China's AI Chip Race Against Nvidia: What You Need to Know

Summary

China is working hard to create computer chips that can compete with Nvidia, the American company that currently dominates the world of artificial intelligence.

Think of it like this: if Nvidia is the best car manufacturer in the world right now, China wants to build its own competitive cars instead of having to buy from them.

Let me explain what is happening and why it matters.

What is the Problem?

In 2022, the United States government decided that China could no longer buy the most advanced computer chips from Nvidia.

The reason was national security—the US wanted to make sure China's military and government did not get too strong in artificial intelligence.

This created a crisis for Chinese technology companies. They could not buy the best chips, so they could not train the biggest and best artificial intelligence systems.

Imagine if you wanted to build a mighty video game console, but the company that makes the best computer parts said you could no longer buy from them. You would have to learn to create your own parts instead.

China's Answer: Make Their Own Chips

Instead of giving up, China decided to build its own artificial intelligence chips.

The leading company doing this is Huawei, a telecommunications and technology company you may have heard about.

Huawei has been designing chips for many years, so they already had the knowledge to try making artificial intelligence chips. The company announced in September 2025 that it would release new chips every year through 2028.

Their roadmap looks like this:

In 2026, they will release the Ascend 950 chip.

This chip will work about as well as Nvidia's H100, which came out in 2022. Think of the H100 as a four-year-old iPhone—still pretty good, but not the newest model.

In 2027, they will release the Ascend 960 chip.

This one will be about twice as powerful as the 950, trying to match Nvidia's newest Blackwell chip.

Imagine going from an older iPhone to the newest iPhone—that is a big jump.

In 2028, they will release the Ascend 970 chip, which will be even more powerful, possibly approaching Nvidia's next generation.

How Good Are These Chips Compared to Nvidia?

Here is an important point: China's chips are not as good as Nvidia's right now, but they are getting better. Think of running a race.

NVIDIA is in first place, and China started running from much further back. They are catching up, but they are still behind.

The biggest problem is how the chips are made. NVIDIA's chips are made using the most advanced manufacturing technology available. It is like having a high-precision factory.

China's chips are made with older, less precise technology. This means China's chips are bigger, use more power, and cost more to make.

It is similar to the difference between building a house with perfect measurements and materials versus building one with less precise tools—both work, but one is less efficient.

Nvidia's chips use a technology called extreme ultraviolet lithography, which is like using a microscopic laser to carve circuitry onto a chip.

The United States government does not allow China to buy these special machines.

So China has to use older technology that is less precise. This is a major disadvantage that will take years to fix.

What About Speed?

When we talk about speed, we mean how fast the chip can process information.

Right now, Nvidia's newest Blackwell chip can do about 26,000 trillion operations per second. Huawei's best chip can do about 1,000 trillion operations per second.

That means Nvidia's is about 26 times faster.

However, Huawei is trying to overcome this by using clever system designs. Think of it like this: if Huawei cannot make individual runners run faster, they can create a relay team with many runners to make up for it.

Real-World Examples

Let us look at actual companies.

ByteDance, which owns TikTok, started using Chinese chips for some of its work in 2024.

They could not use them for training their most advanced systems yet, but they used them for other tasks, like recommendations for what videos to show users. Think of it like this: if you cannot afford a professional chef, you might hire a good home cook for some of your restaurant kitchens.

DeepSeek is another interesting example. This is a Chinese artificial intelligence company that created a smart AI system while using less advanced chips than what Nvidia offers.

They did this by being very clever about how they wrote their computer code. It proved that you do not always need the absolute best hardware if your software is smart enough.

The Software Challenge

Hardware is only half the story. Software is equally important.

Nvidia created something called CUDA, which is like a universal language that tells Nvidia chips what to do.

Thousands of companies have learned this language. They have written millions of lines of code using CUDA. Think of CUDA as English—everyone has learned it, so it is easier to communicate.

China's chips use different languages, like something called CANN or MindSpore. These are like learning Chinese or Spanish—they work fine, but you have to relearn everything.

A company that wants to switch from Nvidia to Chinese chips has to hire engineers to translate all their code. This usually takes two to four months. That is expensive and risky.

Why Does This Matter for the Future?

If China succeeds in making good computer chips, several things could happen.

First, Chinese companies will not have to worry about the United States government cutting off their access to American chips.

Second, computer chips could become cheaper because there would be more companies making them instead of just Nvidia.

Third, other countries that are friendly with China could buy these chips instead of buying from Nvidia, spreading the technology around.

However, there are still big obstacles. China's factories cannot make chips as quickly or as precisely as Taiwan's TSMC factory, which makes Nvidia's chips.

Even if Chinese chips become just as good as Nvidia's chips, the Chinese factories cannot make nearly as many of them.

It is like if you have a really good baker but she can only make five loaves of bread per day, while the American baker can make fifty—the American baker still produces more.

The Bottom Line

China is making real progress in creating artificial intelligence chips that could compete with Nvidia.

They will probably have good enough chips for their own needs by 2027 or 2028. However, they will not surpass Nvidia's technology anytime soon. Instead, the world will probably move from having one company (Nvidia) that makes almost all the chips, to having several companies making chips.

This is actually healthier for everyone, similar to how having multiple car companies makes cars better, cheaper, and more diverse than if one company made them all.

The race continues, and the next few years will be critical. By 2027, we will know much more about whether China can really compete or whether this is just wishful thinking.