Anthony Armstrong stepped down today, or so the word is from the former Morgan Stanley banker who Musk brought on as CFO just six months back, according to The Information. There’s been talk that his parting message was simply a blunt “This is over,” which really seems to put a sharp point on a string of executive exits that you just don't see in the AI world these days. By the time we reached late March, every single one of the original co-founders who wasn't Musk had moved on. Ross Nordeen, who acted as Musk’s right-hand man for operations, posted about “touching some grass” before he went silent. Then back in February, there was that one wave that took out nine of the more senior engineers and research staff in just a few days. Musk has framed the whole situation as a necessary rebuild, but whether you buy that or not, the fact that Armstrong is gone shows this overhaul is hitting the financial side of the house. These were the people brought in specifically to figure out how to bridge the gap between the cash xAI is burning through and what it's actually managing to bring in.
A Trillion-Dollar Ambition Sitting on an Unproven Revenue Core
The numbers aren't exactly a secret. Bloomberg’s reporting indicates that xAI burned through $7.8 billion in the first nine months of 2025 alone. They posted a loss of $1.46 billion in just the September quarter, which looks pretty massive when you put it next to their $107 million in revenue. At the height of it, they were seeing monthly losses that were pushing toward $1 billion. By the end of 2025, the company had built up a fleet of more than a million H100-equivalent GPUs across their Colossus clusters, which is essentially a massive pile of capital that assumes a future where AI works like a utility. Management is guiding for $2 billion in revenue this year, but when you look at the competition, they’re moving at a different scale. OpenAI has already passed $25 billion in annualized revenue, and Anthropic is reporting a run-rate of over $30 billion with more than a thousand customers spending at least a million dollars a year.
It’s true enough that xAI is spending too much, but focusing only on that tends to miss the deeper problem. The real issue is the structural path they've taken. They’re trying to act like a frontier research lab, a massive infrastructure builder, and a product distribution company all at the same time. The trouble is that they haven't shown they can dominate even one of those areas on its own yet. In spite of that, they’ve already locked themselves into a cost structure that you’d only expect to see in a company that has already won the market. Having that kind of mismatch between where you are and what you’re spending is a whole lot riskier than the kind of cash burn you'd see at a typical startup.
Where the Money Is Moving — and Where xAI Is Not
The AI market as a whole is going through a shift. The real profits are starting to settle in areas like enterprise spending, agents designed for coding, and tools that sit directly inside existing workflows. Axios has noted that Anthropic managed to capture 73% of the money coming from companies that were buying AI tools for the first time. Meanwhile, OpenAI is trying to reach $2.5 billion in revenue from advertising so they can stop relying so much on subscriptions. Both of these rivals are having to pivot their business models because they've realized that the sheer cost of running these models is too high to be covered by simple subscriber growth.
Back in March, Musk was open about his frustration that xAI’s work on coding tools just wasn't holding its own against things like Claude Code or Codex. This actually matters quite a bit because coding is one of the more obvious ways to get into actual corporate budgets. Business Insider reported that one of the senior leaders told the staff that they were “clearly behind” the rest of the pack. If that gap doesn't get filled, all that computing power they’ve built up might look impressive in a data center, but it isn't going to do much to generate a real stream of cash.
The SpaceX Merger Makes xAI Safer and Harder to Analyze
Then there’s the fact that SpaceX officially acquired xAI in February. That deal put the combined value at about $1.25 trillion, with xAI’s portion sitting around $250 billion. Already, SpaceX has started filing for an IPO where they’re aiming for a $2 trillion valuation. What this merger really does is give xAI a much deeper pool of capital and more leverage when it comes to procurement. The vision is to weave Grok into everything from X and Tesla to Starlink and various defense projects. They’ve also already secured a $200 million contract with the Department of Defense and are moving into other government work.
Of course, the downside of the merger is that it tends to hide the actual economics of the business. It allows for a lot of internal support and lets the management team put off the hard pressure of actually becoming a profitable company. While that might make xAI feel like a more stable entity, it makes the case for investing in it a lot more complicated. Over in Memphis, their Colossus 2 site is currently facing the threat of lawsuits over air pollution, which is just a reminder that building AI infrastructure now involves the messy reality of power politics and local permits. These aren't just things that hurt your reputation; they’re real costs that hit the bottom line.
The Investor Conclusion
In a sense, xAI has built the factory, but they haven't yet proven that they have the right to operate it at this scale. There’s revenue coming in, sure—mostly from things like subscriptions, APIs, and government work—but we still don't have much clarity on the quality of that money. It’s difficult to tell how much of it is actually high-margin and profitable, or if it’s just relying on the fact that Musk is propping it up through his other companies. The merger with SpaceX is likely to make those questions even harder to answer.
The fact that the original founders have left is a more serious issue than it might look on the surface. In AI, those early technical leaders are the ones who carry the intuition and the history of how the models were actually put together. You can’t just replace that by moving boxes around on an organizational chart. Losing that kind of talent at the same time the company is admitting it’s behind on coding is really a question of whether they can execute on their goals. xAI might get better at the mechanics of shipping products, but they risk losing their creative sense of direction and the trust of enterprise clients who need reliability.
Seeing Armstrong leave today is the clearest sign yet that the real pressure on the company is financial. xAI is going to have to find an answer to a very specific question before their next cluster goes live: can they find a way to make the revenue catch up to the scale of their ambition? Until that happens, every new headline or milestone in computing power is probably best seen as just a bigger bet on something that hasn't paid off yet.
not investment advice
Sources: https://www.theinformation.com/articles/spacex-woos-investors-c-suite-shakeups-continue