Why Silicon Valley Can’t Stop Looking Over Its Shoulder at China
The United States dominates the world of artificial intelligence in almost every respect. The most powerful artificial intelligence systems are made in America. The United States has far more AI data centers than any other country. And the vast majority of computer chips used to build AI were developed by American companies.
However, perception is another matter.
In recent weeks, China appeared to be eroding that considerable lead. The competition got tougher last week when a Chinese start-up released a system nearly as powerful as leading American technology but costing much less. It was the second time in about a month that the Chinese company had succeeded.
In a speech on Friday, Chinese leader Xi Jinping hailed Beijing as a champion of the new AI global order. “The development of artificial intelligence should not be a solo performance by one country, but a symphony of international cooperation,” Mr Xi said.
As with other cutting-edge technologies, from consumer electronics to electric cars, China is demonstrating that it can make something almost as good, but at a much more affordable price. That’s in Silicon Valley, where industry veterans know that when it comes to technology, good enough often beats the best if it’s much cheaper. This has been true since computers replaced mainframes, and it could happen again with AI
But concerns about China’s growing AI prowess appear to be less about what the country’s technologists are doing right and more about what the U.S. he could do wrong American companies could be building their product the wrong way. They could be spending too much and charging too much for their products. The Trump administration could also meddle unnecessarily in the AI market.
“It’s still the US race to lose, but it’s getting pretty damn close,” said Rehaan Ahmad, co-founder of Silicon Valley startup alphaXiv, which has been using the latest Chinese technology for the past few weeks.
Last month, AI startup Anthropic shut down its two most powerful systems after the government unexpectedly demanded that the company bar access to foreign nationals, including its own employees. Anthropic’s close competitor, OpenAI, also said the administration interfered with a recent release of its technology.
Like some AI researchers, administration officials have worried that the technology could fuel cyberattacks or perhaps even help create biological weapons.
The administration lifted its restrictions, but Anthropic and OpenAI still tightly control who can and can’t use the technology. Executives from both companies have called on the government to regulate artificial intelligence in some way.
On Friday, Mr. Xi specifically mentioned the so-called open source method that Chinese companies use to develop their technology. It was a remarkable moment when a world leader appeared to take sides in a decades-long debate about the tech industry.
On one side are those who think new technology should be tightly controlled, and on the other are those who think it should be freely shared and distributed — the open source approach that Mr. Xi has championed. Without open source, it would be extremely difficult for Chinese companies to catch up.
“The most authoritarian government produces the most egalitarian models, and the most democratic government should be the breeding societies that are the most authoritarian,” said Rayan Krishnan, chief executive of Vals AI, which evaluates the performance of the latest artificial intelligence technologies.
Companies like Anthropic, OpenAI and Google hope to recoup their aggressive spending by selling increasingly useful, but often more expensive, technologies to businesses and consumers. China offers similar technology for free. This means that companies anywhere in the world – including the United States – can operate this technology at a much lower cost.
On Thursday, Chinese start-up Moonshot AI released a new artificial intelligence technology that is nearly as powerful as the leading US model, Anthropic’s Claude Fable 5. In line with other Chinese AI companies, Moonshot said it will soon open source the technology called Kimi 3. The start-up is also selling a version as an online service, just as US companies do.
A day later, the technology-focused Nasdaq was down 1.4 percent and the S&P 500 was down 1 percent, with Alphabet, the parent company of Google, and Meta, the maker of Facebook and Instagram, down between 2 and 4 percent. In the minds of some tech experts, this was a silly overreaction.
“Nothing has fundamentally changed,” said Perry Metzger, a software developer who has worked on AI for more than two decades. “The problem is that most investors have no idea what any of these systems actually do, how any of the companies make money, or what any of these reports mean.”
The Kimi model is at the cutting edge of Chinese open source technologies that are almost as powerful as high-end American systems. Last month, another Chinese startup, Z.ai, released a model that was quickly adopted by startups and independent developers across Silicon Valley.
After its release, six of the top 10 most popular AI systems on OpenRouter were closely watched AI model ranking — were Chinese technologies.
In some key areas, the latest systems from companies like Anthropic and OpenAI still outperform Kimi and other Chinese models, according to benchmarks conducted by Vals AI and other independent companies. But since high-end Chinese models are generally open source, they cost significantly less to use.
In early 2024, a third Chinese startup, DeepSeek, first raised alarm among investors and tech executives by releasing a surprisingly effective open source system. Stocks saw a similar fall. Companies like OpenAI soon accused DeepSeek and other Chinese companies of misappropriating data from their AI systems to speed up Chinese technology development.
Last month, Anthropic sent a letter to two US senators accusing Chinese tech giant Alibaba of a “reckless” and “illegal” attempt to copy its technology through 24,000 fraudulent accounts.
Using data from one system to train another — a process called distillation — is common in artificial intelligence development, including in the United States. But Anthropic and OpenAI’s terms of service prohibit anyone from secretly collecting data for distillation. Anthropic called on lawmakers and regulators to explore ways to curb the practice.
But some experts believe that distillation will become less important as companies build systems designed to act as “AI agents” — digital assistants that can use other software to perform tasks. Training agents requires a lot more than distillation, these experts say.
Other experts have long argued that Chinese systems will always lag behind the best American models because US export controls restrict the flow of specialized computer chips needed to train AI technologies. However, Moonshot, Z.ai and DeepSeek are spending millions to access chips in data centers outside of China.
Various Chinese companies have also started building their own specialized chips, and AI startups like Z.ai say they’re starting to use these chips at least as a way to augment the chips they’ve been able to acquire from the United States.
Even as China waives U.S. export controls and other restrictions, many U.S. executives, lawmakers and politicians continue to call for regulations that tightly control the use of U.S. technology. But some experts believe that these regulations could eventually drive more people to Chinese models, as they can use the technology however they want.
“It’s really hard for American companies to operate when they know that the Chinese models are getting stronger,” Krishnan said. “Something like Kimi 3 is purely open source. There’s no way to control who uses it.”