Elon Musk is betting his business empire on AI

THOSE WHO believe in Elon Musk are convinced both by his vision to go where no one has ventured before and his ability to pay for it—what some call the “Elon backstop”. Mr Musk’s announcement on February 2nd that he will merge SpaceX, which builds rockets and sells satellite broadband, with xAI, his artificial-intelligence lab, was not short of ambition. The world’s richest man declared that the new company would “extend the light of consciousness to the stars”. Back on Earth, however, it is getting harder to see how Mr Musk’s numbers add up.

The transaction values the new entity at $1.25trn; investors in SpaceX will be entitled to 80%, with the remainder going to xAI’s owners (Mr Musk holds a controlling stake in both). The stated rationale behind the tie-up is that the companies will work together to launch a fleet of data centres into space, giving xAI a big advantage in the race to develop cutting-edge models while furnishing SpaceX with a new line of business. More immediately, combining the two could further boost interest in a public listing expected this summer.

By bringing SpaceX and xAI together, however, Mr Musk is saddling a money-spinning space champion with a loss-making AI laggard. At the same time, he is reshaping Tesla, the carmaker he runs, into a “physical-AI company” focused on self-driving taxis and humanoid robots. If the latest wave of AI proves as transformative as some expect, these bold gambles might just pay off. If not, Mr Musk’s business empire could well be in jeopardy.

Start with the mega-merger. SpaceX is a jewel. In 2025 it launched close to 4,000 satellites into space, accounting for about 85% of the global total for the year (see chart 1). It can send objects into orbit far more cheaply than any competitor. Starlink, a satellite-broadband service that is its primary source of revenue, has around 9m subscribers worldwide, according to Deutsche Bank, more than triple the number two years ago. The company also has lucrative government contracts. All together, it reportedly generated as much as $16bn in revenue in 2025 and around $8bn in operating profit (before depreciation and amortisation).

It is a far different picture at xAI. The AI lab last year made on the order of $500m in revenue from its Grok models; OpenAI, maker of ChatGPT, brought in about $13bn. X, the social-media platform with which xAI merged last year, brought in perhaps an extra $3bn in sales. Even so, the business as a whole is reportedly bleeding cash at the rate of about $1bn a month as it invests vast sums in data centres.

The company brings with it other problems, too. X is under investigation in the European Union and Britain over potential breaches of data regulations and for its launch over Christmas of an image generator that was widely used to produce nude deepfakes, including reportedly of children; on February 3rd its offices in Paris were raided by French authorities. Mr Musk has denied that the company has done anything wrong. If courts find otherwise, the EU could fine it up to 6% of its global revenue, while Britain could fine it up to 10%.

Then there are its various debts. Last year xAI borrowed $5bn to fund its data-centre binge. With Valor Equity Partners, a long-time backer of Mr Musk’s ventures, it has also set up an off-balance-sheet vehicle financed by around $3.5bn of debt in order to buy yet more AI chips. The merger with X last year left the AI lab with another $12bn or so of borrowings remaining from Mr Musk’s purchase of the social network. SpaceX, for its part, is on the hook to cover $2bn of interest owed by EchoStar as part of a deal last year to acquire mobile spectrum from the struggling satellite company. These combined obligations will strain the business at a time when xAI continues to operate at a loss and SpaceX is investing heavily in its new “Starship” launch system, which is behind schedule.

A shot of equity from a public listing would help ease the burden. The merged company reportedly has plans to raise $50bn at a valuation of at least $1.5trn. It is a lofty sum even by Mr Musk’s standards. Tesla is valued at $1.5trn, but last year generated $95bn in sales—around five times as much as SpaceX and xAI combined. Some fuddy-duddy institutional investors will balk at the price tag. Others will be turned off by the association with Grok. But they will not be Mr Musk’s target audience. His pitch will no doubt present data centres in space as a prelude to factories on the Moon and cities on Mars. Retail investors will lap it up.

Mr Musk’s willingness to merge SpaceX with xAI shows how committed he has become to dominating the AI industry. It is personal: he loathes Sam Altman, the boss of OpenAI, which Mr Musk co-founded and is now suing for abandoning its original non-profit structure. Mr Altman is also eyeing a giant listing this year.

In theory, using SpaceX to build orbital data centres might help Mr Musk gain the upper hand over his rival. Mr Altman has reportedly sought to acquire, team up with or build a space company to rival SpaceX, and has long mused about the benefits of space-based data centres. Google, a search giant whose Gemini model competes with both ChatGPT and Grok, is planning to send a test satellite containing its in-house AI chip into orbit in 2027.

Mr Musk is eager to get a head start. On January 30th SpaceX filed a request to the Federal Communications Commission, an American regulator, to put a 1m-strong constellation of satellite-based data centres into orbit. Mr Musk argued that within two to three years, the cheapest place to provide computing capacity would be in space, by harnessing solar power that is undiminished by the atmosphere. Starlink satellites can then beam the data back down to Earth.

Yet much is still to be proven. The main question, says Sir Peter Beck, founder of Rocket Lab, a smaller rival to SpaceX, is which is cheaper: the cost of electricity on Earth, where power is scarce, or the launch costs to get to space, where energy is abundant? For now, the latter are prohibitive. In a study last year, researchers at Google said the launch cost per kilogram was not likely to fall to a level equivalent to the cost of running terrestrial data centres for at least a decade. A significant increase in computing capacity will be required for xAI well before that.

There are also a number of technical hurdles. Orbital data centres will need large radiators for cooling, and cosmic rays could damage equipment. Chris Kemp, the founder of Astra, another rocket company, notes that AI chips tend to quickly become obsolete and need replacing. “You are going to have to refresh your satellites every few years, which further compounds the problem,” he says.

Tesla, which in the past few years has stopped guzzling cash, could be enlisted to help. On January 28th the carmaker stated that it had invested $2bn in xAI. The two companies are increasingly sharing software, data and chips. Some speculate that Tesla could even be folded into the rest of Mr Musk’s empire, though doing so would be complicated by the fact that he does not own a controlling stake in the company, and as its chief executive recently secured a pay deal worth up to $1trn that may be put in doubt by a merger.

Even so, Mr Musk is steering his carmaker straight into the AI hype-cycle. It will soon cease making its Model S, its first mass-produced electric car, and Model X, its gull-winged SUV. Together the two models accounted for just 2% of Tesla’s vehicle production in 2025. More telling was the fact that the factory space currently devoted to them will be repurposed to manufacture Optimus, Tesla’s humanoid robot. Mr Musk has set a target of making 1m of them a year by the end of 2027. At the same time, Tesla is investing heavily to develop its Cybercab, a two-seater self-driving taxi that is set to go into full production in April. Mr Musk has declared that by the end of this year his robotaxis will have broken free from a handful of testing areas and be available for up to half the population of America.

In reality, self-driving taxis and humanoid robots will take years to mature into cash-generating businesses. In the meantime, they will require vast investments just as Tesla’s core business is stalling. Vehicle sales in 2025 fell by 9%, the second year of decline (see chart 2). In Europe they plunged by a quarter. Some buyers have been turned off by Mr Musk’s political antics. The deeper problem, though, is stiffening competition in electric vehicles, both from legacy carmakers and Chinese newcomers. Tesla’s remaining line-up, starved of investment, will continue to become less attractive.

As with SpaceX, then, Mr Musk is gambling Tesla’s future on AI, and his belief that he can harness his existing businesses to dominate the technology. Plenty of doubters have scoffed at his grand ambitions before. But never has Mr Musk put so much on the line. ■