The age of impunity has come to an end.

I think there's a lot of catastrophizing on Twitter about the data center thing.

But I think if you talk to people who are on the ground actually doing this stuff, it's not that they're so much concerned that this means they're not going to build any more data centers.

I think it's more like, okay, we need to reset our our public relations strategy here.

Like the tech industry always had a PR strategy around data centers, which was like, come in, you know, with these NDAs, like sneak into these towns and we're going to like, you know, we'll get the deal almost there and then we'll throw in some sweeteners and we'll fund the public schools and everybody will be happy and we'll move on.

That kind of worked in the cloud era, but It was never going to work in the AI era where you're talking about scales that are so much larger.

And so I think that this just means like there's going to have to be a lot more just like upfront PR slash deal sweeteners right off the bat.

tax, which I'm quite surprised about.

This week, to my great surprise, Xbox executives kind of seem to have leaned into the trend, um, and you wrote about it.

Yeah, yeah.

They started Xbox Labs, and it was basically...

It's really almost not a, a, a, a new thing.

It's like this is, this stuff happens within Xbox.

They experiment with the kind of thing that you were talking about internally anyway.

But I think it's them saying to the world like, "Hey, look, like we, we're on board here.

And, and the CEO of Xbox wrote a kind of open letter where they said, "These experiments challenge long-held assumptions about our industry, the tether between software and hardware, the boundary between creation and play, and how people discover and experience games." I think those second two things we've spoken about, but what does she mean about the tether between software and hardware?

I don't necessarily actually think that they're blockbuster games, but I, I thought it was interesting that Xbox was leaning in, but what's your interpretation of that story right now?

Oh, yeah, I totally agree.

I was talking with someone, I, I think someone at Xbox about that- You know, earlier in the week, I think Asha is probably the right CEO for Xbox right now, because I think she's like, you know, she's a little more forward-thinking, she's flexible.

Um, and I think they're like, yeah, I mean, if you can't beat them, join them.

Sure.

But I mean, yeah, I think you totally can.

But I also think it's like who...

It's like there's not some law that's being violated by this already.

Obviously, you know, private sector... billionaire? What's he doing at NASA? And what's the public-private relationship in space in the US today, Reid?

Well, I mean, he's certainly helping Elon, right? There was a lot of tension with Elon and the Biden administration.

And I think Jared's helping Elon.

There's less tension between SpaceX and NASA.

But, you know, I haven't heard any sort of controversies about him.

He seems to be, you know, doing a fine job.

We'll see what the effects of his term are.

So I do think that building a scooter company ironically kind of prepared him to just do some sort of hardware startup generally.

[laughs]

It's very similar to the big security boom that happened, like, a decade ago when...

I don't know if you remember this, Natasha, there were just, like, 8 million security startups that did, like, telemetry.

They essentially, like, all did the same thing, and then every company in the world would, like, hire them, of course, just usually as a pilot and not giving them very much money.

But, like, then they'd get all this data, which turned out to not really be that useful in cybersecurity, and a lot of them just, like, disappeared.

I feel like it's a similar thing, except it took, it took, like, decades for that to materialize after the internet, and, like, in AI fashion, everything happens so quickly that this has, like, popped up overnight, and it will probably lose its mystique very quickly as well is my, my prediction.

Oh, you did?

... incredibly useful for me already.

[laughs]

Um, who's, you know, who's actually, like, you know, cares about safety? Would the Chinese go along with an agreement? Etce- et cetera, et cetera.

I think, personally, the, the likelihood of some sort of agreement between the US and China to, to slow the, you know, slow down or pace the frontier of AI development is, is super unlikely.

Um, I just think it's, it's also just so difficult to, you know, stop the, the advance in software, right? This is not like, like chips, right? That could be-- that can-- that you can't have export controls on software.

It's just a, you know, a thing you can send over the internet.

So I, I, I think it's super unlikely.

Of course, you know, AI lab chiefs You know, presented in front of the UN, and they, not surprisingly, called for, you know, some sort of regulat- some sort of vague regulation, you know, an agreement between all the nations to, you know, to make sure that AI is safe.

Did they buy OpenClaw, or did OpenAI buy OpenClaw, I forgot?

What's going on with the IPOs? I mean, people are talking about this being the year of $3 trillion IPOs, but it looks like OpenAI and Anthropic might not happen this year, or probably won't happen this year.

Yeah, because I mean, ultimately, like, they're not right now, they haven't really figured out their finances.

I think that's pretty clear.

I mean, you see a lot of numbers get leaked in the press about how much revenue these companies have.

And They're profitable, et cetera.

But none of this is audited gap financials.

These are all leaks to the press to manipulate investors in some way or another.

We could be preventing that future, but if we legislate towards that future, all the stuff we are seeing is just going to fester, I think.

Reed and

Tasha, do you agree with Taylor's take? I do agree.

I think it is kind of a rollback.

I mean, look, my kids are eight and 10.

They're too young to use this stuff right now.

So I sort of had this fantasy in my mind since they were born that by the time they are of social media age, we'll have figured this all out.

Clearly, that fantasy is not going to become reality.

But yeah, I think when you dig into it, it is pretty interesting.

Uh, but if the markup is lower and lower, do we see that sort of, uh, unlimited demand can, you know, persist?

Well, I think there's two or three...

There's, there's actually two or three sort of different things going on there, right? One is who's, who's actually selling the equipment for the data centers, right? The GPU providers.

Um, then you've also, you've got the, the model providers, right? Who are, who are creating the software, the application layer, the, the frontier models for, that run on these things.

Um, and then you're actually talking about the, the data center services providers, right? The, it, you know, just having a data center and being a dumb, you know, essentially like rack provider is not actually like a great long-term business.

It's pretty good now.

But that's why like these companies, like CoreWeave, they wanna offer services on top of the data center, right? So there's actually like three things going on.

... talk through what happened and why you feel that way.

That prompted them to issue the restrictions, which have been seen as one of the strongest examples of US government intervention into AI companies yet

One thing we learned over the weekend was that one of the rationale for the administration doing this was that they had heard that China had gained access to this model via an organization with ties to the Chinese government.

So that was, I think their fear, was that this could fall into the hands of America's adversaries

Anthropic is preparing for a blockbuster IPO, and other AI companies are watching closely to see how the government might suddenly put the brakes on their work

If you just think about it logically, where this goes if there's not some sort of compromise that allows Anthropic to continue offering these models, they've spent billions upon billions of dollars training these things and developing them.

If they can't go and sell them on the market, then there is no business there, and these companies are trying to go public.

So this is hugely important to the American economy, and it's just, you could say it's too big to fail or what have you, but it's too big to shut down

Could you just give us sort of the simplified version of what actually happened here, the sequence of events leading to Anthropic pulling these models offline?

The simplified version, I think, is that Anthropic did a great job of marketing this model, that Mythos model back in April when they announced it, and they said, "This thing is a danger to the world because it is so good at finding software vulnerabilities and bugs that hackers will be able to use this thing." I mean, I was in DC in April right after they announced it.

I mean, people were, were genuinely scared.

And I think when the White House started to get reports from companies like Amazon, which is a big investor in Anthropic, an important partner, saying people have been able to jailbreak Fable-5, I think it actually scared them, and then I think Anthropic saying, you know, "We're not...

We don't think these jailbreaks are that serious," um, not responding with enough urgency, um, you know, it just led to, I think, this kind of cascading series of events that led to the, the export control.

And it didn't help that they have just had one conflict after another with the White House for the past year, which I've, I've been covering.

[laughs] I've been writing about this a lot.

We're gonna go after them." Do you have a view on, [laughs] on who might be in the right or the wrong?

Could you just give us sort of the simplified version of what actually happened here, the sequence of events leading to Anthropic pulling these models offline?

The simplified version, I think, is that Anthropic did a great job of marketing this model, that Mythos model, back in April when they announced it, and they said, "This thing is a danger to the world because it is so good at finding software vulnerabilities and bugs that hackers will be able to use this thing." I mean, I was in DC in April right after they announced it.

I mean, people were, were genuinely scared.

And I think when the White House started to get reports from companies like Amazon, which is a big investor in Anthropic, an important partner, saying people have been able to jailbreak Fable 5, I think it actually scared them, and then I think Anthropic saying, you know, "We're not-- we don't think these jailbreaks are that serious," um, not responding with enough urgency, um, you know, it just led to, I think, this kind of cascading series of events that led to the, the export control.

And it didn't help that they have just had one conflict after another with the White House for the past year, which I've, I've been covering [chuckles] I've been writing about this a lot.

Do you have a view on, [laughs] on who might be in the right or the wrong?

I wouldn't read too much into the Hegseth, Hegseth tweet.

I mean, look, I...

Is this one giant step backward for humanity, or, or are we still confident that the Artemis program is, uh, is on track?

No, I th-- look, I think it's-I, I wouldn't, I wouldn't view any- anything as, like, a step back, right? If it, if it's a test, nobody died.

You know, they learn- they're always gonna learn something from this, right? That's, that's what you do.

You learn and then you, you move on.

I just don't...

I, I think that it's funny, like, when you- when the media sees, like, something blow up, a- an apocalyptic mushroom cloud in a rocket test, it's like the immediate thing is like, "Oh, failure.

Like, look at these companies.

We're gonna go to a quick break, but Reid, before we do that, is there a race between SpaceX and Blue Origin? Does this explosion mean that Blue Origin have fallen further behind? Or is this, is this a two, is this a two ho- two-horse race or a one-horse race with a, with a pony cantering, cantering behind?

... commodity form to understand tokens?

Yeah, I mean, they're kind of like Bitcoin, but they actually have utility, right? Like people, people need them and, and want them, and you sort of realize that when you try to use, you know, the Anthropic products 'cause you run out of tokens pretty fast, and then you find yourself, you know, at 2:00 a.m.

sort of like a heroin addict, like just wanting more tokens.

Um, y- you know, and I, [laughs] I think, I think that's like to me really indicative.

And, and this Elon, the, the whole xAI or SpaceX, whatever we wanna call it, deal with Anthropic where, you know, xAI found itself actually with some extra tokens, and instead of that being a problem, they just unloaded those to Anthropic, which was short of tokens, and that to me was like, oh, wow, like these things are...

They're general purpose.

They're transferable.

Reid, I have one question for you, which is, which has been burning on my mind actually, which is how much surplus value is there in token creation? I mean, in other words, like today, when you assess the price of building a data center, the electricity that goes into the data center, can minting tokens be a profitable endeavor, or do tokens cost more to make than they're worth today?

The harness revolution, what you mean is basically, like, the OpenClaw, Claude Code moment, basically?

Right.

I mean, that's what, like, the, you see this massive increase in token usage in AI, and it's all because of these harnesses that allow you to, to basically, like, pile on the tokens and build, build software with a single prompt, and that's what, that's what Codex does and, and Claude Code, Claude Cowork, and, and yes, OpenClaw.

And somehow xAI is, like, just chugging along with their chatbot on X and, like, and sort of seems to not realize that this is happening.

And they're like, "Oh my God, we have to acquire...

We gotta acquire something to, to get in this game." I was really shocked by that because as fast as xAI moved in the beginning, they seem to, like, move very slowly in this, in this area.

So I don't know, you know, I don't know what that sa- I mean, obviously there's a huge leadership shakeup at xAI, and, and Musk has said on, you know, publicly that he made a mistake and has to sort of redo things.

Oh, good.

There was a big New Yorker piece from your, your magazine, Kyle, which certainly has got people talking, but I'm curious, like what's signal and what's noise? What are you as somebody who's really tapped in paying attention to?

Yeah, I mean, I wrote about it, this in the newsletter today, and, you know, I think, I think OpenAI, when you, when you read, you know, the, the Ronan Farrow piece in The New Yorker, it's like, okay, there's...

This company is clearly just chaotic.

I mean, that is-- like everyone I know who's worked there says it's just, even by startup standards, just kinda crazy to work for.

Um, but I think the, the, the piece that got me thinking the most was the one in the Information which came out, I think, right around the same time, uh, where they were, you know, Sarah Friar and Sam Altman, according to this article, are sort of fighting about whether they should IPO, when they should IPO, is revenue growth slowing? Um, you know, really kind of a worrying sign for the company, and I think what's...

what you have to do, and [chuckles] what I did, like, uh, am constantly doing covering AI, is like, okay, calm down, take a deep breath, take a step back, just think about this logically.

They had literally three x revenue growthOver, you know, from twenty-four to twenty-five, going from, you know, I think four billion to twenty-one billion dollars.

[laughs]

It looked like the future.

[chuckles] Right.

But I, but I think that-Like, the, the defense of Apple from, from the Apple fanboy crowd, which they're, you know, they're very...

They're many, and I'll probably get hate mail for even saying this, but is like, you know, look, they've-- Yeah, I mean, Apple's great at this.

They just watch technology develop, and then they just jump on top of it and like, and, and, you know, and benefit from it.

And I think that's like a very lazy defense of, of what's happening, and it's-- And the other thing is like, well, you've got this phone and, you know, they're gonna just like the-- they're gonna be able to run all these AI models on the device, and so that's gonna be a huge advantage for Apple.

And it's like, I don't think that's really how it's gonna work.

Why? [laughs]

Uh, what's going on here?

Yeah, I mean, I...

They- So they got rid of Sora, their, their video generation tool, and what I thought was so interesting about that, it's, it's funny 'cause it kinda ties in with what Natasha was talking about with, with what is OpenAI gonna do now with social media.

There was sort of a social media app that was part of this.

But I thought was, what was so interesting was, like, what the real reason is that they just c- they don't have enough tokens to go around, right? And they're seeing this, there's a huge profitable, you know, boom right now, like with OpenClaw, which we talked about last week, you know, just causing people to use way, way more compute, and they're kinda seeing where this is heading, and they have to...

They're making tough decisions about where they wanna allocate their compute, and this was not a profitable area.

Like, it's just, you know, they had this deal with Disney.

Like, what is the big takeaway here for you?

And what are the competitive threats?

I think it's actually, it's funny.

The thing about Nvidia is it's almost, like, too simple of a story, right? It's like there's these things called tokens that we're all using, right? That's what, that's what we're using when we, when we query ChatGPT, and people can't get enough of them, and they're willing to pay a lot of money for them.

And you could see with, like, this OpenClaw stuff, there will actually, the more tokens you use, the better it gets, and people are paying for it.

And it's like, okay, so these companies are making a product that is pretty profitable, has good margins, and everybody wants, and they can't make enough of it.

Like, that's just a good business.

Like, and I, I...

That, that, that's [laughing]

I think some people would maybe say, "Well, why are we having non-Americans come in, these Middle Easterns come in and control the algorithm?" Um, what do you make of the owner- the ownership? Uh, who's really in control here?

Yeah, I mean, that- that's a- that's another interesting ...

I think that's gonna definitely raise some questions.

Now, if you've been following the Gulf region, which, you know, Semafor has, there have been a bunch of deals between US tech companies and, you know, MGX, and, and G42, and other Abu Dhabi entities, um, UAE entities, really.

So this is...

And this has been what the, the Chinese relationship with the UAE has been this big sort of question mark around there.

Should we even allow, for instance, Nvidia chips to be shipped there? Um, the US government now, you know, has, it's been, it's, it's sort of under Trump, has been a bit more open to dealing with, with those entities.

Now, maybe that's not large enough to attribute that this is because of the TikTok deal, but I'm certain that there's definitely a question that investors are asking, like, "What does this actually mean for TikTok, the product?" And perhaps could it be, you know, worsening something that's already, already working very well? What do you think?

Let's just give you, you know, the bullet points." Then you go away.

away.Yeah. No, I, I, I, I see what you mean.

Um, I actually, I like Gemini too.

Like, I, I use that.

I think it's good for, like you kind of said, anything visual.

Like, I used it to try to, like, redecorate my house, which, you know, was like just, just in mayhem.

And so it was really helpful for that.

Reed, i- is this going to be the year and why ha- why is Tesla and, and not really viewed in the same league as Waymo? I guess, it's not as good?

Reid, what do you think? And what did you make of Meta's event last week or two weeks ago?

I was there.

And yeah, despite all the hiccups in the demo, I think it's a really interesting product.

Actually, I thought the best part of what Meta did was the wristband.

It's like this new user input that I think, you know, if they sort of open that up, I could see developers coming up with like really creative ways to use that.

Right.

But I thought I read your deep research report on this, and I thought that it was actually it kind of hit the nail on the head when it said, that these, the two, you know, the two companies, Apple, Google, that really control the mobile ecosystem have a tight grip here and they're not gonna let people come in and innovate.

And I wondered, are you seeing different perspectives depending on what part of the world you're coming from on AI, or is it pretty uniform right now?

The big question we've been talking about on this show for a long time, which is, what is the valuation of these chatbots and these m- from foundational model companies and when is it gonna be justified? And I guess if you gobble up the whole web, you're okay.

Well, you know, like, so you mentioned sports scores.

That's an interesting one.

Like, people used to bu- uh, used to pay for newspapers so they could open it up and see the sports score, right? And then, you know, online publications started putting the scores up.

And now I think the way it works is if you, if you Google a sports score, you just see the score right there, and I, I believe they have a deal, maybe it's with ESPN, but, like, eventually, it's, it's gonna...

I think they (laughs) just get those scores directly from the leagues and they're just gonna be paying the league for that information.

This is not really a, it's not, that's not really, like, a valuable service that somebody, that some publication is providing.

That's, that's the question, right?

Does the fact that all these share prices have gone down as much as they have, you know, does the fact that Nvidia could lose a cool half trillion dollars [chuckles] in a couple of months, uh, does that impact the tech industry's willingness, ability to continue to invest in AI? What do you think, Reid?

Reid?No, I, I think that the investment in AI will continue.

Um, you know, there's a...

I think there's two things happening.

I mean, one, for sure, like if you look at the private markets, there's plenty of money to go into these, you know, in, in, into building, you know, building out AI infrastructure, et cetera.

Um, I think there, there may be a question around like is it gonna be the public companies that do that? Because, you know, Amazon is, is spending, you know, $100 billion this year on AI infrastructure, [clears throat] for instance, and Microsoft is not far behind that.

You know, all these companies are spending huge, huge dollars on building out their AI infrastructure, which there is huge demand for as well.

So what have you found and, um, and which part of the executive statement should we believe the most?

So would you take a step back for the lay listener and explain what open source is and why it's relevant to the DeepSeek story?

Yeah.

I mean, so open source is – we call it open source.

A lot of people call it open weights models because they don't actually give you all the information.

They don't tell you exactly how these models are trained, but they tell you – essentially how they work and they allow anyone to run these models anywhere.

So I can go right now, in fact I have, you can go on these websites like Hugging Face and you can download this model and you can run it on your home computer.

Hopefully you have a very fast one.

tech companies, maybe OpenAI, which makes ChatGPT, maybe even others going to sue them? Is there that kind of recourse? Shut them down because they have stolen their code?