Wrong questions being asked about AI ecosystem

For much of the past three years, the global artificial intelligence story has been framed as a United States' one. OpenAI, Google and Anthropic have dominated headlines, while investors have poured money into the companies supplying the chips, memory and data centers needed to power the AI boom.

But a new question is beginning to emerge: What happens if Chinese AI models are better, and much cheaper?

The instinctive answer is that this must be bad news for the existing AI winners. Yet the reality may be more complicated.

Chinese AI models are increasingly finding users outside China. Their appeal is not that they are necessarily the smartest models in the world. Rather, they are often "good enough" for many everyday tasks while costing a fraction of the US alternatives. In some cases, usage costs are only 15-20 percent of comparable US models. At the same time, the performance gap between Chinese and US models continues to narrow.

This should not be surprising. Chinese developers have focused intensely on efficiency. Instead of pursuing the most capable model at any cost, many have prioritized delivering useful performance at the lowest possible price. They have embraced collaborative innovation, streamlined model designs and found ways to use computing resources more efficiently.

For consumers and businesses, this is unquestionably good news. Cheaper AI means more people can afford to use it. More companies can experiment with it. More applications become commercially viable.

What is perhaps underappreciated is that China's contribution to the global AI ecosystem extends beyond lower pricing. Chinese AI developers have helped broaden the industry's focus from a singular race for the most powerful model toward a more practical question: How can advanced AI be made accessible at scale? In many ways, this echoes China's role in other technology sectors, where innovation has often been driven by affordability, deployment speed and mass adoption rather than by pursuing the most premium product.

As Chinese models become increasingly capable, they are expanding the range of choices available to enterprises and developers around the world. Greater competition tends to accelerate innovation, encourage cost discipline and reduce barriers to adoption. From a global perspective, the emergence of multiple centers of AI innovation will be healthier than a market dominated by only a handful of providers. A broader and more inclusive ecosystem means technological progress can reach businesses, consumers and emerging economies more quickly than would otherwise have been possible.

The more interesting question is what it means for the rest of the technology ecosystem.

Many investors assume that if AI becomes cheaper and more efficient, demand for the underlying hardware must fall. After all, if each AI query requires fewer computing resources, shouldn't fewer chips be needed?

History suggests otherwise.

When technologies become more efficient, people often use them more, not less. Think about mobile data. As networks evolved from 2G to 3G, 4G and now 5G, the cost of transmitting data collapsed. Yet total data usage exploded because people discovered entirely new ways to use the technology — streaming music, video, gaming and social media.

AI may follow a similar path.

The critical equation is simple: Total memory demand depends on how much memory each AI task requires, multiplied by how many AI tasks people perform. The first number will almost certainly fall as models become more efficient. The uncertainty lies in the second number. If cheaper AI leads to dramatically more usage, total demand for underlying infrastructure could continue rising despite greater efficiency.

Important distinction

There is another important distinction that is often missed.

Many of the Chinese models gaining attention globally are open-weight models. Put simply, they can be downloaded and run by others rather than being controlled exclusively by the company that created them. If these models become successful in the US or Europe, they will still need computing infrastructure to serve users. The data centers running them are likely to be located close to customers, not in China. As a result, the success of Chinese AI models does not automatically imply weaker demand for data centers, memory chips or cloud infrastructure.

Finally, it is necessary to avoid conflating two separate debates: Chinese AI and Chinese semiconductors.

China's progress in AI software does not automatically mean Chinese-made chips will dominate Western data centers. Those markets remain heavily influenced by geopolitics, supply chains and export restrictions. The more relevant question may be whether Chinese chipmakers can gain ground in traditional electronics markets as regional suppliers increasingly focus on the lucrative demand created by AI.

In other words, the rise of Chinese AI should not be viewed simply as a challenge to today's winners. It may instead reshape where value is created and who captures it. The biggest opportunities — and implications — are likely to emerge not from the technology itself, but from the second-order effects that follow.

The author is the head of EM and Asia equity strategy at UBS.

The views do not necessarily reflect those of China Daily.

If you have a specific expertise, or would like to share your thought about our stories, then send us your writings at opinion@chinadaily.com.cn, and comment@chinadaily.com.cn.