal contracts with NASA and the Department of Defense, among others — the kind of relationship where governance questions don’t stay theoretical. There are, then, tons of regulatory scrutiny risks tied to conflicts of interest and the sensitivity of SpaceX’s contracts.

There is, of course, at least one more twist, because Musk never makes anything simple. XAI has a contract worth as much as $200 million to provide Grok products to the Pentagon, and U.S. Defense Secretary Pete Hegseth said Grok would be integrated into military networks as part of an AI acceleration strategy. That’s the upside case in its sharpest form: Grok graduates from “internet goblin” to “defense tool,” and SpaceX sells itself as the infrastructure backbone of next-generation national security.

The problem is that the market will be asked to believe both stories at once. Grok as a defense-grade asset. Grok as a consumer-facing model that has already triggered cease-and-desist demands and multi-country investigations. SpaceX as a pristine engineering brand. SpaceX as the corporate parent of a product that regulators keep describing in language no IPO roadshow wants to repeat.

Grok’s myriad — and incredibly serious — reputational risks don’t necessarily guarantee a catastrophic outcome for SpaceX. But it changes the risk profile in a way that’s hard to hand-wave away. A company preparing for an IPO can usually pitch around losses and capex if the growth story is clean. It’s much harder to pitch around a consumer product that arrives with an ever-growing and ever-spiraling controversy pipeline, especially when the parent company’s customer list includes governments that are famously cautious about reputational shrapnel.

SpaceX has always sold trust. It sells, “We will put the payload where we said we’d put it — and where you paid us ungainly amounts to do so” to commercial customers, and it sells “We will not embarrass you” to governments. Grok sells immediacy. It sells, “Ask me anything (and we do mean anything)” on a platform that treats virality as a feature, not a bug. If you’re pricing the combined company for an IPO, you now have to underwrite both.

Because the IPO story, if it comes this year, won’t just be about launch cadence and satellite economics. It’ll be about legal exposure, regulatory patience, and whether “move fast” culture can coexist with “don’t break national security clearance” reality. It’ll be about whether Grok’s next scandal lands as a platform issue, an xAI issue, or — now that the corporate line is one line — a SpaceX issue.

This is the bet, stripped of romance: that the market will treat Grok’s liabilities as temporary turbulence, while treating orbital compute as inevitable runway. SpaceX has pulled off bigger engineering feats than reputational integration. But engineering has a gift that governance doesn’t: Physics is consistent. Regulators, investors, and the internet are not.