Who Will Win the U.S.–China AI Race Over the Next Decade?

Have you noticed something strange about the AI race between China and the United States? China’s large AI models keep getting closer and closer to America’s, yet never quite surpass them.

Every time a new generation of ChatGPT comes out, we catch up. When Sora made its debut, we quickly followed. When OpenAI released o1, we soon had a competing product.

Always half a step behind. Always a few months behind. Always right on their heels.

Some people will surely say that China simply lacks the computing power to overtake the United States. But look at it another way: what if deliberately staying just behind is a strategy we are pursuing in plain sight?

Just how far has America gone in its obsession with AI?

The five tech giants’ disclosed debt alone has exceeded $700 billion. Add in commitments such as long-term data center leases and bulk chip orders, and their total financial obligations approach $3 trillion.

Even so, they keep borrowing money to pour into computing power.

OpenAI’s Stargate project alone plans to burn through $500 billion.

Meta has poured $50 billion into just one AI data center.

Nvidia, too, continues to invest tens of billions of dollars, taking major stakes in AI companies and tightening its grip on the entire computing supply chain.

Across America right now, enormous data centers are springing up all at once. A single large AI data center can consume as much electricity as a town of 100,000 people.

With no clear path to consistent profitability, investors still refuse to pull back. America has practically gone all in on AI, betting that it will win.

Does this playbook sound familiar? It immediately brings to mind the old Star Wars program.

In 1983, Ronald Reagan announced that the United States would build a multilayered missile interception system in space, commonly known as Star Wars.

The plan sounded like science fiction, but the Soviet Union took it seriously. To avoid being left behind, it felt compelled to follow suit, pouring enormous resources into a space arms race.

We all know how that ended: the Soviet Union was exhausted by this war of attrition and eventually collapsed.

America’s approach to AI today follows exactly the same playbook as its Star Wars program back then.

You build data centers? I’ll build more.
You deploy a server cluster with 100,000 AI chips? I’ll deploy 200,000.
You burn through $50 billion? I’ll burn through $100 billion.

With the Soviet lesson staring them in the face, can America’s elites really not see the risks? Why do they keep charging down this road with their heads down?

Of course they understand what is at stake. But they are doing a different calculation.

The AI industry certainly has enormous potential. But for American capital, openness and sharing are not where the big money lies.

The real money lies in monopoly.

First, use staggering amounts of capital to erect a barrier so high that everyone else loses hope. Once an unassailable monopoly has been established, close the gates and squeeze the rest of the world for profit.

Europe, Japan, South Korea—you want to develop large AI models? Fine. First, put up $300 billion to $500 billion to build your data centers. Then absorb tens of billions in annual losses. Then pray that your technology can keep pace.

Can’t do it? Then pay a premium for my products and remain permanently dependent on me.

This is the strategy America has always excelled at, pursued in plain sight: I have the most money, so I’ll use capital to clear the field first.

And frankly, it has worked.

Look at Europe. Apart from a handful of second-tier open-source models, does it have any impressive large AI models to show? No.

What about Japan? It is in an even worse position, barely able to muster the resources to follow along. And South Korea? Samsung is making an effort, but the gap separating it from China and the United States is plain to see.

What America never expected was that one country would fight its way to the table against the odds—and do so at less than one-tenth of the cost.

That country is China.

With extremely low computing costs, highly efficient engineering and deployment capabilities, and an enormous talent pool, China has stayed right on America’s heels.

You spend 10 billion training a model; I spend 1 billion training one that performs almost as well.

You use 10,000 chips for training; I can catch up with 5,000.

Your model is indeed powerful. But mine is only a little behind.

AI needs electricity. China generates more than twice as much electricity as the United States, at much lower prices, with a seemingly inexhaustible supply.

As for talent, well over half of the world’s AI professionals today are yellow-skinned people of Chinese descent. It is ethnic Chinese talent in Silicon Valley competing against ethnic Chinese talent in China—even the Indians have long since fallen behind.

This is what makes America’s position so uncomfortable.

It thought it could spend everyone else out of the game, leaving even its Western allies far behind. Yet China—the competitor it most wanted to eliminate—remains firmly seated at the table.

And now Wall Street is getting anxious.

Capital demands returns. You have already invested trillions of dollars, built all those data centers, and consumed enormous amounts of electricity. Where is your killer app? Where is your steady revenue? Where is your return on investment?

Whenever these questions arise, America’s AI giants repeat the same line:

“Just give us a little more time. Put in one more round of investment. Once we completely crush Chinese AI, we’ll earn back everything we’ve spent.”

Once. Twice. Three times.

Wall Street bought the story. And even if it did not, what choice did it have? Surely all that money could not simply be allowed to go down the drain. The only option was to keep raising the stakes. Perhaps this time, they would win.

That is the sunk-cost trap. The more you invest, the harder it becomes to stop.

America initially invested in AI to win. Then it became a matter of being unable to afford defeat. Now, it is simply unwilling to admit that its enormous investment might come to nothing.

Today, American AI is far more than just an industry. It is a pillar of the U.S. stock market, a source of confidence in the American economy, and one of the last major bulwarks of dollar hegemony.

In recent years, seven tech giants have accounted for 80% of the S&P 500’s gains.

Americans’ pension funds, mutual funds, and insurance assets are all heavily invested in these companies.

Beyond that, global investors’ confidence in America is tied to them as well. And every growth story these companies tell revolves around AI.

Who would dare let the AI bubble burst? Not Wall Street. Not the Federal Reserve. Certainly not the U.S. government.

During the 2008 subprime mortgage crisis, it was only the housing bubble that burst. America’s core foundations—its technological, military, and economic dominance—remained intact. After sacrificing a few unfortunate firms like Lehman Brothers, America pulled through again.

But if the AI bubble bursts this time, America’s very foundations are inside that bubble. Tell me: what will it have left to weather the collapse?

Only when you understand America’s predicament can you grasp China’s long-term strategy.

Do not fight for first place in the short term. Do not chase the spotlight. Simply rely on the lowest possible costs, solid engineering capabilities, and a deep talent pool to keep following close behind.

When America accelerates, we keep up. When America slows its expansion, we maintain our pace. When America widens its lead over other countries, we widen ours too.

As long as we stay at the table, we can remain in an unbeatable position and quietly wait for our opponent to make a mistake.

So do not envy American AI companies’ sky-high valuations or enormous fundraising rounds. Every dollar borrowed today will eventually have to be repaid.

Many people will surely say: why not just default? Do you really expect America to play by the rules?

But that misses the point. Finance rests on creditworthiness. Even if America can refuse to repay its debts, once that trust collapses, nobody will be willing to lend to it next time.

If America has to exhaust its financial dominance to escape its debts once, that is a trade we come out ahead on, however you calculate it.

During the Star Wars era, America wore down the Soviet Union through a war of attrition. In today’s AI race, history may repeat itself.

Only this time, the country being dragged into an endless contest of attrition is called the United States.

America cannot win, and we are in no hurry to win. The ultimate outcome of this competition was already written the moment we decided to enter the game.