I’ve spent the last couple of issues explaining the scandal surrounding Nvidia and what it could mean for America’s lead in artificial intelligence.

But I’ve also seen some surprising online chatter wondering if China’s AI models are starting to fall behind, and if the Nvidia scandal could make this problem even worse.

So I decided to look into it.

And I’ve found some compelling evidence that America’s AI advantage may be bigger than we realized.

China’s Growing Chip Problem

Back in July, I told you about Chinese AI startup Moonshot, which had released a powerful new open-weight model called Kimi K3.

It could compete with some of America’s leading AI models at a fraction of the cost.

But there was a catch.

Moonshot reportedly had to stop accepting new subscribers because it didn’t have enough computing power to keep up with demand.

That might have been our first sign.

Then last month, an independent testing firm called Artificial Analysis changed how it measures AI performance. It started using tougher tests with questions and answers that AI developers haven’t seen before.

Anthropic’s leading Claude model scored 58 points. But Moonshot’s newer Kimi model only scored 46, while the strongest models with publicly released weights managed just 32.

That doesn’t mean Chinese AI has suddenly gotten worse. But it suggests the gap between America’s leading AI systems and China’s best open-weight models may be wider than earlier tests indicated.

And closing that gap will require enormous computing power.

It takes thousands of powerful chips to train an advanced AI model. And once that model is ready, you need even more computing power to serve millions of users.

Turn Your Images On

American companies can buy Nvidia’s most advanced chips to accomplish this. But Chinese developers can’t, at least not without getting around U.S. export restrictions.

As I’ve written about before, China has found some clever ways to make their models more efficient. Its developers have even used American AI systems to help train their own.

But at some point, you need advanced hardware to take that next leap.

So just how dependent is China on American chips?

Last week, Reuters examined more than 150 Chinese government documents, construction contracts and investment plans. It identified more than $3.2 billion worth of Chinese AI computing projects that call for restricted Nvidia chips.

One proposed facility in Fujian province is worth $873 million and calls for Nvidia’s H100 chips. Another in the Shaanxi province, worth about $423 million, lists both H100 and H200 processors.

Those figures cover the cost of the entire facilities, not just the chips. And some haven’t been built yet.

Still, it’s surprising how much China is counting on American technology to build its AI infrastructure.

And with the Nvidia scandal threatening that access, Reuters reports that Chinese developers have been scrambling to rent computing power overseas.

That’s one way to work around U.S. export restrictions. Instead of bringing advanced Nvidia chips into China, developers can send their AI work to computers in other countries.

But that workaround is now under scrutiny. Last week, Treasury Secretary Scott Bessent confronted Commerce Secretary Howard Lutnick over his department’s handling of suspected illegal chip shipments to China.

Turn Your Images On

At the center of the dispute was Megaspeed, the Singapore-based company connected to the Nvidia scandal.

According to Axios, President Trump had approved tougher restrictions on suspected chip diversions three times this year. But officials accused Lutnick’s Commerce Department of holding them up.

Meanwhile, Congress has been looking for ways to close that loophole.

In January, the House passed legislation by a vote of 369 to 22 that would give the Commerce Department greater authority to regulate foreign access to restricted American computing technology through overseas data centers.

Turn Your Images On

That bill still needs Senate approval. But Washington isn’t waiting for Congress to act.

On Monday, the Trump administration issued new rules requiring overseas data centers serving Chinese companies to identify their end users. It also imposed additional licensing requirements on advanced chip sales to several countries, including Singapore and Malaysia.

The rules took effect immediately, targeting China’s access to American computing power through third countries.

That could make it even harder for China’s AI developers to get the hardware they need.

But advanced chips aren’t China’s only disadvantage. American AI companies also have far greater financial resources.

For example, Anthropic is expected to raise more than $90 billion through its upcoming stock offering and related investments. That’s vastly more than Chinese AI developer Z.ai has raised.

That kind of money gives American companies the ability to build bigger computing systems and develop more advanced models.

Of course, I wouldn’t count China out of the AI race.

Its companies are still releasing competitive AI models. And firms like Huawei are developing domestic processors that could reduce China’s dependence on Nvidia.

Chinese companies appear to be focused on making AI affordable for a much wider audience rather than trying to build the world’s most powerful models.

And that strategy could pay off handsomely.

But the latest benchmarks suggest America still has a substantial lead when it comes to the most demanding AI tasks.

And with our advantages in advanced chips, computing infrastructure and funding, our lead could become even harder for China to overcome.

Here’s My Take

America’s chip restrictions might be slowing China down, but they’ve also pushed its developers to get creative.

And that makes me wonder what American companies could accomplish by combining that same ingenuity with the world’s best chips.

We already have the tools to lead the race to superintelligence.

Now it’s time to put that advantage to work.

Regards,

Ian King's Signature
Ian King
Chief Strategist, Banyan Hill Publishing

Editor’s Note: We’d love to hear from you!

If you want to share your thoughts or suggestions about the Daily Disruptor, or if there are any specific topics you’d like us to cover, just send an email to dailydisruptor@banyanhill.com.

Don’t worry, we won’t reveal your full name in the event we publish a response. So feel free to comment away!