# Musk's AI Gambit: XAI Struggles, SpaceX IPO, and Orbital Data Centers

**Podcast:** FT Tech Tonic
**Published:** 2026-05-27

## Transcript

Before we start this episode of Tectonic, we want to know what Tectonic listeners think of the show, so we've put together a short survey.
It's your chance to give us feedback and help shape future episodes.
And doing the survey will also give you the chance to win a high-end pair of Bose headphones.
There's a link in the show notes of this episode, plus all the T's and C's.
Okay, on with the show.
Well, Musk has always been a bit of a doomer in the sense that he thinks AI.
could prevent the end of humanity.
This is the FT's San Francisco Bureau Chief, Stephen Morris, talking about Elon Musk.
He's a big science fiction fan.
He always references Ian Banks' culture novels as kind of the utopian idea of the future of AI, where machines create super abundance and the rest of the humans just live for parties.
But he also likes to talk about Terminator, the flip side, Skynet, you know, deciding that the easiest way...
to prolong its existence is by removing humanity.
And he uses these kind of pop culture science fiction references a lot.
Musk has been warning about the dangers of AI for years.
But at the same time, despite those fears, he's been determined to be a central player in the creation of the technology.
He was an early investor in Demis Hassavis' DeepMind and one of the original founders of OpenAI.
But with Musk, there's always an inherent contradiction.
You know, I was speaking with Sebastian Malaby, who has written a book about Demis Asavis.
And I said, you know, you've spoken to Elon a lot, you know, Demis.
If Musk is so worried about AI, like, why is he right out there on the frontier trying to build one of the biggest, most dangerous models?
And he said, well, as far as I can tell, like, Elon is both terrified of this technology, but also terrified about not being in the vanguard of the next big thing.
You know, so is electric.
cars, it was autonomy, it's spaceflight rockets.
AI is the next frontier and Musk just cannot bear to like not be there himself.
Elon Musk was quick to recognize the significance of AI, but he always seemed to find a way of missing out.
By the time ChatGPT came along, DeepMind had long been sold to Google.
Sam Altman was the boss of OpenAI and Musk was out in the cold.
But now he has his own AI lab, XAI.
part of his giant SpaceX company, which Musk is hoping to take public in the largest IPO in history.
AI is at the heart of SpaceX's future plans, putting data centres into orbit, and, in the words of the IPO documents, the emergence of new trillion-dollar markets on the Moon, Mars, and beyond.
So can the world's richest man make up for lost time, take AI into space?
and put himself back at the vanguard of artificial intelligence.
This is Tectonic from the Financial Times.
I'm Murad Ahmed, the FT's technology news editor.
A handful of Silicon Valley companies are vying to lead the world in artificial intelligence.
In this season, I'm talking to the FT's expert tech reporters about each of them.
In this episode, XAI, SpaceX, and Elon Musk.
To talk about Elon Musk and XAI, I spoke to Stephen Morris and to Hannah Murphy, who writes about technology for the FT in San Francisco.
You'll hear from her in a moment.
But first, I wanted Stephen to recap a bit of Musk's history with AI, starting with a famous debate he once had with Larry Page, a co-founder of Google, at a party back in 2015.
Yeah, it's his birthday.
I think it's his 44th birthday party.
His then wife, I think Tallulah Riley, the actress, had organized it for him.
And he was having a debate, a late night debate around a fire pit with Larry Page, as you said, one of the co-founders of Google.
Apparently, Larry Page has some disorder of his throat or vocal cords, which means he has to whisper very quietly.
And everyone at the party was trying to listen into this increasingly.
frenetic debate about the future of AI.
And the way that Musk tells the story is that essentially Larry Page wanted to bring about the AI singularity and merge AI technology with humans to create, not androids, I forget what a human merged with a machine is called.
What's the name?
Cyborg?
Cyborgs.
And Elon Musk didn't like the idea of this at all.
He is all about the human, which is probably why he has so many children.
And he doesn't want them to be merged with AI and for humanity to essentially lose part of itself.
So they start arguing more and more.
And eventually, Larry Page, according to Elon Musk, calls him a speciest, which means he's somebody that prioritizes this human over the machine.
And Elon Musk goes, well, of course I do.
And after this, they were no longer on speaking terms.
And Elon Musk concluded from this exchange that Google was a dangerous company to have.
control of AI technology.
And right from the beginning, it looked like this was going to be the default position.
And Google had more money than anyone to invest in it.
It bought the leading lab, DeepMind.
And that, at least in Elon Musk telling, is why part of the reason why he determined to try and build all of these rival efforts, whether it's through Tesla, whether it's OpenAI, or whether it's now XAI.
because he just doesn't trust Google.
And he's been repeatedly frustrated in his efforts to do things in AI.
He was an early investor in DeepMind, and then it was sold to Google.
Which he tried to block.
And he was a co-founder of OpenAI, which he explicitly set up to compete with Google and maybe develop AI with more attention paid to the potential risks.
Obviously, OpenAI went on to be a big deal.
So how come he wasn't part of that?
He was the biggest donor who donated $38 million to them in rent and chip payments.
And he lent his name to the company, helped them recruit.
I mean, he definitely had a leading, very prominent role in the founding of this company.
But when he decided that the only way to compete with Google was raw money and the computing power that came with that, he wanted them to either merge with Tesla, his electric vehicle company, which of course develops real-world AI in its full self-driving system and develops its own chips.
or he wanted to be given a 90% stake in the for-profit company, essentially sidelining the other founders like Sam Altman and Greg Brockman and Ilya Stutskova and taking control for himself.
They wouldn't give him this.
Part of the reason was they didn't want to be cut out of the monetary upside, but also they didn't trust Elon.
I mean, it's difficult to not know a lot about the man because of his preeminence in this current moment in capitalism.
But Elon is a difficult person to work with, I think it's fair to say.
And giving him 90% control of the company was essentially signing away their own destinies, their ability to control.
And there were serious questions about whether Musk actually cared about AI safety.
He says he does, but does he actually care?
Or did he care more about winning and beating Google?
And what was he prepared to do in order to get there?
So they had a big argument when he couldn't get control.
They essentially refused to accede to any of his demands.
And he left the board, I think, in early 2018.
And OpenAI went on to launch ChatGPT and kickstart the AI revolution.
And so how does Musk respond to the rise of ChatGPT?
Well, he's obviously very annoyed.
He's a competitive person.
And, you know, he wants to get back in the game however he can.
So he founds his own company, which he calls XAI.
Almost everything in the Musk world has an X in it.
Tanner and I have written a number of stories kind of chronicling, you know, the brutality of working for Elon Musk as he fights to catch up rivals in AI.
Yeah, let's talk about that.
So Musk now has his own AI lab called XAI.
It's got its own chatbot called Grok.
And he appears to want to directly compete with the likes of Google and OpenAI.
But the company seems to have a very different culture to...
other AI labs and there have been a few controversies and problems.
Hannah, you've also covered Musk over the years.
What's going on?
I mean, it's hard for Stephen and I because how many times can you write the same story that there's been a reshuffle in the executive rankings that a huge number of people have left?
I mean, it's the same story every couple of months.
You'll speak to AI staff at...
an XAI who will tell you that they will learn that they are supposed to do a product release about a new product they haven't really heard much about on Twitter by Musk posting it on Twitter, not even inside the company being told or given a clear timeline.
And I think labs like OpenAI and DeepMind have for years been slowly, steadily, consistently building up processes, a tech stack around rolling out these models that works.
And, you know, Elon himself has said he doesn't believe in research, the term researcher.
He thinks that's a relic and you should all be engineers working on the product now.
But it is that is at odds with some of the more sort of academic bent of cutting edge frontier breakthroughs.
And then you also have the sort of organizational chaos.
And then you have Musk saying that.
Grok itself has to embody his free speech ideals.
I asked Grok itself if it was behind.
In fact, I said, why is Grok behind the other models?
And it pushed back on this and said it wasn't meaningfully behind.
Oh, really?
But then it added, Grok trades some, in quotes, safe slash polished traits.
for maximum truth-seeking, real-time knowledge, and less refusal.
So this is the idea that the model shouldn't be moralizing or paternalistic, and it should be as close to totally freewheeling.
I think there's an unhinged mode, as it's called.
But we've had a couple of crises, two that spring to mind.
One is, last summer, it started kind of glomming onto various conspiracy theories around white supremacy and… talking about Hitler a lot, even when it wasn't sort of prompted directly and it even described itself as Grock Hitler.
Wasn't it Mecha Hitler?
Mecha Hitler, yes.
Mecha Hitler.
And then at the beginning of this year, you have a big quarry around non-consensual deepfake imagery.
So the model making deepfakes of real people, but in various states of undress or sort of sexually graphic somehow.
And that was typically targeting women and also children.
So that caused a huge backlash.
And in both of these instances, it was Musk who sort of propelled these crises forward by, in the case of the non-consensual imagery, Musk had tweeted a deepfake of himself in a bikini and then suddenly hundreds of other people tried to do similar.
He'd also...
on the Mecca Hitler piece.
Both cases, it's sort of he, the man, spurring on this crisis, which then becomes a huge distraction and presumably internally having to deal with regulators, press, etc.
knocking at your door.
So we should say that Elon Musk has said subsequently that any use of Grok to create illegal content is, well, illegal, and therefore people are going to have to be responsible for that, and he doesn't condone.
that at all.
But this issue that you bring up, Hannah, has shown, you know, there are so many of these conflicting things about him.
He wants to be a free speech absolutionist and be at the forefront of the AI race and push people incredibly hard, but also keep all the talent with him.
Do you get the sense, you know, is there an audience for what Grok is becoming?
Is there a way of turning this into a viable business?
I think for me, a signal was last week, speaking to someone, a source, and we were talking about Meta, and they noted that while Meta has been trying to catch up, it's still a little bit behind OpenAI, Gemini, Anthropic, but it's made some traction recently.
And they were saying Elon simply has not.
It's rare to hear someone say that.
Mark is doing better than Elon.
And this tells me that, you know, Elon really is beginning to trail in this respect.
I mean, there will always be Elon obsessed Twitter fans and users who will, I'm sure, use Grok.
Sometimes I will use it for research because it is good at scouring Twitter itself for anything people have said, but it's not in areas like coding, like perhaps, you know, having it.
particular edges in particular areas with the cursor acquisition as a bit of an acknowledgement that it is behind in a certain area.
And I don't think we have a very clear sense of exactly what is the consumer and or enterprise target here.
Stephen, very briefly, what's the cursor acquisition for those uninitiated?
So cursor was an AI startup that specialized in coding, which is the most persuasive, most lucrative business case.
for AI so far.
Musk always said that the grok was going to be hyper-focused on coding and was going to catch up and then lead the frontier.
This hasn't happened.
So he's turned to cursor.
First, he hired a couple of their senior staff to turbocharge his own efforts.
He then dispensed with, I think, all seven of his co-founders at XAI growing frustrated with their performance or them growing frustrated with his demands.
He's then got an option.
to buy this startup cursor for $60 billion, which is a huge increase in its valuation.
And as Hannah said, you don't buy a rival coding business for $60 billion if you think your internal efforts are going well, do you?
So Musk is facing a lot of challenges with his AI business.
We've had controversies with Grok, and he's struggling to compete, particularly when it comes to the business cases for generative AI like coding.
But there's one massive event that we're expecting this year that could change everything.
Because XAI is now part of Musk's other company, SpaceX.
And SpaceX is going public this year in what's expected to be the biggest IPO in history.
So after the break, we'll talk about what that might mean for his AI ambitions.
So look, must the man of many contradictions hear.
We have told the story largely of woe, that he...
has constantly had the vision to know that AI was going to be a really fundamental, important technology, try to get in on it.
Maybe some personal failings have led to him not being able to really kick on.
And yet he's the world's richest man who's on the verge of his greatest ever payday with the SpaceX IPO, which looks like it's going to be massive.
And XAI is part of SpaceX now.
How to explain all of that, Stephen?
You wrote a great piece looking at what the SpaceX IPO means.
Just talk us through how massive it's going to be and what Musk is trying to get out of it.
You know, three companies in one.
X, the social media network X, the AI, model developer XAI, and then of course SpaceX, which is part rocket launch business, part satellite broadband business.
The cynical argument is that X was losing money.
It didn't have much traction.
bankers and investors and experts, you're facing quite steep losses on the money that they put in to help Musk buy it.
And then XAI is burning, as they say in the industry, cash burn, spending more money than they generate in revenue.
Tens of billions a year.
So what do you do?
You know, people are looking at XAI and thinking, is it going to be able to raise money?
Like, does it have a viable product?
What is Grok?
Who is it for?
So then you merge it with SpaceX and you say, actually, this has been the plan all along.
We're going to build data centers, orbital data centers in space.
And that's how we're going to beat Google because they can't do this because we're the only ones that have rockets.
We already have a network of 10,000 Starlink satellites up there, all connected by lasers.
We're building an even bigger rocket, which can send even bigger Starlink satellites up.
And this interconnected cluster of telco technology will eventually be turned into giant floating data centers that don't need to be cooled down because obviously space is very cold.
Don't need to pay for energy because the sun is so much stronger outside of the atmosphere.
So that's kind of like...
The optimistic bull case, you know, if you're looking at the bear case, you know, he hasn't actually got the Starship rocket to work yet without either failing or exploding.
Nobody quite knows how to make a data center.
One of the biggest problems is keeping all of these chips in a big cluster functional.
If you have to send up an astronaut or a robot up there to fix it in space, that seems to me to be slightly more difficult than sending somebody in a hard hat into a data center.
And, you know, his competitors for once have far deeper pockets.
You know, Musk is severely outgunned on multiple fronts.
And whilst we haven't even had the SpaceX IPO, which is set to raise somewhere between 50 and 90 billion at a valuation of 1.75 trillion, people are already talking about the eventual merger of Tesla, which is a $1.4 trillion company with SpaceX, which would then put together an entity, which if you believe the valuation would fit very nicely into the top five companies in the world.
What we're describing here with Elon Musk is somebody who's spread quite thin.
He wants to run a space exploration company, an electric car company, a humanoid robot company.
He wants to build a free speech social media group that dominates all others.
And he also wants to conquer AI.
Hannah, you've watched Musk try and run multiple companies at once.
How good is he at juggling all these balls all at the same time?
I feel like to compare him, as I tend to these days, to Mark Zuckerberg, the Mark Zuckerberg trope is that when he fixates on something, it's like the eye of Sauron and he will have an obsession, one obsession at a time.
And at the moment for Mark Zuckerberg, it is AI.
And I think it is actually fairly similar with Musk.
From where I sit, I think X is getting a much less airtime.
There was a year or two around the takeover where...
He was obsessed with it.
And then there was the sort of the Trump campaigning where he was very, very active and the company itself was getting a lot of attention and resources.
And now it feels like that's sort of falling by the wayside.
And it feels like he's definitely moved into a more expansive extraterrestrial, as Stephen says, moment where he's trying to elevate his pitch beyond just a consumer or enterprise AI.
Stephen, what do you think is Musk's ultimate goal when it comes to AI?
What does he ultimately want to achieve?
He's trying to build, everyone's trying to build AGI, which is this vaguely defined concept of artificial general intelligence, which is loosely understood as when a machine can perform almost every task.
better than a human.
But right at the core of his belief, which is somewhat at the core of these other sort of AI philosopher kings like Dario Amadei at Anthropic and Demis Asabis at DeepMind, they all kind of think that they are the only people that can be trusted with AGI.
They're the only ones that fully understand how the technology comes together and can...
make sure that it's controlled on the downside.
And a lot of people that work for Musk in particular don't necessarily agree that he should be the guy with his finger on the button.
People that Hannah and I have spoken to from incidents like Mecca Hitler to the indecent images of children, Musk is willing to push the boundaries in order to try and make a splash and catch up with his rivals.
So it will be really, is going to be a fascinating...
you know, I would say the next one to five years.
I mean, most of these AI labs say that AGI will be reached somewhere between the next 12 to 36 months.
But in terms of like the next frontier of like data centers in space, I flip on a weekly basis between whether I think this is all marketing science fiction nonsense or whether Musk, who has often achieved the seemingly impossible in the past, can pull this off again.
And once he executes this IPO, he's certainly not going to lack for funds in the same way that he has over the last few years.
So it's going to be a very busy period, I think, for the Financial Times tech team, but also a very interesting one.
Hannah, as we try and assess where Musk is and where he might...
end up going.
It's clear, whatever we think about his political views, he's an extraordinary entrepreneur and he'll be a very important figure in the world of business and technology and will be remembered as such.
But do you think he'll be remembered as a titan of AI into the future?
I think that we cannot say because...
It seems like someone's out ahead and it seemed like, for example, just until recently, Google was way behind and then suddenly they sped up and then suddenly everyone was talking about Anthropic and OpenAI who nobody thought anyone could usurp and who were the clear front runner sort of trailing behind and re-orging.
So I think it's ever morphing and there's sort of many different visions and maybe many different visions can win and Musk will.
perhaps have an edge on that more sci-fi vision that Stephen outlined.
Stephen, I'll give you the final word.
Can he do it?
Yeah.
Can he win the AI race?
I don't necessarily know it's one of these, like, I know tech has trended towards monopoly or oligopoly in recent years, certainly.
I think there's room for more than one winner.
And I'm loathe to put my voice on a podcast saying that he will be one of the losers as opposed to one of the winners.
But I think his failure to gain any traction, despite years of effort, does suggest to me that he may lose very important parts of the AI race to rivals.
I think he's in a much stronger position on like robotics and autonomy.
And of course, if the future of data centers is these vast orbital things up in the outer atmosphere, then he's definitely in the strongest position.
for that.
But I don't know what Grok is.
Like, you hear stories about what people use it for.
There was a big spat between Anthropic and the Pentagon, which was because of concerns about mass domestic surveillance and autonomous lethal weapons, the contract that Anthropic refused to.
Musk just signed that immediately, trying to get some more revenue for Grok.
But somebody in the government told us that the only thing they use it for is Grok acting as like a bad AI that's attacking government systems.
So, you know, whilst Musk is probably more philosophically aligned with the current direction of the Pentagon, even his bedfellows there don't actually think his technology is good enough to be used in intelligence or military settings.
So that kind of just shows you where he is.
Hannah Murphy and Stephen Morris, thank you both very much.
Thank you.
Thank you.
That was Hannah Murphy and Stephen Morris talking to me about Elon Musk's AI ambitions.
You can read their reporting on Musk at ft.com and there are some free-to-read articles in the show notes.
Thanks for listening.
We'll be back next time on Tectonic with the last in the current season on Silicon Valley's AI Labs when we'll be talking about Mark Zuckerberg.
First it was social media, then it was advertising, then it was supposed to be the metaverse.
Now Meta is spending billions and billions and billions on AI.
So will it pay off?
Find out next time on Tectonic.
Tectonic is hosted by me, Murad Ahmed.
The producer is Edwin Lane.
The executive producer is Topher Forhez.
Sound design by Breen Turner and Sam Giavinko.
Original music by Metaphor Music.
Cheryl Brumley is the FT's global head of audio.
To be a good teammate, you have to not only take criticism, but you have to be a little softer when you give it.
That's Lloyd Blankfein, the former CEO of Goldman Sachs, on the leadership lesson it took him decades to learn.
Every episode of Executive Decisions goes inside the moments that define a leader's career.
Executive Decisions with me, Steve Sedgwick.
Listen and watch now wherever you get your podcasts.