The $800 Billion Threat: China’s Plan to 'Old Navy' US AI

The top open source AI models used to be American.

Now that is changing.

Today, the top five open source AI models are all Chinese.

And that is not the only story of AI evolution coming out of China.

Chinese companies are also showing that they can do more with less.

DeepSeq is using a new format that has allowed their training models to run 30% faster.

Meanwhile, Alibaba last week said that its new computing pooling system cut the number of GPUs needed to run its AI models by 82%.

So this is...

quite significant, Scott, following the conversation we had last week about how we're seeing all of this data center demand, and we don't really have enough energy to power the data centers.

Energy costs where data centers are being built have tripled,

And as a reminder, if you want to build AI, you need to build the data center that powers the AI.

OpenAI wants to build a chip network that would consume 250 gigawatts of energy, which is equivalent to a quarter of America's entire grid capacity.

Point being...

a lot of AI and not enough energy to power the AI.

And then we ask this question, okay, well, then maybe we need to figure out a way to either get more energy through solar or nuclear or perhaps drill, baby, drill, or you figure out a way to make the AI models more efficient.

You figure out how to consume less power, how to do more with less.

And it appears that

that China is making some headway on that front.

Alibaba just last week announcing this new system.

It cuts their GPU usage by 82%.

Basically, they can use 200 GPU chips and they'll perform at the same level of 1200 GPU chips.

Don't ask me why or how they can do that, but that is what they're saying.

It appears that this may be a trend in China.

So Scott, your initial reactions to China winning the efficiency race or so it appears.

I was thinking in the editorial call yesterday, when we were going over the stories, I kind of got my mind blown thinking about, as I heard you guys talk about AI and the energy requirements and open source-less, open weight models.

And I've got this, what I think is this, I don't want to call it a revelation, but something that I thought was really exciting.

And I thought, wow, maybe we do see something here in the matrix.

And that is while everyone's talking about the gating factor is going to be energy and that there's no way that there'll be enough nuclear power coming online to quench the thirst of Sam Altman's unbelievable juggernaut called OpenAI.

And I thought, well, maybe he's just trying to manifest it in these big energy projections that are running up the stocks of electric companies and scaring everyone that we're not going to have enough power.

I wonder if, well, innovation can come from the other side, and that is people can start producing chips and LLMs that just require dramatically less energy.

And then the other thing that sort of blew my mind, and now I think as I thought about it and stepped on it, I think it's actually a pretty decent

thesis is that if you look at the US economy, it's run for profits and shareholder value.

The life in America is so much better if you're wealthy than if you're middle class and your life in the middle class is so much better than if you're in a low income household that we make a series of incremental justifications every day where it's all about, America is now all about the shit you have or specifically your ability to buy more shit.

Our economy and our society is run for profit.

The Chinese economy is run for control and also long-term geopolitical power.

They make 50- and 100-year plans, which we would never have the ability to do.

And part of that is because we pay the price for having governments, democracies that turn over and sometimes go zig and zag.

But what I think is going on or what I think is going to happen is the following.

China is really fed up and sees America as their enemy or is sick of America having this sclerotic trade policy that is really damaging them and then saying, you can't have our chips.

And they've said, okay, we've had it with these guys.

So what do they do?

They're much more strategic, and quite frankly, Xi is 10 times smarter than Trump, and also has this incredible weapon in that he's willing to kind of think middle, think kind of medium and long-term, which Trump has absolutely no ability to do.

And I think what they've decided to do is the following.

I think they are going to flood the market with cheap, open-weight models, AI models.

And if I were advising Xi and I said, okay, if you think of America as an adversary or America,

Essentially, America has become a giant bet on AI.

Specifically, 40% of their entire market is from the valuations, the exceptional valuations of 10 companies related to, directly related or related adjacent to AI.

And if we take down, if those 10 companies go down 50, 70, 90%, like they did in the dot-com era, they are in a global recession.

Trump is out of vogue.

We don't even need to stop buying soybeans.

They are in a world of fucking hurt.

How can we do that?

Pretty easy.

Let's start pumping out a bunch of models that require less electricity, less power, less processing power, and are 90% as good.

And another personal anecdote, my first strategy client, I started a strategy firm called Profit when I was 26, and my first client was The Gap.

And they wanted a new brand.

And we came up with working with this really intelligent guy named Mark Bucco.

Was that a strategy there at the time?

We came up with what I thought and ended up being a great strategy.

And that was, I just read Peter Drucker's book and he said, every major business shift in society is largely reverse engineered to a demographic shift.

And as someone who had personal experience with this, I said, one of the biggest demographic shifts in America, and this is 1993,

is the explosion in single parent homes.

And it's almost, when we say single parent, we mean a single mother.

And single mothers, we did some research in some focus groups, are very cognizant, self-conscious, and focused on dressing their kids well, because they're insecure and self-conscious about the fact they don't have as much money and the dad isn't around, and they want the kid to feel good about themselves.

And so we identified this segment, and it was this fast-growing segment, and we said, okay, here's the positioning.

80% of the gap for 50% of the price.

And that positioning took Old Navy from zero to a billion dollars faster than any retailer in history.

And that is probably, if you look at the fastest growing businesses in the world, oftentimes they're that 80% of the biggest airlines, America, Delta, and United, for 50% of the cost, that's Southwest, now the most valuable, or was the most valuable domestic airline.

I think that the Chinese tech sector, under the direction and encouragement of the CCP, is about to Old Navy the shit out of the US economy.

And they've done it across BYD, electric vehicles.

Generally, that's their business strategy, whether it's cell phone towers or clothes.

But I think they are going to spend a ton of money and time and put their most talented scientists, of which there are a ton, to work on the following.

Let's fucking go.

We've been going for the heart and lungs after red states with these –

trade policies, whether it's rare earth minerals or canceling our contracts for soybeans.

Now let's go for the fucking jugular and let's neuter their AI industry, specifically the valuations, and let's release a ton of near premier quality

LLMs, open weight that anyone can use for near free, that maybe requires much energy, but most likely don't, and certainly aren't nearly as expensive.

And let's fuck with America's big bet here.

Let's make this big bet

not pay off.

I think that is exactly what is happening.

I would add that part of the reason why this has happened is because of the export controls that we put on chips to China.

And this was the deep-seek story that we discussed at the beginning of the year, where China had less capable and fewer chips to work with because of these export restrictions.

And it

forced them to get leaner and more efficient and work with what they had.

That is exactly what they've done.

And I think to your point, they are doubling down on that.

And I'm just going to go through some of the data that we've seen on how China is winning in this AI efficiency game.

So I just talked about Alibaba and their AI model, which is called Quen.

They're using 82% less GPU chips because they figured out this system that helps them do that.

Again, as I said, I don't know exactly how that system works, but that's what they've told us.

DeepSeek, they are also optimizing for efficiency.

They're using this thing called the FP8 format, which essentially just cuts down the number of the decimal count on all of these long numbers that go into these models, which allows them to reduce the amount of usage of memory.

and it allows them to run their training models a lot faster.

GLM is a model that is produced by this other Chinese AI company called ZAI.

They are also using a lot less energy, and all of these efficiencies are being reflected in the costs of these models.

So just to go through the costs here, the cost per million tokens, if you're using OpenAI's model, which is GPT-5, the cost per million tokens is $10.

Compare that to ZAI's model, which is called the GLM 4.6 model.

The cost is $1.75.

For Alibaba's model, Quen Plus, it is $1.20.

For DeepSeek, it is $1.10.

Put another way, all of these Chinese models...

are nearly 10 times cheaper than their equivalents in the US.

Now, I'm sure the US models are better and we'd want an AI engineer to come on and confirm that to us.

But what is clear is that they are putting exactly, as you say, the old Navy strategy.

80% of the value or the quality or whatever you said for not 50% of the price, but for 10% of the price.

So they are doing to AI the same thing that they did to electronics, the same thing that they did to apparel, the same thing they did to consumer goods.

It's the same thing again.

It is reduce the costs dramatically.

And at the same time, and

sort of upstream of that, reduce the amount of energy required to create these models dramatically.

And they seem to be way ahead of the ball on this.

And I do agree with your point that this could be, it could gut the American economy, which, as you say, has become incredibly reliant on AI.

Now, I want to go through in a second

just the history of efficiency in the business world.

But I will pause there and see if you have any reactions to those numbers.

So earlier in the year, we predicted that the rivers of capital into the US were going to reverse flow.

That was mostly sort of right.

And that is EU stocks have done really well.

But also, to be fair, American stocks are up 12%, 13%.

The S&P is up.

So clearly, there's still massive flows of capital into the U.S.

The most dangerous own goal over the medium and long term is there really has been a chill

placed on the inbound rivers of human capital, the inflow.

And that is some schools, some colleges are projecting there's gonna be 20 to 40% fewer foreign applications.

Every talented PhD student and every world-class researcher in the world had one thing in common.

They had either come through US universities, been trained here, or would seriously consider coming to work for our universities.

And now what's happened, China's figured this out.

China produces 120,000 PhDs every year.

That's three times what the US produces.

In 2024, China's AI research publications, the output, the amount of peer-reviewed credible research matched the combined output of the US, UK, and European Union, and now commands more than 40% of global citation attention.

And AI, which let's be honest, is very IP and research heavy, technical heavy,

They've got more coal in the furnace, and they're going to use it strategically to reach that type of scale that you're about to talk about.

Talk about the importance of scale in our winters.

what I would describe versus scale is the importance of efficiency.

And if you look back through basically every great company in history and what they were good at and where they innovated, they all basically have one thing in common, and that is they were all incredibly innovative when it came to cutting down costs.

And we can go back as far as the invention of the car.

I mean, you look at Ford as a great example.

Their great innovation was the conveyor belt, the assembly line.

And they basically, what they did was revolutionize how you make a car.

And what used to take 12 and a half hours of manual labor, they were able to reduce it to just 90 minutes.

And because of that, in less than a decade,

Ford cut the price of cars in half.

And then a decade later, they cut the price of cars in half again, to the point where by the 20s, more than one in two cars in the world was a Ford.

That's how they did it.

They just innovated on, let's make it more efficient to build this thing, not let's make the coolest and sexiest and fastest car in the world.

McDonald's, same thing.

They reinvented the kitchen assembly line.

They reduced the menu to just a few items.

They cut the cost of the hamburger in half.

They became the biggest restaurant chain in the world.

Ikea, another great example.

Their big innovation was just how you ship the product.

Let's put them in flat, small packages.

They shrunk the size of the box,

by 50%, which cut the logistics costs almost in half.

And they can now ship 10 times more furniture than their competitors for the same amount of gas.

Now it's the largest furniture retailer in the world.

Walmart, another good example too, cutting down logistics costs so they can sell products cheaper than their rivals.

SpaceX, a newer example, their big innovation, let's just reuse the rocket.

Let's make it cheaper and more energy efficient to do this.

And they cut the cost of launching stuff into space by more than 95%.

And as we've discussed, now they have 90% of the space launch market.

So I think when we look through history...

The most impactful companies, the companies that achieve the most amount of significance and dominance and revenue, it's the companies that figure out a way, how can we do more with less?

And I think that is the big question that America now needs to ask itself, especially when we look over what's happening in China, where China is laser focused on addressing this problem.

And I think if you look through history, what you would assume is that they're going to win on AI if they figure that out before we do.

So there's this great economist at Stern named Bruce Buchanan, and he absolutely blew my mind and gave me a model through which I run almost every strategy and every decision I make economically or in terms of positioning a company.

And that is all shareholder value, all stakeholder value is a function of the relationship between three lines.

at the very top, perceived value.

In the middle, the price you're charging, and the bottom line is the cost to produce that product.

And companies can only increase stakeholder value by doing one of two things.

They either push down the cost line, right, through scale.

That's what we're talking about now.

Hamburgers for less money, putting satellites into space, payload into space at a lower cost per kilogram.

assembling a computer for less than anyone else, Dell, right?

And then once you push that line down, you can pass on the savings as Walmart does and immediately lower the price concurrently.

And that creates a gap between the price you're charging and the perceived value, which is greater, which should result in more market share, which is how these companies have created hundreds of billions of shareholder value.

It's like, oh my God.

So the old ad that was most effective for Walmart was if you start shopping at Walmart, it's like getting a promotion.

and that is your quality of life, you're not gonna have to buy Budweiser, you can buy Heineken.

You can now buy steaks instead of hamburger.

It's like you got promoted.

Because of that scale, because we're able to lower our costs, we're merely gonna pass it on lower prices and the margin, if you will, of perceived value relative to price is greater, more market share.

The flip side, and what actually America is, I would argue, better at,

is taking the top line or perceived value and pushing it up through branding, through merchandising.

Loosely speaking, or very reductively, China wins by pushing the bottom line down.

Scale and passing along those price advantages to their customer.

And America, generally speaking, has value-add products

through advertising, strategy distribution, whatever it might be, scarcity or artisanship pushes the perceived value line up.

I mean, open AI appears to be the king of the perceived value in the AI space.

I mean, we keep on seeing these benchmarks, and in a lot of cases, different models are actually beating GPT when it comes to reasoning and speed, all these things.

And yet OpenAI is the premier brand in AI.

And I wonder if that is a function of, as you said, maybe they are really good at doing the stuff that Apple's really good at, which is the branding and pushing up the perceived value.

But I wonder if when you're talking about gigantic revolutionary technologies that expand beyond just

retail, something like this, something like AI, which is supposed to take our entire economy, the global economy, into an entirely different direction, whether it is imprudent of America to spend their time on perceived value or

versus figuring out, okay, how can we deliver this to the most amount of people for the least amount of money possible?

That to me seems like the right direction to go with when it comes to AI, especially if this is going to be as transformative as they say it is.

It's an interesting question, but traditionally, the American economy has been about value-add as opposed to scale.

Think of value-add is synonymous with brand value.

Name a global brand that's come out of the second largest economy, China.

It's tough.

Isn't that weird?

Think about it.

There's great French brands.

Louis Vuitton.

BYD, I think, is probably the sexiest brand coming out of China right now.

We hadn't heard about that until 24 months ago.

Some people say TikTok.

Some people say Alibaba.

But America probably has, whether it's Coca-Cola or Harvard, America has like hundreds of amazing.

We're really good at media.

We're really good at creating value at the end of the day.

Our consultants are the most expensive in the world.

right?

Our schools are the most expensive in the world.

The digital world, I think to your point, has become more about scale and open AI was first zero to kind of, you know, whatever, zero to, what is it?

Zero to a million users in like five hours or something, whereas other technologies have taken months, if not years.

Zero to 10 billion in revenues faster than anybody else.

So these digital products that are frictionless, I agree with you, it's more about scale, but generally speaking,

American products are usually considered globally the premium product at a higher price and greater margin.

Just to put a button on this, where do you stand on the US-China AI race?

I mean, China is winning on energy capacity and energy build-out.

They're also winning on efficiency.

We are probably winning on quality and, in some cases, scale.

and brand value.

We do have some progress on efficiency.

Google's developing these alternatives to the GPU called the TPU.

I think you're going to hear a lot about this TPU over the next few months, and they are more energy efficient.

They just had this deal with Anthropic.

Amazon is working on their Trillion chips, which are supposedly more efficient as well.

We're getting there, but I think it's safe to say...

China's winning on efficiency and they're winning on energy.

So given that, where do you stand on the race?

What's going to happen, do you think?

My emotional reaction is China's going to beat us because I think Donald Trump's policies are just incredibly sclerotic and short-term and just head up your assery like nothing before.

The biggest own goals in history have been committed, in my view.

Having said that...

I have a tendency and I think we have a tendency to overestimate the policies of a current administration relative to the underlying economy, which just churns on.

And the US economy, biggest gears in the world that grind on more.

It's like, have you seen those TikTok videos of those things that crush metal?

Like you put a car into the thing and it starts like eating it alive, like it's a tomato.

I mean, it barely even feels it.

It's like, oh my God, watch this.

What?

It can do, I mean, it can just eat anything.

That's like the US economy.

The US economy just appears to be so resilient and so powering on.

So my answer would be, okay, does China win with scale and low cost or does America win in AI with value add?

I think the answer is yes, and that is,

If you want to order a puffy winter jacket, it's going to be hard to get something for less than if you get it from Shein or another company that has manufacturing in Southeast Asia.

At the same time, people are still going to love North Face.

The existential threat here is a function of our valuations.

And we've been saying this forever that the cloud cover for the Trump administration right now and the cold comfort that people have in terms of confidence to spend more money, especially the wealthy,

who the top 10% now account for 50% of the consumer spending, which is basically economies resting on the top 10%, the top 10% consumer confidence comes from where the stock market is.

And the stock market is up solely, solely because of 10 companies, 77% of that growth.

So they don't need to win.

All they need to do is show that these companies aren't gonna dominate the world

Built into these valuations is an assumption that these 10 companies are going to own all of it.

All of it.

And so if they still have the niche products, still amazing businesses, still get premium margin because people want to say, I'm on ChatGPT or Claude.

I don't want to be on DeepSeek.

I don't want to be on, you know, it's not great self-expressive benefit.

Or if you're an American company, there might be security concerns.

Fine.

But all they need to do is say, we're eating into their share.

So I don't think there's a winner and a loser here.

What I think is...

Both will find their niches, if you will, or both will find a market.

The problem is, is that these companies get cut in half and the stock market goes way down.

And when it becomes clear, these companies aren't going to have like 97 points of share of the global AI market.