Elon Musk's xAI Is Failing

Elon Musk is planning to take SpaceX public this summer at a reported valuation of $1. 75 trillion. SpaceX on its own only generated $18 billion of revenue last year, which doesn't even come close to justifying this valuation. In an attempt to inflate SpaceX's valuation, Musk is trying to capitalize on the AI bubble. In early 2026, he merged SpaceX with XAI. XAI owns a social media platform, which used to be called Twitter, as well as the Grock AI chatbot. Musk has made bold claims about the current and future capabilities of the Grock AI chatbot, claiming that it will soon achieve artificial general intelligence and be worth trillions of dollars. The problem is, Grock is not very useful and almost nobody uses it. In 2024, XAI borrowed tens of billions of dollars to build a massive 300 megawatt data center in Memphis, Tennessee. Because so few people use Grock, this data center is mostly sitting idle. In April, The Information reported that the Colossus data center has a utilization rate of just 11%. In May of 2026, XAI signed a deal with the rival AI company, Anthropic, to lease the Colossus data center. This was a shocking reversal. Just a few months ago, Elon Musk publicly accused Anthropic of being racist, sexist, evil, and criminal. More importantly, now that XAI is leasing out the entirety of its main data center, how is it going to train and run its Groc AI models? How can XAI even pretend to be a serious player in the AI industry after this? In this video, we'll look at the pathetic failure of XAI, and what this means for SpaceX's upcoming IPO. Before we get into the video, there's a quick announcement I want to make. In addition to Wall Street Millennial, we also have a second channel called Broken Business Models. There we make unscripted videos about companies we believe to be fraudulent or overvalued, as well as various topics that didn't make the cut for the main channel. If that at all sounds interesting to you, make sure to check out Broken Business Models, which you can find linked in the description below.

Elon Musk founded XAI shortly after the release of Chat GBT. XAI released his chatbot, Grock, in November of 2023. Musk was quick to make grandiose claims about Grock. In mid 2025, he claimed that it was smarter than PhD students in every single academic field. In other words, they trained it on a bunch of PhD material and Grock memorized the answers. In reality, Grock's use cases are very limited. It is heavily integrated into the X social media platform. If you go on X, you'll quickly see that Grock's main use case is AI generated porn. Elon Musk is desperate to monetize Grock in any way that he can. They launched an AI anime Girlfriend, but most people view this as cringe and very few people are willing to pay for it. Musk saw the success of Chat GBT and wanted to get in on the AI bubble, but it doesn't seem like there's any overarching business plan. In February of 2026, Musk went on the Stripe podcast where he failed to explain what XAI's competitive advantage is. — So, what is XA's plan to win? — You should tell you on a podcast. — Yeah. Spill all the beans. Have another Guinness. — It's a good system. — People sing like a canary. Um, all the secrets, but it's just — Okay, but in a nonsecret spilling way, what's the plan? — What a hack. — Well, when you put it that way, um, I think the way that Tesla solved uh, self-driving is the way to do it. So, I'm pretty sure that's the way. — It sounds like you're talking about data like Tesla driving because of the — we're going to try data and algorithms. — But isn't that what all the other lines are trying like what's if those don't work I'm not sure what work we've tried data we've tried algorithms but — uh yeah — it seems like a competitive field and I'm like what is how are you guys going to win is like my big question. I I you know I I think we see a path to doing I mean I think I know the path to do this because it's kind of the same path that Tesla used to create self-driving. Um you know instead of driving a car it's driving a computer screen. So Musk has a super secret plan of how XAI will create godlike super intelligence and he's confident in his abilities because Tesla was able to make full self-driving. Tesla has been promising full self-driving next year for the past 10 years, but to this day, they still only have level two autonomy, which requires constant monitoring by a human driver because it makes mistakes so often. Musk has a long history of lying about Tesla's autonomous capabilities, and the company has already had to pay hundreds of millions of dollars worth of legal settlements. Currently, Tesla is offering limited robo taxi services in geofenced areas, but they have to use human safety drivers. Earlier this year, they pretended to remove the human safety driver. In reality, they had the human safety driver following right behind the robo taxi to monitor it. In that same podcast interview, Musk fails to explain XAI's business strategy. What will XAI's business be like? Is it going to be consumer enterprise? What's the mix of those things going to be? Is it just going to be similar to other labs where Yeah. This — you're saying labor makes sense — corporations. — Corporations goes to DVON. — Revenue maximizing corporations to be clear. — Those GPUs don't pay for themselves. — Exactly. Um but yeah, what's the business model? What are the revenue streams in a few years time? — Um I think things are going to change very rapidly. Like I'm staying the obvious here. — Um you know I call AI the supersonic tsunami. I love iteration. Um so really what's going to happen is especially when you have humanoid robots at scale um is they will just provide they will make products and provide services far more efficiently than human corporations. — Musk cannot give a straight answer to even the most basic questions. Will XAI focus on the consumer or enterprise markets? What are its major revenue streams? Musk responded with a bizarre word salad about supersonic tsunamis. In reality, there is no secret master plan. Musk is in way over his head. A few months ago, Bloomberg reported that in the third quarter of 2025, XAI generated $17 million of revenue and incurred a net loss of $1. 46 billion. That's the most recent financial data we have for the company. Musk is not happy about this performance. In March of 2026, he tweeted that XAI was not built right the first time around, so it is being rebuilt from the foundations up. This rebuilding involves mass layoffs. According to the Financial Times, Musk has been frustrated by slow progress in XAI's model development with performance lagging far behind competitors like OpenAI, Anthropic, and Google. So, Musk started firing large numbers of researchers, including very senior members of XAI's founding team. The staff turnover at XAI has reached the point of being almost absurd. In February, XAI hired a man named Toby Palon, who was previously a researcher at Google's Deep Mind. He was hired for a very senior position at XAI. He was in charge of a project that was supposed to develop AI agents for software development. Musk fired him after just 16 days on the job. Musk has unrealistic expectations, and the firing seemed to be almost random. Many of XAI's top researchers quit to join rival AI startups because XAI's work environment is so toxic.

In 2024, XAI built what Musk claims is the world's biggest AI data center. It has 220,000 Nvidia GPUs that consume 300 megawatts of electricity. They call this data center Colossus. This costs tens of billions of dollars to build. When SpaceX acquired XAI, it also acquired all of XAI's debt. SpaceX reportedly now has $20 billion of debt. Probably most of this came from the Colossus data center. So now SpaceX has a massive data center in Tennessee, which reportedly only has 11% utilization. XAI does not generate nearly enough revenue to cover the data center's interest expense and operating costs. So in May of 2026, XAI agreed to lease the entire data center to Anthropic. The terms of the deal are not disclosed, but it is believed that Anthropic will be paying XAI billions of dollars per year. This deal is surprising for a number of reasons. Just a few months ago, Musk went on a vitriolic Twitter rant against Anthropic. In February, he said that Anthropic's AI hates whites, Asians, especially Chinese, heterosexuals, and men. This is misanthropic and evil. That sounds pretty bad. So, how did Anthropic become so evil? Anthropic becoming evil was not due to any conscious decision on the part of the company's leadership. It was more of an accident. According to Musk, it is inevitable that every company becomes the opposite of its name. When they chose the name Anthropic, it was inevitable that they become misanthropic. They should have named their company missanthropic so they can become anthropic instead. Keep in mind that Musk has a long and welldocumented history of substance abuse. This can likely explain most of his bizarre and nonsensical tweets. But Musk also has some more legitimate criticisms of Anthropic. He accuses Anthropic of being guilty of stealing training data at a massive scale. This is true. Last year, Anthropic agreed to pay $ 1. 5 billion to settle a class action lawsuit brought by authors whose books were used to train Anthropics AI models. Anthropic's copyright violations are a legitimate criticism, but this is a case of the pot calling the kettle black. In December of 2025, XAI was sued by another group of authors alleging copyright infringement. At the same time, XAI is suing to try to stop a California state law, which would require it to disclose its training data. They're probably trying to cover up the fact that they too are using massive amounts of copyrighted texts to train Grog. SpaceX reportedly lost $5 billion in 2025. This is pro- fora to include XAI's financial results. Musk is desperate to improve SpaceX's financial performance going into the company's much anticipated IPO. Leasing the Colossus data center to Anthropic is one of the only ways for SpaceX to bring in more revenue. Musk is happy to accept Anthropic's money regardless of how evil he used to claim the company was.

In public, Musk likes to hype up the capabilities of XAI. Remember in the Stripe podcast, he claimed to have a secret plan which will allow XAI to out compete its rivals. In 2025, he said that Grock was on the verge of surpassing both OpenAI and Anthropic in terms of capabilities. In the ongoing Musk Open AI trial, Elon Musk is forced to testify under penalty of perjury. What he truly believes is very different from what he tells to the public. In the trial, Musk was asked how XAI's current capabilities compare with OpenAI and other rivals. Musk said that Anthropic is currently in the lead. Open AAI is second, followed by Google and then a number of open-source Chinese models. XAI is far down the list. He characterized XAI as a much smaller company with just a few hundred employees. In January, Musk claimed that XAI built a new AI data center called Colossus 2. He claimed that Colossus 2 is the world's first 1 gawatt AI data center. However, satellite images show that it only has cooling capacity for 350 megawatts of GPUs, making it only slightly bigger than Colossus 1. Elon Musk previously claimed that XAI's demand for computing capacity is so great that they'll have to launch 1 terowatt of AI data centers into space. But if that's true, why are they giving up half of the computing capacity they already have on Earth? In April 2026, SpaceX agreed to acquire the AI coding startup Cursor for $60 billion. Technically, they only purchased the option to acquire Cursor, but they have to pay a $10 billion breakup fee. If they don't, the acquisition will be paid for in SpaceX stock rather than cash. Curser operates an AI native integrated development environment, which is popular with software engineers. In the past, Cursor did not develop its own AI models. They hooked into the Open AI and Anthropic APIs. They basically purchased tokens from OpenAI and Anthropic and resold them to end users for a lower price, thus resulting in negative gross margins. Recently, they launched their own AI model called Composer 2. Cursor did not develop Composer 2. They licensed an openweight model called Kimmy K2. 5 from a Chinese company and rebranded it as Composer 2. Because Kimmy 2. 5 is open weight, Curser can run it on its own data centers. But Curser doesn't have This is why they agreed to be acquired by SpaceX. They want to use XAI's Colossus data centers to run the Kimmy K 2. 5, which they hope will be cheaper than paying for ChatGpt or Clawed APIs. XAI only has two data centers, Colossus 1 and Colossus 2. The entirety of Colossus 1 is now being leased to Anthropic. Once SpaceX acquires Cursor, Cursor will presumably use Colossus 2. If Anthropic is using Colossus 1 and Cursor is using Colossus 2, that doesn't leave a lot of space for XAI to train its own Grock models. It looks like Elon Musk has finally given up on Grock. He realizes that he's very far behind his competitors. Grock will never be a commercial success. To the extent that XAI builds any more data centers, they will likely rent them out to their competitors. Of course, Musk will never admit this. He will continue to pretend that he has a secret master plan to turn Grock into a god-like super intelligence. He needs his gullible fanboys to believe this so they buy shares in SpaceX's IPO. All right, guys, that wraps it up for this video. What do you think about XAI? Let us know in the comments section below. As always, thank you so much for watching and we'll see you in the next one. Wall Street Millennial signing