AI Infrastructure Race: Alliances, Power Fights and a $7B Deal

We're in the show. Hey, everybody. Welcome to techstrong gang.

It's Tuesday, and as always, we have a fair assortment of AI-related things, but boy, every time I seem to turn around, somebody's doing something to somebody, and the soap opera continues. Let me welcome our guest today, Tom Hollingsworth. How you doing, buddy?

How's things? They're good. It's been so long since I've seen you, like a whole day.

Tom might be approaching a regular status. Who knows? Sid, how you doing?

Good to see you always, my friend. Yes, sir. Good to see you, Mike.

All right. Chris Blask, I can't tell by the background, where are you? Still in Canada, hacking another space together.

All right. Well, then there's a whole new venue. And then, of course, Kate Scarcella joining us who, we just call her Ms.

Right. How you doing, Kate? All right.

Let me get started with this thing. So, apparently the United States government is drafting a letter to warn its allies that when they committed to being part of something called PAX Silica, that that included AI models, and that therefore they must lean in towards the AI models provided by the United States. At the same time, though, there's all these open weight models that are coming out of China, and I think the latest one I saw was out of Alibaba, claim that they already have a million downloads for this thing.

And I guess I'm going to kick this to Chris, but have we stumbled into something that is largely now an AI Cold War? Is that what's going on here? And is there any middle ground?

Well, there sort of has to be middle ground, right? So you know I'm a cybersecurity person. I've spent the last decade in focusing primarily on supply chains, right?

And this is AI alliances increasingly becoming a supply chain alliance, right? What dependencies am I going to build into my systems? And dependencies are a real thing, right?

And there are, as we discuss here weekly, the other big platforms aren't going away, big nations aren't going away. Small countries can't build some of the same infrastructure. But you need to know wherever you are, what dependencies you have, how you maintain operations, resilience, and security, and everything that matters to you should those suddenly become unavailable.

Because, again, as we discuss on this very slot every week, the US government turned off a version of Anthropic, and you're Switzerland or Kenya or wherever, right? So is it a matter of choosing sides entirely and putting all your dependencies in one basket? That doesn't sound like good thermodynamic management to me.

I think part of the issue too is the choice seems to be, I don't know, Hobson, but Tom, I'm going to kick this to you for a minute. The concern out of the models out of China is that they have been tuned or trained in a way that is spreading ideology and propaganda and maybe malware, who knows. And then this other side of it is a concern that somebody in the White House is just going to turn off access to AI whenever they decide that somebody's upset about some other policy decision that has nothing to do with AI.

And I can't help but wonder if this is just going to push more people towards maybe looking for alternative sources for AI models. What do you think? I think it absolutely will, and this is not the first time that we've seen the US government do this, right?

Just think about any kind of defense contractor situation. We've dealt with this with allies before, where they've gone and bought surface-to-air missile systems from Russia. And we were told, "No, if you bought the SA-400, then you can't have our Patriots, because even though they're different technologies, we can't let you do that.

" And I think part of it is they don't want anybody to do a real bake-off between the two to understand which one's better, where the pluses and minuses are. But I completely agree with you that there's apprehension because when you've locked somebody into something, you have the capability of directing how they perceive it, right? I couldn't imagine anybody creating an AI model that is deferential to a specific group of people or has influence from a certain group of people, and definitely not if the creator of one of those models spends his every waking minute tweaking it so that it doesn't tell the internet what an idiot he is.

No. That could never happen, ever. No, this is the same gamesmanship that we've seen about a lot of things, right?

We have to have an adversary to fight against, and we can't take the chance that people can think for themselves. So we effectively lock people out from the beginning by saying, "No, you have to pick a side here. Don't do the comparison work.

" Yeah, because the Russians never stole any technology from us back in the '80s, like Baron or any of the other shuttle projects that they had or any of the other fighter plane projects that they had, but we've never stolen anything from them either, right? No, we could never do that. That's the problem here, is we don't want everybody to see that it's really just a house of cards under the hood, and we're stealing each other's best ideas because when you put them side by side, you'll figure out exactly how identical they are.

Right. So Kate, is there any rumbling in the open source community that you're hearing about all this? Because you would think that they would kind of step up in the middle of this and say, to Tom's point, if this is not really differentiated technologies, then why are we all running around trying to build all these various AI models when maybe we should just build one or two that are good enough for everybody?

Well, I can't say that I heard from the open source, but what I have heard is from Bloomberg Surveillance, and at the end of the day, what you find is it's about the economy, stupid, right? It's just about the money. And that's what we keep finding out, that is something good enough.

And I hate to think that we're like that, but we almost need something, we need it fast, and is this good enough? Is this going to get us to where we need to be before we apply the additional 10% to make it maybe a little bit different to fit our project and what we're trying to do? So that's it.

At the end of the day, it really is about the money. And just as what Tom said, they're not trusting us at all. They are not trusting in the United States government.

" And that's what companies can't have, so. Yeah. So can I jump in here?

Sure, go right ahead. Anytime. So I think there's three things here that we need to think about.

One is that AI is becoming the substrate for geopolitical power, right? That's very clear here. The second point, I think, is the real race is infrastructure stack.

And the third point is the US is muscling itself back into the game when China was winning with the model game and making that the core theme. So the US is saying, "No, this is not just about models, it's about who controls things like critical minerals, semiconductors, manufacturing, accelerators, networking, and all of that," right? So I think that's where the game is shifting.

So while China was trying to make this all about models, and we're going to race to the bottom with our deep seek and other models, EI, and all of that, it's going to be a different play. " And I think the other point is countries are being made to choose and decide which AI ecosystems they want to operate within, right? So people are being forced, other countries, whether it's, I don't know, could be Europe or India or whoever else, it's like, now you've got to make a decision.

Who do you want to align with? So I think that's another interesting dynamic that we need to think about. So to Sid's point, Tom, I'm going to come back to you, but it is becoming a larger game, right?

China and the US are now conflating not just AI models, but precious metals and all these other things that you need to go and build infrastructure. " I think it's actually a point that's coming pretty soon, because this is exactly how the original Cold War started, right? We had two sides, and you had to pick which one you wanted, or I don't know, Brezhnev rolled tanks in or something like that.

But more importantly, that created its own little ecosystem for people, right? Because we saw that with Cuba. They were basically being propped up by the Soviet state for the longest time, because that's basically how it all worked.

And I think that China is going to start doing that with AI technology, with electric car technology, with access to precious metals, with support for just about anything. And it's the best weapon that they've got, right? You can be with us, or you can choose them.

And because you're in our sphere of influence, we can effectively isolate you. And that creates other economic problems for them. And that's really, I think, where a lot of this conversation comes back to.

This isn't about an AI model. This isn't about a precious metal storehouse. This is about effectively creating a partnership that will help you weather any upcoming economic storms more than any kind of military action.

I think that the next World War is going to be fought in that realm, and it's going to be pretty brutal if you pick the wrong side. But right now, we don't know which side is the right one to pick, and the people aren't doing a good job of helping us differentiate between the two. And we are.

We're entering into an AI Cold War. If you look at everything that preceded World War II, I think we're totally entering into a world war, and what does that look like? And we're not getting it, as I've said on the show about this asymmetric type of power that we're not understanding yet.

As far as how we're fighting it, I think it's going to be drone-like. But AI, and since we're on this topic, what I like about the one model that China's talking about is that you can have a local model. So back to Quiet Wire and Chris.

Having that local AI model is really, really important, and it helps build this base that we need. " And I think our friends in Canada with Mark Carney is one of the ones who has started this. Mm-hmm.

In reality, all times are unprecedented, right? And I love history. Always looking at where we've been to this point, where is the trajectory taking us.

History does rhyme a lot, right? And I agree with what everybody said. There is a huge global conflict going on.

But just walk around the differences between this and 1930, or World War I, or other conflicts in the past. It's cliché to say it, but again, we have spaceships and things now, and you can communicate with people more than 5 miles, 50 feet away, is huge. So from a resilience perspective, from a country that is not the two we're talking about.

There's 300, 400 countries out there. Almost all of them are not US or China. Almost all humans live in countries who aren't those two countries.

And they're sitting there right now in whether it's the third largest country or the smallest, have the same AI tools that everybody else does, right? And they can say, as they choose, people there, the government, the folks can say, "I want to use these AI tools to see what resilience looks like to me. " And these are the sort of capabilities that even in the West, even in the US, over the last 50 years have been hard to do.

Shout-out to the late great Jim Peurifoy and folks at Argonne National Labs 26 years ago, who did one of the few holistic infrastructure dependency models for the United States. Water, power, transportation, everything else. Because it's hard.

Now, the smallest island nation, if they have the right folks, if they use these AI tools reasonably well in this era, in this next one, two, three years, can put together a national plan that says, what do we do if the US and China both cut off our AI models, for example. And maybe have plans that change this game so it isn't the Cold War again. It isn't just picking sides.

Maybe it's a different game. Yeah. Is there any irony in this, Tom, to you?

Because especially from the US perspective, I look at this and I go, didn't we just spend the last year and a half or so telling all these foreign countries that we were cutting off aid? " So at some point, the leaders of these other countries must be laughing at themselves because now they're looking at it and going, "Well, I got the US on one side willing to pay us to use their AI models, and the Chinese are willing to pay us to use their AI models. " No, it's flipped on its head because the person who's currently leading it has a transactional view of the world.

If you want something from me, you have to give me something in return. There's no such thing as an altruistic kind of bond between two peoples. There has to be a winner.

" But unfortunately, we are where we are, and unless we can replace the president with an AI, I think we're kind of stuck. Yeah. I think it's definitely the new world order where AI race is going to be the next battlefront.

But I also think in the interim, we ought to think about how existing battlefronts are going to be energized and enabled by AI, right? So you think about weaponry, military. You think about the world economy, world currency, geopolitical alignment.

All of those things are going to be AI-enabled. So the question then becomes who is going to be winning that race, right? So if I'm building drones, is it AI-enabled?

If China's building drones, are their AI-enabled drones better than our drones? I think those are the kinds of things that are going to be important going forward, right? So not just the AI race in its sort of vanilla form, but where AI creeps into every day interactions between countries that are going to be important, right?

And that's where the battlefront is going to be. But you know what's sort of funny to me at the end of the day, is how Kazakhstan, we've had Europe, we've had these people actually joining with China. They're actually saying, "We hear what you're saying, United States, but you know what?

" So there you go. Well, that's Tom's point, right? If we've got a leader who's basically using bullying tactics- Yeah ...

to force people to make a choice in a certain direction, right? So in the case of Europe, even if they want to lean towards a certain way, they cannot completely do that because of other dynamics that we just talked about and Tom highlighted. Yeah.

So that's the thing to keep in mind. So what is the probability of this scenario playing out? And I'll throw it to Chris.

Basically, China's giving open weight AI models away to anybody who will take them, and they're funding that and passing that along. " So then essentially, United States government and the taxpayers wind up funding the adoption of AI models in various countries around the world, while US citizens are paying for the privilege of using those very same AI models. So at some point, does this just get completely ridiculous?

I could see something incoherent like that playing out. And I don't joke. So this is taking us right into B block, right?

So this is becoming physics, right? Economic physics. If a nation state, like for example ours, the US, chooses to make it a black and white thing and require dependencies, and, and, and, then we can find ourselves in positions that aren't econ- You're on mute.

Yes, I lost mic. You got to stop waving your hands, man. Right.

But- Danger, Will Robinson. Yep. So, and again, this is my last comment.

We're a bunch of geeks. We spend a lot of time talking about model weights and everything else, and we've forgotten what these systems do. AI systems are really, really good at coherence.

You can throw something at even a crappy AI system and say, "Does this make any bloody sense? " And it'll fill in the gaps, and it'll make things make sense. These are the things we're talking about.

I know I can take your question and put it into an AI system, anyone, much less the sort of good ones I like, and the answer is, no, that doesn't make any bloody sense. We can debate it as humans, but A and B do not equal C. Yeah.

It's not the way things work. But resilience, this is in the grid. And again, B block, I'll let you kick in after I intro, after I start, but we've been building systems for the last 80 years, post-World War II.

And in the US and North America, grids and power systems and everything else are getting integrated. We have industrial economic systems that have done quite well with externalizing a lot of cost and dependencies without really putting it together, Kate, Cleveland, and environmental stuff. We're at a point now where we can add up all those issues.

We can solve things like environmental issues and economic issues, but they have to make sense. When we're doing things just because we used to, or we're loud, or a crazy politician wants to do crazy things, and they don't make any bloody sense, other people just won't buy into it because they don't have to. All right.

Well, we keep referring to the B block, so I'm going to shift that gear right now because we've been talking about politics on a global scale, but as Tip O'Neill once pointed out, all politics is also local. And what we have now is a bill in Congress that somebody's introduced to create something that feels like a data center bill of rights for local communities so that they can sit there and say, "We have the right to refuse whatever data center project it may be," and all the implications that go with that. And this is happening against the backdrop where data center projects continue to be funded.

There's just huge projects that are out there that OpenAI is throwing money behind, and this all looks good on paper, but then it also turns out, well, we don't have enough power to support any of this stuff. So, we got more paper data center projects than we'll ever be able to find any energy for. So Tom, I kind of want to ask you, what's your take on what's going on here?

Does this bill, A, have any legs, or is it probably just going to die in the vine, and it's just theater, and C, or B, is this all a non-starter? Because in the end... Oops, we lost our host.

Well, Mike, I'll go ahead, and I'll start off by talking about the fact I honestly don't think this bill's going to go very far. With all due deference to Congressman Connor, this is kind of performative, and we know it is because, again, all politics is local, and the bribes at the local level are much cheaper than bribing an actual Congress person. Allegedly.

I have to remember to add allegedly- That makes sense ... to these things now. But I don't understand, why do we have to build these data centers?

Because if this was a prison, or a wind farm, or a solar farm, suddenly there'd be a good reason to reject it. What's so special about the data center, other than the fact that, allegedly, it's going to bring so much extra revenue around here? Couldn't be anything to do with the fact that the companies are either threatening to move it somewhere else or encouraging that maybe the county commissioner who's going to get voted out for approving the project will get a nice cushy job somewhere where they can receive some stock options or something like that.

I think that ultimately, this is the problem, is it feels very crooked. It feels very backroom deal kind of thing. And realistically speaking, to your point, why should we spend all of this money to build all these facilities that we won't be able to power for three or five more years, that we won't be able to hook up to any kind of an infrastructure?

It couldn't have anything to do with the fact that all of this money's already been kind of allocated for these things, and if we don't do that, then the shareholders might get mad that their investment only grew 8% year over year this year instead of 10% year over year, and then therefore, they're going to get really mad and might sell the stock, and then people won't be trillionaires anymore. I think that we have to take a really long, hard look at what's going on. Just think about it.

We're stressing the electrical grid. We are running roughshod over communities to just basically clear land as fast as we can to put up a giant Amazon style warehouse that nobody's going to work in, aside from what? 15 people to just keep the lights on and sweep the floors.

RAM is insane. Hard drives are insane. All of our computing cycles are being pushed to their absolute limits.

I even saw a story yesterday about how China has ramped up production of synthetic diamonds to be used as a cooling substrate for these things, so there's a possibility that the diamond market could crash. " We aren't adopting consumer focused AI fast enough to make these build outs necessary, because I know that nobody out there wants to hear it, but there's only so much AI use that enterprise can do before it stops. And we have this mentality where we have to build things bigger and stronger and faster, and wait a minute, I need to go check on that dark fiber that we buried 25 years ago that was going to be the solution to all of my problems for the internet, or all of those rent houses that I bought in 2008 because you know the market's looking up, and if I don't buy houses now, I'm going to be left out in the cold.

And if you know what the first two things have in common, you know what the third thing we're talking about right now is going to have in common in a few years, when suddenly all of that compute power and all those data centers and all of those things don't matter one bit. Yeah, I think- Exactly. And then...

So, I think this is more than this. It's about a social contract with the community. Nobody complained when Verizon put in a 5G switch in your community.

Why? Because it's the scale. It's the amount of stress it puts on the existing community's resources, whether it be power supply, be it...

They're talking about data center in my town. Heck, 90% of people have a septic tank in their backyard. We just don't have the resources to support a data center.

So bringing communities into the conversation is extremely important. Where are these data centers being built? Who's paying for supporting the infrastructure?

What are the economic benefits? What are the impact of the environmental and quality of life? So I think that's the contract that is more important than the data center itself.

And the second thing is, it's no longer about just GPUs and compute power. It's about the amount of power draw. It's not about who can buy the most accelerators, but who can actually secure reliable megawatts of energy required to run these massive accelerator clusters.

So there's a data point in one of the reports you sent that says US developers have submitted requests representing 1,066 gigawatts of data center power demand, whereby the utilities can only supply about 28% of that. Where does the rest of the 72% power come from? Those are the questions that one needs to ask about.

So I think the new metric that we need to think about is economic output per megawatt. Okay, you want to put a data center, it's going to take massive amounts of power in terms of megawatts, but what is the economic value per megawatt that's going to be invested, both from the local economy perspective and the state economy or whatever it cascades into. So that's really the metric that we should think about.

So I'm not sure I agree that this is completely performative because you're starting to hear Republican communities complain about the same thing. And it's coming up because people are noticing as they bring their kids back to school that, for example, the price of the laptop has gone up, and that's part of the blame of the data center. So, I don't know.

Tom, do you think that there might be some sort of bipartisan push at some point because Republicans represent poor districts? Mike, it's adorable how you believe that there can be consensus on something that's controlled by business. And I'll give you a perfectly good example of this because, unlike the rest of the country, I have had a lot of experience dealing with a corrupt organization, and they're actually building a turnpike about two miles north of my house.

And they're wiping out a whole bunch of houses, and nobody wants this thing. Zero people care about them putting this turnpike through, and yet somehow it's still being built. People have showed up to community meetings.

They have voiced their concerns. They've talked about how corrupt the fact is that the current sitting governor is actively pushing for this, and they know he's going to end up with some kind of a contract after this is over with. And yet somehow, people that I know are being forced to move out of their houses, businesses that have been here for decades are being told that they have to sell off and move.

How is it that something that's so insanely unpopular is getting done anyway? And the answer, of course, is that the right money is going to the right people. So yes, we could pass this law.

Yes, we could empower local organizations to do these kinds of things. And I promise you, everyone, not only on this call, but listening to this podcast, can think of at least two things in your town that nobody wanted that somehow still got passed anyway because, well, someone put a survey in front of us, and it really is important. And pay no attention to my ties behind this economic development lobbying group.

That has nothing to do with why I voted to support this. I want to put some hope in the end of this segment, right? So the great futurist, author, friend of the show, David Brin.

My favorite book of his, "Killing People" It's a fascinating book. But one of the things he built into that is in the not too distant future, organized crime and conspiracy gets really, really hard. And it's not a main plot point, he's just extrapolating, and I discussed this with him when I met him through you guys.

But I always thought that's sort of obvious. And what I hear in this, I don't disagree necessarily with anything anybody said. I think it's a little more complicated than that, but what I hear is absolute distrust in the system because we know we can't see.

Tom, what exactly is the reason for that turnpike, this data center? Who was really consulted? And we have lost faith, again, like the Prime Minister here said, the post-World War II era is over.

Part of it is that we thought we could build these institutions and have faith in them, and we didn't. And we thought we could have democracy and free systems and free speech and really see what's going on. And it turns out it just takes too much bloody time.

And this is one of these things that I agree with, with David Brin. I think one of the things we find in the next handful of years out of AI is that individuals can find those things out. You don't need to take six weeks to get a response to get the PDF for the one decision in the one council in state, local, municipal, whatever it is.

And we're working with people today in the legal world who are showing that you can see through this mess of documentation bureaucracy to see what really matters to you in the time you have actually to do it. And I think that changes things. Tom, as you were saying at the start of this, I think about people who are really benefiting from the corruption embedded in the system.

Think wisely, because in three years, you might find out that everyone around you knows exactly what happened, and that your cover of taking too long for people to figure out and not being worth their while goes away. And I'm looking forward to that. " And I thought, good for them.

I've always liked them. Good for them for pushing back, asking for the real analysis on how much power they're going to really need. And they're coming back and actually discovering that they don't need, big surprise, as much power as they're requesting.

And so, hey, let's leave it to them at the end of this segment. Yay to power companies. " Well, here we are.

Yeah. So I'm going to shift, though, to a C block, which may offer some hope in this conversation, so let's see where this goes. But what we have seen now is that Stripe is acquiring a company that provides essentially a router to LLMs.

It makes it easier to swap them in and out as you need be. And they're laying down a cool $7 billion. I'm not sure why Stripe felt the need to go buy this company, but it raises the concept and the notion of, well, are these AI models essentially going to become disposable?

And we can just swap them in and out as we need them, and therefore, the race to zero is now on because the only way that they're going to compete is by lowering the cost of these things, and that will drive the choice of whether or not I use one versus the other. Sid, what do you think? Is that possible?

Is that where this is all headed? Yeah, I think the deal validates an important shift in the AI operational model. Enterprises are becoming less interested in putting all the eggs in one model or one basket, if you will, and are interested in multi-model consumption.

Interested in infrastructures that can dynamically select these models based on price, performance, latency, and all the availability and all the good stuff. So I think this goes back to the conversation we've been having for the last, what, three or four shows. It's all about the control plane, stupid.

So that's what it is, really. So you asked why did Stripe buy Open Router? It sits between the business and customers for payments.

So I think Open Router could put Stripe exactly between, hello, applications and model providers. And I think Open Router provides that sort of access to more than, I think, 400 models, I think the article said. So that's a massive amount of sort of gateway power, if you will, for lack of a better term.

And I think model commoditization could make this whole idea of, and especially with the first segment we talked about the prices of models falling rapidly with China coming into the game and all of that. So I think model commoditization is actually happening. So I think the router is going to be critical, and that gateway functionality is going to be key.

So the company that controls the routing layer would gain significant advantage in this game of AI-infusing workloads and AI spend, and so on and so forth. All right. So Tom, if our votes won't matter, will our wallets make the difference in this conversation?

Yeah, of course they will. I think the key here is I would like to congratulate OpenAI and Anthropic for doing this to themselves. Here's why.

Originally, we had an unlimited all-you-can-eat buffet. We can do whatever we want with these models, and yeah, Anthropic's not making any money. Who cares?

Well, now the people who are backing these companies are like, "Whoa, wait a minute. " So what did they do? They came up with a way to monetize the token.

People who are using more tokens are going to have to pay more. It is a brilliant way to monetize it. But what happened was is that as soon as you make the token the unit of work, now you've created an arbitrage situation where someone can say, well, if a token doesn't necessarily mean a token in every place, then if I can make the tokens make more money in other places, then I can make money off of that delta.

So the Open Router thing is kind of brilliant. If I can take tokens or units of work from one location and spend them in another location and save you a little bit of money, then I can take a little off the top to get paid on that. And when you are dealing with an all-you-can-eat buffet, that's like buying a gift card to Best Buy.

I can have all the money in the world, but I can only ever spend it at Best Buy or Amazon or wherever. This, for Stripe, conveniently enough, is like when you have a Visa gift card. Yes, it's probably not as effective in the long run.

You pay more fees, at least the processors do. But it's valuable everywhere. So what Stripe is effectively saying is we have a platform that can process millions of transactions, a huge volume, which is something that would be important when you're trying to farm out these tokens to go have work done.

Instead of worrying about, am I going to do this with Anthropic? Am I going to do this with Google? Whomever.

Let us worry about that. Tell us what your constraints are, tell us what kind of things you want to be able to do, and we'll just take a little bit. Come on.

What they're going to do is, like any other payment processor, they're going to charge the AI models 1 or 2 or 3%, or whatever it happens to be, for consideration in the system. And I'm sure that as the companies either build out their capabilities to make one look more attractive than the other or, Mike, like you said, cut prices to make it look like more traffic should be steered over here according to Stripe's algorithm, that that will create a whole kind of bidding war behind the scenes. But the important thing for the AI companies to remember at the end of the day is you did this to you.

All right. Well, Sid, walk me through this a little bit because theoretically, this scenario should have already played out in the land of cloud computing, right? We have all these standard virtual machines out there.

Most people are using a small variation of the same type of one, and yet the price of a VM in the cloud, regardless of who it's from, seems remarkably stable, and it doesn't continue to decline as rapidly. In fact, it's probably going the other way because of all the things we've talked about in A block. So is this going to be a race to zero for the LLMs, or will they somehow or other magically come up with the price of a token, to Tom's point, doesn't actually move dynamically in the way of a market, it just kind of suddenly is constant?

Yeah. Ironically, this begins to look like the evolution of cloud infrastructure that we've witnessed. To your point, Mike, customers ultimately want portability, price performance, price competition, abstraction from underlying suppliers.

Right? In the world of cloud, we have the best cloud instance, right? Whether a smart instance or whatever instance, right?

So I think it's going to become more and more of that. So whoever can build that multi-cloud management system that we talked about back in the day when we worked in cloud, that can be the broker across these multiple cloud providers and get the best value for the best cloud instance for a given workload that's running at a certain time of the day for that particular transaction is going to be the winner, right? So I think that's where this is going to head to.

However, there is a contrarian view to this whole notion, which is I'm reminded of my networking days when Radio Forum had built a multiprotocol router that bridged across IP, IPX, DECnet, DLSW, and AppleTalk, right? But in the end, IP won and the rest of the protocol did not even matter, right? So the question is, is this a dominant model that ends up being the working model or one of two models?

We don't know whether that's going to be the case. I'm just presenting a contrarian point of view that all this will not matter at all, right? I'll just jump in and say the contrarians are wrong, right?

Again, I think of models as units of torque. I can replace a unit of torque in a device with a different one, a different electric motor, gas motor, diesel motor, whatever. They all have characteristics, but it's all about doing something with it.

Right? We've over-focused on the models, but I agree with Sid, and I like the way both of you just said that because Tom there's nothing more to say but yes, and effectively probably good. Right?

But my next question, of course, is there just going to be this one control plane that we all have to deal with that's going to abstract the models, or are there multiple? And can I see exactly what they're doing, and can I compete my interest in using them based on their ability to let me know what I need to know? Or is it going to be another opaque control plane that gets back into block A and B?

Well, there's more, and I'm going to toss this to Kate. So what are the odds that the United States looks at all of this and concludes that this is a race to zero, and we might as well accelerate the merger of OpenAI and Anthropic because this is a global concern, and we're competing with nation state funded organizations. And while we're at it, we can also stabilize the cost of a token because, well, it's in our interest to make sure that remains artificially high.

And for a kicker, the United States will take a 20% stake in the entire combined equity. What do you say? Yeah.

That's the way our wonderful capitalistic market is working right now, right? Is taking stake in companies, and honestly, I- I thought that was socialism. Sorry.

Thank you. It's allowed. My point exactly.

And I think one of the problems is that nobody's telling our leadership what is happening from a technology standpoint, and so I think they just don't understand. And at the end of the day, us having the ability... This is what we saw with cyber mercenaries, right?

When it comes to cybersecurity, that our borders are almost gone. And the way that we think about cybersecurity is not necessarily, even though in Europe they keep trying to set up these boundaries and everything else. But I think at the end of the day, as long as I'm able to communicate with whomever I want to communicate with, it almost just takes these borders and just crushes it, and everything becomes an open border.

And so with that, our leadership in the United States can definitely do that. I don't think it's a wise choice because I think it's like we have, I would consider this almost like a Visa platform. Look, you get Mastercard, you get Discover, you get...

So we always have, thankfully, multiple choices. And at the end, I think this all becomes more infrastructure, and that's what we've been talking about most, right? On this show, we talk a lot about how all this is becoming infrastructure.

And really, whoever laid out all the electrical lines and everything else, it just became a part of our working, everyday society. And I think that this will be the same way, and I think the ability to enhance it, make it more robust. This is something where I think we're just so stupid, where I think that we can really help by opening up, giving us the ability.

I always talk about the two guys in the garage. We want that type of innovation, not innovation where government is just pouring money into something that, my gosh, if it doesn't work, get out. Just let us not fund a failure.

And this is what I feel like. We fund failure. Let things that work, let it thrive.

Yes. Was it just this week? The CEO of Politico, I think it was, I saw it and it has been mentioned to me, so I think it's resonating.

" Yeah. Because six weeks from now, I'm right there with him. I love forecasting, but starting to get into eight quarters, 12 quarters, 3 years.

Yeah. The shape of that, you have to have an idea in your head, but moving at the speed of government in today's world? Okay.

Good luck. Yeah. Good luck with that.

I think the world will move faster. Totally. We have to adjust to that way of thinking, Chris, everywhere from government to business, to even us as individuals.

A year plan? Are you kidding? How much changes in a year for us?

Everything is so different, and if we don't embrace the different, we're going to be left behind. Yeah. This is the turn of the century, just as much as the internet.

Because the internet was around before that, but the turn of the century, it really literally changed everything. Geopolitics, physical, right? The world is-- You can see it from space.

This is one of those times. Yeah. For the next couple of years.

Yeah. I think we as a nation are developing AI paralysis as a consequence- Yeah ... of the competition with China.

What happened to good old American innovation, right? Exactly. Why are we responding to China because they're taking a model and bringing down the cost of the model?

Why can't we build a commoditized model, right? Yes. And build a whole ecosystem around it.

Yes. That at the end is going to win. Guys, I got to end this conversation because I know it can go on forever.

I would just say, it's a complicated world, but if you start voting with your wallet and work backwards, it gets a lot simpler, a lot quicker. Hey, guys, thanks for sharing your insights as always. You guys were awesome.

I'm sure a lot of people now have their eyes crossed because there's so much to think about, but hey, that's the function here. That's what we do. And I want to thank you all for watching the latest episode of the Techstrong gang.

Please stay tuned for the rerun of the rest of the Techstrong TV lineup, and as always, we'll see you, God willing, tomorrow.