Nvidia’s CEO, Jensen Huang, is warning that China isn’t far behind the US on AI development at all, and in fact is “very, very close” to delivering competitive products and hardware for the growing AI industry. He went on to praise Chinese tech giant Huawei, calling it “one of the most formidable technology companies in the world,” and primed to rival Nvidia’s own AI chip production efforts. He warned that the country has made tremendous progress in recent years and could prove a real rival for US alternatives.
Until recently, Nvidia was primarily known for its gaming graphics hardware. Now, it is one of the largest companies in the world and at the forefront of AI training and hardware design, with its H200 graphics cards powering some of the most advanced AIs around the world. However, Chinese AI developments made waves in recent months for their capabilities and efficiency, and now Nvidia’s CEO warns they’re catching up on hardware, too.
“There’s no question that Huawei is one of the most formidable technology companies in the world, and they’re incredible in computing,” Huang told reporters, via Business Insider. “They’re incredible in networking technology and software capabilities, all of these capabilities to advance AI they have. They’ve made enormous progress in the last several years.”
Credit: Nvidia’s H200 GPUs are some of the most capable for training AI.
This comes amid the US government restricting sales of Nvidia’s China-specific H20 GPUs to the country, over fears it could help accelerate China’s AI developments. Huang’s warning seems to suggest that without Nvidia’s involvement, Chinese companies will fill the gaps instead. At the forefront of those developments is Huawei, a Chinese telecomms company initially rebuffed by Western countries in its attempts to contribute to the 5G rollout. It’s now producing AI hardware, with its CloudMatrix 384 system launching this year, offering alleged competition with Nvidia GPUs in raw performance.
However, while Huawei and other Chinese companies are developing new AI hardware at pace, some analysts point to their lack of established software ecosystems as a potential stumbling block. Where Nvidia’s CUDA has a robust pedigree of software integration, Huawei’s AI chips do not. That could hamper scalability, even if the AI hardware it produces is as capable as it claims.
It’s important to remember, too, that the job of CEOs such as Huang and OpenAI’s Sam Altman is to generate profit for their investors. That happens for Nvidia if it sells more GPUs, or if companies invest in ChatGPT, which may be more likely if Western companies feel China is poised to outcompete them. That’s not to say China isn’t developing AI hardware and that Nvidia and its partners have some stiff competition coming down the pipe. But Nvidia’s goals with announcing this are unlikely to be altruistic.