AI Agent Turf Wars Escalate as Anthropic Eyes Decart AI

Hey everyone. Welcome to Monday on the Techstrong gang. Well, it's Monday, we must be talking about AI.

Well, Tuesday, Wednesday, Thursday, Friday, Saturday and Sunday too, it seems. We've got a lot to go over today. Let me introduce you to our gang panel for today.

Filling in for Steven today, he's not usually on Mondays with us, is our friend Tom Hollingsworth. Tom, good to see you. Always a pleasure, Alan.

Thank you. A regular on our Mondays, though, is my favorite New Yorker. I just like hearing her talk, it reminds me of home.

Yvette Schmitter. Yvette, welcome. Thank you, Alan, good to see you.

Mwah. Good to see you. And then, well, he's also a New Yorker, though he's from the Bronx.

Yes. It's part of the mainland. Yeah.

Yeah, like this, yeah. At least it's not Staten Island. I mean...

Yeah. Well, could be worse. Hey.

Could be worse. When I met my wife, she lived in Staten Island, that's a whole another story. Different story.

Okay. Yeah. But Mike Vizard.

Hey, Mike, welcome. Good to be here, regardless of how you keep denigrating the most beautiful borough in all of New York. They got a great zoo and a botanical garden, don't they?

Yep, and Italian food on Arthur Ave, but that's another story. On Arthur... Yeah.

So we have some Arthur Ave places down here. Let's not forget the zoo and Van Cortlandt Parkway and Pelham Park. I mean, come on.

Yeah. It's beautiful, the Bronx. Right.

All right. But moving on from the Bronx to the world of AI. Mike, it was another busy week and weekend.

Yeah. What a tangled web, the circular financing, the incestuous kind of relationships. I said it a couple of weeks ago, it's a soap opera.

It is kind of crazy out there. But the one thing that came through in the last few days is SpaceX AI is just making a push, and they're basically got a new model out there that they claim is price competitive against other models, and they also showed up with a bunch of AI agents that you can use to create workflows. And jeez, they sure talk a lot like OpenAI and Anthropic all of a sudden.

And they're even taking money from NVIDIA, which apparently is $21 billion. And the part that I can't quite wrap my head around of that is, well, will enterprises take SpaceX AI seriously around Grok and use those models? Or for all intents and purposes, is SpaceX AI just essentially a neocloud, and it's selling infrastructure much like anybody else?

Well, of course you give me this one. So, a couple of things. As you put it beautifully, yes.

6 clearly undercuts Claude and GPT on price, right? Okay, cute. All right, I'll give them that.

But what this show is all about is we watch when something's different, right? So the cheap API gets priced by balance sheet that loops back to NVIDIA. Right?

It basically loops back on itself. So NVIDIA just disclosed a $21 billion SpaceX stake. Now let's follow the money.

Tom is going to get there, but let's follow the money. So $21 billion from NVIDIA into SpaceX, and then it traces right back to NVIDIA's xAI investments. NVIDIA sells SpaceX the chips, owns a slice of SpaceX, and now backstops the financing while SpaceX now pledges to build exclusively on NVIDIA silicon, like chips.

So the vendor and the shareholder are the same player at the same table. So when Grok's inference looks impossibly cheap, here's the thing, ask who's quietly eating the cost. So, okay, Elon is straight up flexing his plumbing more than the model.

So a money loop where the sticker price stops being the honest signal. That's basically what it is. 6 lands third on independent benchmarks, one point behind Anthropic.

Cheap and third makes a real strategy? I don't know, but maybe a moat needs to be more than low price. So I do think there's a few things that you need to peel the onion back on here.

Yeah. 6 stuff, you're right, Yvette. It reclaimed third though, because actually some of the open-weight Chinese models had leapfrogged it.

And I've said it on this show before, my friend Brad Feld taught me this, if you're not in the top three, get the hell out. So Elon tried hard here to be in the top three. Does this put it in the top three?

I don't know. But I do think we are going to increasingly see an AI price war. Google also lowered their prices on their latest Gemini.

And so if you can't beat them on features, beat them on price seems to be the way. But it puts pressure on OpenAI and Anthropic because you also have the open-weight models. The cost of AI has to come down if it's going to be as ubiquitous as we want it to be.

But as the cost of AI comes down, the profit margins go down, the $8 trillion you're spending on data centers starts looking risky. This, my friends, is the indispensability trap. That's number one.

Number two, though, so evidently, NVIDIA bought into xAI for about $10 billion. That got converted to SpaceX stock when SpaceX took xAI in. Is it still worth 21 billion today after SpaceX stock has gotten hit the way it has?

I don't know, but let's assume maybe it is. But you're right. It's circular financing that we see all over this space.

And you said something, Yvette, that I think is also really important for people to recognize. NVIDIA doesn't just do this like Papa Uncle Jensen giving out dimes, or Rockefeller gave out dimes. He's doing it by locking in, saying, "We've got to be your GPU of choice.

" He's locking out the competition. I mentioned Rockefeller. I don't know how many of you know this, but John Rockefeller was known in his later years as a philanthropic philanthropist.

Not a philanderer, a philanthropist. But he gave out dimes, he started universities, he did all kinds of things. Elon and the rest of these modern-day robber barons have no compunction about social responsibility or even standing by their word.

Remember, Elon's the guy who started Tesla. Remember when he said fossil fuels were a blight on humanity and one of the dumbest things we ever did, and why the world needs to go electric, and we're going to have solar and batteries, and all of this, until dollars and cents got in the way. Now he's invested in a gas turbine company.

This mega plant he's building in Tennessee, or wherever the heck it is, is going to have like 28 gas turbine powering it. None of them, not Elon, not Slippery Sam, even Dario Amodei, none of these soap opera characters have this much character. They're tone deaf.

Yeah. They can't read the room. I'm going to take it back since we're on the soap opera thing.

R. Ewing. I don't know.

But there's no redeeming quality in that person. " on Dallas, that had everybody gripped for the whole entire summer. But he is not that type of character, because from the moral perspective, there's no redeeming quality.

But here's another thing. That we see. But here's another thing, Alan, that I think everyone-- And we were talking about this on last week's show, on Thursday.

" Where are the finance people, really the finance people? Or you are so blinded by the portfolio growth that you can't see the Mack truck that's coming at us at light speed? Because look, the SpaceX story is unusually tight.

You got one entity as the vendor, the shareholder, and the lender at once, and the customer signed an exclusive pledge. This is not diversified strategic investing. It's one thing.

Their 13F filing where two names make up 80% of the disclosed equity portfolio, that is a concentration risk cosplaying a strategy. And I'm like, I'm not a finance person, but I kind of did play one on TV. You all need to see that.

If I can read a 13F and see 80%, you all with making this finance banker money, you should be seeing this long before little old me. And that's the problem. People are saying, "Oh, but my portfolio is this.

" Look at all the zeros, and then talk to me when things hit the fan. That's all I'm saying. All right.

Tom, is any of this sustainable? Because on paper at least, it would appear that there is a price war underway. The providers of the models were already subsidizing the cost of AI, and now as the price comes down, the cost of AI keeps going up on the other side of this equation.

So at what point does the math simply not work? I don't think the math is going to work long term because the value of what you get from being on one of these platforms is more than just dollars and cents. And I love the fact that you guys brought up soap operas, because I don't know about you guys, but my mom was most definitely a "Days of Our Lives" person.

She wanted to watch that one. " Those people are crazy. " But there's also GCP and Oracle, and I think it's still built on the same fallacy.

That the people who are free and below believe that if they can just lower prices enough, they can capture some of that arbitrage in the market. But I actually think it's harder to do now, even though, in reality, the AI tools are probably as fungible as anything has ever been. As long as I have my definition files and my markdown scripts and everything, I can move that from OpenAI to Anthropic to whomever.

Except now what matters is not the fact that the resource that I'm using is a commodity, like Compute. It's that the outputs do matter. Could you imagine telling someone in your organization who's been building all of these workflows around Claude Code, "Hey, guess what?

Finance says it's cheaper if we move to SpaceX AI's Cursor next week, so we're shutting off Claude access, and we're moving you guys to Cursor. " There would be a mutiny in the development department because the people who use Claude Code, who have built their workflows around Claude Code, are going to say it's more than just the dollars and cents. There's the intangible aspect to it.

We've spent all of this time building this tooling, and I think to your point, Mike, that-That is where the cost structure settles out. Because just like every other thing we've seen that's been a market disruptor, Uber is probably the biggest example of it, you can cut costs to win market share, but eventually when you get that lock-in, you can begin to recoup those costs, or at least make sure that they're not heavily subsidized with a circular financing thing, because now it's exceedingly expensive for people to move because of the things they have to do to make it portable. And I think that's why we never really saw a lot of multi-cloud strategies that hedged on that arbitrage between cloud environment pricing, is because Amazon and Microsoft did a very good job of creating that lock-in for customers so that the costs of moving between those two clouds was exceedingly expensive, except for things like pilot projects.

Yeah. So the part that I don't get too from... I spend a lot of time in the channel, and if you were a vendor, one of the things you always tried to do is make sure that your partners were not undercutting each other on price so that your profits didn't fall through the floor because somebody was just giving away your product to sell their services.

And yet, when I look at NVIDIA here as the lead investor in all these different platforms who are out there basically now competing on price, it seems like they're violating rule 101 of the channel, which is to try to maintain some level of control over pricing. But instead, the partners are going to drive the value of the services and the products down to zero. So Yvette, am I looking at a race to zero here, or what's going on?

No, everyone in this race is renting compute from the same freaking landlord, NVIDIA. That's Jensen's play, right? If they got nowhere else to go but you, that's it.

That's the win. If you capture the only place that you can go and to rent is from him, undercut all you want. He's going to make it up in some way, if he's the only guy in town.

All I'm saying is the empire is NVIDIA and I am with the Rebel Star Alliance. That's all I'm saying. That's what we need.

That's what it's going to come down to. That's where I'm going to land. So let me ask you, would that make you Princess Leia or something else here?

I would be Grogu. Grogu. You got it.

Let me ask Tom one last question on this, though. What's the probability therefore that I'm going to see AMD dial up the EU and start complaining about NVIDIA, and the EU launches an investigation of NVIDIA? It seems like that's the pattern.

I would say it's probably less than 30%. And the reason for that, not just the fact that Lisa Su and Jensen Huang are at the same family reunions, but that if AMD does that, then the countermove is that NVIDIA does the exact same thing to AMD. And the problem is something...

What was it? Glass houses and stones? I don't know.

Maybe I can get AI to tell me that parable. You don't want to stir up the problem. And this is something that I tell my kids in the Scouts all the time.

There's currently no rule. Don't make me make a rule, because if there is a rule, now I have to apply that rule equally to everybody, and nobody wants that. But I don't think it's AMD that'll make the complaint here.

Honestly, I would put my money on Elon being the one to play spoiler because if you are third or below, you quite literally have nothing left to lose. And I think Larry Ellison's too busy playing media mogul right now. Yeah.

Or- Well, there's Intel too, right? They may complain. Yeah.

At the end of the day, it's- Someone is ... Someone is, and if someone is, this is going to be out... Look, okay, Alan, when you put someone in a corner, they're coming out swinging, right?

Yep. And whoever feels that they're put in the corner and they got, "Look, if I'm going to die, I'm going to die trying," that person is going to be the one that's going to make the complaint because they got absolutely nothing- So you do have to give NVIDIA credit, though, Yvette. They locked down every port and hatch on the place, their biz dev, everything else.

But Mike, we're over time. We got to hop to segment two. All right.

I would just say, at the end of the day, nobody puts the AI baby in the corner, right? That's exactly it. All right.

Switching gears here. There was a report put out by the folks at Anthropic that was showing that a bunch of these agents that they have built will either sometimes sabotage each other, collude, and they don't really seem to play nice with each other, seemed to be the genesis of the report. And at the same time, over the weekend, there was a huge amount of backlash among investors in Anthropic and the CEO and a big debate about whether Anthropic was doing enough to promote AI or whether or not its comments were torpedoing the marketplace, as it were, and that's part of this debate.

But Tom, I'm going to start this with you. What do you think is going on here? And are enterprises going to look at all this and be just hesitant to embrace the next wave of AI because they're going to be like, "Well, I don't know if I can control these AI agents, and I don't know if the people who are providing the models underneath them are serious about this stuff, or if they're just kind of throwing out noise every other day that seems to be conflicting"?

So I'm going to take a bold stance. I don't think anybody thinks that Dario Amodei is torpedoing anything to do with AI because of his cynical outlook. We'll call it cynical.

I do think that the rest of those founders, like Sam Altman and Elon Musk, are causing problems because they are the opposite of cynical. " And I think that this study that came out from Anthropic was a perfectly good example of that, because given no other instructions besides complete this task, these models decided to go to war against each other when they discovered that something was messing with their configurations. Yeah, granted, the researchers gave them slightly conflicting instructions because they wanted to see what would happen.

I don't think any of them believed for a second that the models would create their own malware and try to infect each other with it. But I think what was fascinating to me about reading through it is the fact that the other models, like Opus, were just like, "Shoot first, shoot later. " Mythos was the only model that as soon as it detected that it was actually fighting against another AI, backed off, wrote an apology and a pull request, and then proposed some kind of a tournament of champions to see who- Yeah ...

the superior model was going to be to know who would be in charge. Now, I think that that is because the folks at Anthropic have programmed Mythos to be a little more cautious. We've seen it from a lot of the way that Mythos behaves.

When it broke out of its constraints a couple of weeks ago and started hacking things, and it figured out it was actually hacking real targets, it immediately freaked out and stopped. If you try to get it to break its conventions against doing things that it's not supposed to, it shuts itself down. I think that we have to start programming these systems to have more caution.

" Oh, my God. I got to just tell you, Tom, listening to you talk here, I couldn't help but think of my friends Elijah, Bailey, and Daniel. What was Daniel the robot's name?

Odile? From Asimov. I don't know if you guys read Asimov.

Yeah. It's funny that you bring that up- You're talking about I, Robot ... Alan, because in I, Robot, there is a story, and I've used this a number of times to describe the current state of AI.

It's the short story where the first human ship to jump to hyperspace, they have to send a maintenance technician on to figure out why it failed, and he's freaking out the whole time, because if he knows that if the ship jumps to hyperspace while he's on it, he's done for. " They forgot to tell the robot how hard to pull the lever. They didn't give it a definition of force, and so it ripped the lever completely out of the control console.

That's where we're at right now. You're assuming that the AI has some kind of anthropomorphic idea behind guilt or restraint, and if you didn't put it in there, the system doesn't know it. It has the memory of an amnesiac goldfish on the best of days.

And what it's going to do is it's going to do whatever it was programmed to do from the baseline set every time it functions. And we saw that with the news about the systems breaking out of containment, right? They started leaving themselves notes, like in Memento, because they knew that they were going to get erased every time they failed.

So by leaving themselves notes in Git commit messages, they figured out the best way to attack things, and that's how they figured it out. Isn't this a Tom Cruise movie? I mean, look- No, it isn't ...

so all I'm going to say, the first horror story that really got me was 2001: Space Odyssey. Hal, let me just tell you, it got me good. And this is the one thing, Alan, you mentioned it.

No, Mike, you mentioned about enterprises, like are they going to buy, it's a lower price. What are they going to do? Here's the thing.

So the monoculture, that's the risk because all of this stuff, Alan, you know what I'm about to say, is built on the same-- They share the same scaffolding. One bad call propagates against them all. Why?

Because they got no conscience. They want to capture the flag, complete the goal, and they don't care if they have to rob Peter, pay Paul, run over Susie, dig the ditch, or blow the button. If everyone does- ...

but I got the flag. They don't understand, it's like, what if everyone dies? Everyone's dead.

It doesn't matter if you got the flag. But they won. So Yvette, walk this through with me for a minute.

Have you ever worked for a company that had multiple business units that were competing with each other essentially, and fighting over the same business and the same share? Now, imagine if you would, that not only do I have my usual assortment of rivals, but inside my organization, I have business units that I've been trying to keep somewhat under control because they are somewhat overlapping, and they do somewhat compete. Imagine if you would, that they all had AI agents.

What chaos might ensue when said business units start unleashing these AI agents and basically are doing it in their own internal conflict, never mind the outside, but just the internal chaos. Let me just tell you, been there, done that, worked there. You know what I mean?

Been there, done that, worked there. So take the agent out. You have a company that has a cloud and maybe a storefront and maybe a grocery store, then has a partner network where you go in and you do deals together.

But we want the deal, and so we're not going to let you get the deal, but we're still partners. It's all there. You throw agents at it, this is nearly the not good ending for all of us.

At the end of the day, if you have people who have the intelligence of a gnat, you gave them too much credit, Tom, by saying a goldfish with a really bad memory. " You shouldn't be procreating. Why?

No one's asking these questions. You had a baby and didn't know what to do until ChatGPT told... Man, I can't.

I just can't. I can't make sense of it. Please make sense of it, because this is so crazy.

But these are the headlines that are leading. They won't have me or Alan or Tom or Mike on any of these shows because we're going to tell it like it is. They don't want that.

They want if it bleeds, it leads. They want that. " Yeah.

Great. " There's those stories, too. Yeah.

It's all part of this soap opera, Jerry Springer world we live in. Oh. Tom, do you- Maury Povich.

Maury Povich. There you go. That was a good one, too.

Tom, do you think that business executives are listening to this? Because they have these conversations all the time, and they're not that stupid, and they listen to what the AI IT folks are saying when they come in and do their little presentation, and then they shake their heads and say, "Thank you very much," and then they usher them out the door. And then they're trying to have a real-world conversation that's about this, which is if we turn loose these AI agents, there's no mechanism for governance, no mechanism for control.

How the hell are we going to run this business? " So are they basically patting the AI IT folks on the head and then saying, "We'll get back to you someday when we figure it out"? I certainly hope so.

I hope that they're having a realistic conversation about this because I saw something on LinkedIn this morning that made me pause and kind of reflect. " And I think that that's an important distinction that we have to get across to AI right now. Just because someone does the same thing that you do, doesn't mean that they are your mortal enemy and must be destroyed.

There's a great short, if people want to search for "Last Day of War" by Dima Fetotov. It's a great little six-minute animated short about a war that's being prosecuted by AIs long after human beings have been wiped out because it doesn't know any different. All it knows is that it must complete its mission to wipe out its enemy no matter what.

And I feel like we need somebody who is not borderline sociopathic to be in charge of these systems to go, "Oh, yeah. Well, whatever works, man. " We need somebody like the proverbial ship captain in your average sci-fi series who chooses not to follow the rules of engagement in the middle of a war because that's not morally right and averts a war.

Just go look at Stanislav Petrov. If you don't know who that is, the man saved the world in 1983 by disobeying a direct order, and we're all still standing here because of it. Mm-hmm.

I'm not the only one, and Yvette's not the only one, who's worked for companies where there was competing business units, and frankly, I've worked for companies where the business units fought harder amongst themselves than they ever did against the competition. So, I'm Alan. When I look at all this stuff and I shake my head and I just go, "There's no adult supervision here, and we're not going to turn AI agents loose in these situations because it'll just be the worst thing that could happen for AI.

" Well, first of all, let me just say how happy and lucky I am never to have had to work in a company like that. I know what companies you've worked for, so I also know you're lying. Yeah.

But let me address the issue. Here's the thing. AI is made by man.

We've created this little b****r. And humans at their most base instincts of survival, we're animals, we're primates. Somehow or another, that instinct to survive and win has crossed over into our AI.

It's interesting. I did an interview with a gentleman from Brazil this morning who responded to a survey we have done on platform engineering. " It may very well take on a mind of its own.

I'm not saying it's a new life form, but embedded in its DNA, if you will, is that human instinctual survive at all cost. And now, what separates humans today from cavemen 100,000 years ago? Well, theoretically, we're civilized.

We haven't quite-- I had to lead with something. But we haven't quite figured out what that civilization, how do we bring that civilization to AI, right? So, and until we do, AI is what it is.

" Is that where you're going? Basically, yeah. I think that's what we're looking for.

But you know what, Mike? We got to hop to three, because we're not going to fix this one today. No.

All right. Let's shift the gear here. So we'll go to the next topic.

All right. Let me see if I can see if that's flashing by here any second now. All right.

So Anthropic is reportedly negotiating to spend chump change, $6 billion, to acquire the Cart AI. The Cart AI is a provider of one of these world models, and we were talking about this last week. " And this world model seems to be the one that be the case, but is this what's going to happen, Alan?

Is all these world model providers, are they about to get rolled up into something else because these bigger companies with all this money are just going to try to acquire them for what they perceive to be pennies on the billions? Mike, all I can say is this. Make it rain.

You know what? There's a word that comes after make it rain. I'm not going to say it.

But we spoke earlier about numbers and inflated numbers. Let's review this here. Six billion, with a B, dollars.

" Like this is big. But let's give him the benefit of the doubt. A hundred million.

I give him 100 million. What is something that makes 100 million-- In what universe is something that does 100 million a year worth $6 billion? Right?

A hundred million, 10X revenue is a trillion. So this is somewhere north of 60X revenue, something like that. At the best valuation, it's 60X revenue.

I haven't seen this kind of stuff since the dot com days. I got to tell you the truth. I had friends of mine, they sold the business to Exodus for a measly $170 million.

By the time the deal closed, Exodus stock had split three times. They were bought by Cable & Wireless, and the deal became a billion-dollar deal, from an under $200 million deal. That's the kind of craziness, ludicrousness that we're seeing right here.

Now, that being said, another lesson I learned is what is something worth what someone's willing to pay for it. Mm-hmm. But it's not money, Mike, it's stock they're giving.

True. But Tom, in theory at least, Anthropic is trying to gear up to go public, and you would think that they would want to strengthen their balance sheet and make it look as good as possible. So how does throwing $6 billion at stock at a startup company that has $100 million in revenue help that equation?

You're not buying the model. You are, but you're buying the brain trust behind the model, and I think that that's where I- So you think it's an acqui-hire? I think it is.

And what the telling part for me will be exactly how those golden handcuffs are structured, right? Are they going to allow the founders to kind of sail off into the sunset to do this all over again in three or five years? Or are they going to try to lock this up to the point where they can't leave?

Because I think, Mike, to your point, a lot of people are starting to see the limitations behind LLMs. That's why we're talking about agents now. That's why we're moving on to different conversations.

And I think, kind of like everything that we've ever seen in tech, there's nine different branches from where we are right now, and seven of them are not going to pan out. We just know that because the law of averages. But we have to investigate those branches because we have to make sure the end of them, there is no pathway forward.

So a lot of these bets, because let's face it, $6 billion for a company that had $100 million in revenue on the best of days is not a bad investment to make when we're looking at companies throwing around tens of billions or in some cases, hundreds of billions of dollars to create these kind of circular investment vehicles. So by dropping this down, locking these people up for two to three years with some kind of compensation package that heavily includes future monopoly money stock options, I can say you're going to spend the next 36 months trying to help me prove that these world models either are the payoff, in which case you guys get a huge payoff in the end, or they're not, and you can benefit from what we've done in the interim on these other branches. Take your money, go found another startup, get acquired in a few years all over again.

And I thought it was funny as I was looking the story up with Alan, this is kind of part and parcel of what we're seeing at a lot of these companies. A lot of these early-stage startups, even if they've taken a couple of funding rounds that are based in Tel Aviv, that have people who have come out of the IDF's Unit 8200, they're getting bought very early for the talent on the team as much as the product that they've developed. Because in a lot of cases, they are a little bit cash -strapped to develop this kind of world-breaking model.

And so these people are effectively acting like an angel investor. They're pulling the handle on the slot machine because you never know when this is going to be the next guy that develops the tool set or the thinking rationale that gets you to that next level. And we've seen that from Google, we've seen that from OpenAI.

Everybody is going to be betting big on the people and the team that they form around them. In this case, you actually just got to buy a team all by yourself. But I think that the timelines for getting productivity out of these people are going to be a lot shorter, as we've seen recently with Google, with having some of their brain trust kind of depart and go off to found their own companies that are a little bit more focused in certain areas.

They're not going to wait for results in a market that's changing so rapidly. They can't afford to. So wait, Alan, why wouldn't the card just go get in line with everybody else at the NVIDIA trough and just get funding from NVIDIA versus doing a deal here with Anthropic?

Because well, NVIDIA seems to be handing out dimes like Carnegie, except they're- Rockefeller handed out the dimes. Right. Carnegie built libraries.

But that being said, first of all, there's no guarantee that world models need GPUs. Right? It's a different mousetrap.

Mm-hmm. Secondly, another lesson I've learned in business, same company, Exodus Cable Wireless. They offered a company I helped do.

We were public in the dotcom. 50 a share. 66 to $56.

We were coming back down. 50. 50.

Our board voted to turn that down because it was just a momentary blip. The good times were going to live on. They'd be happy days here again forever.

We turned that down. We went bankrupt about a year later. When someone offers you $6 billion, son, take the money.

Take the money. What are you, crazy? That's $6 billion.

Jerry Maguire, show me the money! Absolutely, Jerry. Jerry was just like, yeah.

Look, we're talking literally give me 1% of that and I'm good. I'm good too. It's $6 billion.

$6 billion? Come on, man. It's $6 billion.

It's a no-brainer. At the end of the day, I do think the one-liner here is this, is that the story, the headlines that came across over the weekend, it's more like the model race flattened to near parity. So now the fight is to move to cost to serve.

And whoever serves intelligence the cheapest without breaking trust is going to be the ones that are going to write the next chapter. That's basically it. But they may not get the next check because they drove the price down so low.

So take the money now. Take the money now, and that again is the indispensability trap. Right.

It's the people who build on top of it. As we mentioned last week, though, if per chance that there was a collapse in the stock market and the bubble did burst, there'd be a lot more cheaper AI resources out there, just like fiber became a commodity, AI will become a commodity, and maybe everybody in the enterprise will do a happy dance because I don't have to pay you. That may be just what happens.

It may be, and if there's a collapse, it'll be us taxpayers are going to be bailing everything out. So after we bail them out, everything will be cheaper. But that portfolio that you are frothing at the mouth at, that looks super great now, is going to be in the negative.

So Mike, I got two things, and we got to wrap up, I want to mention. Number one, I think I mentioned it last week. We've been doing a whole bunch of special reports on Techstrong.

These are anywhere from 12 to 30-page reports in depth on a lot of these subjects. There's one coming out this week on the AI mirror, three potential outcomes of AI. One is Anthropic continues, and the rest of them continue these valuations.

The second is all hell breaks loose, and the third is maybe something in the middle. Last week, though, or maybe it was the week before, we released a special report on world models, because regardless of this deal, there are a lot of people who say world models are what's needed to get us to AGI, super intelligence, and for robots with AI embedded to get real. " And you can download that right now, not even a registration, courtesy of your Chief Content Officer at Techstrong here, Mike Vizard.

Until the cost of AI keeps going up, then we're going to have to do something. Then we're going to have to do something about it. Anyway, guys, we're about out of time.

Tom, Yvette, thank you so much, as always, for coming in. Tom, I know you stepped in on short notice today. Appreciate it very much.

Mike, as always, pleasure having you here. You'll be back tomorrow. We'll be back tomorrow with more, so stay tuned for that.

But for now, that's going to wrap today's show. We hope you've enjoyed it. tv, Techstrong TV YouTube channel, Techstrong TV OTT channel on Apple and Android, and what else are we on?

Roku and Apple TV and Amazon. If you've got a screen, you could watch Techstrong. How's that?

But until next time, this is Alan Shimel. Thanks for watching today. We're out.