It is no secret that the arrival of ChatGPT has revived the interest of many of the big technology companies in betting on the artificial intelligence (AI). This race has also become a new scenario of rivalry between the United States and China, countries that seek to position themselves as leaders in this promising sector.
the asian giant, according to SCMP, is at a significant disadvantage. Despite its enormous economic and industrial muscle, China depends on certain US technologies to develop its projects in the field of AI on time and, consequently, present its national alternative to ChatGPT.
The trade restrictions promoted by Washington have been increasing, becoming the cornerstone of the technological war between the first and the second economic power in the world. Beijing receives blows on different fronts, but the inability to access certain NVIDIA products is one of the hardest.
A key player in the world of AI
When we think of NVIDIA we imagine some of its graphics cards such as the GeForce RTX 4080 and RTX 4090, but the American giant does not only design products for the mass consumption market. NVIDIA is one of the largest (and elemental) players in the AI computing industry.
Training the artificial intelligence models needed to build chatbots like ChatGPT requires vast computational power and scalability. The processing units NVIDIA H100 and A100 are considered one of the most suitable solutions for this type of work and, furthermore, they are tremendously in demand.
Expert forecasts indicate that OpenAI, the AI company run by Sam Altman, has used some 20,000 of these AI chips since 2020 for training tasks. Now, with its future projects on the horizon, GPT-4 and new models, it is expected that it will need more than 30,000 chips which, by the way, are banned from entering Chinese borders.
Under the argument of “keeping advanced technologies out of the wrong hands,” the US Department of Commerce established late last year new export criteria that limited the most advanced technologies from NVIDA and AMD. This move did not go unnoticed in China.
From the Asian country they did not hesitate to respond to that measure. They said it would “hinder international scientific and technological exchanges and economic cooperation, and have an impact on the stability of global industrial and supply chains and the recovery of the world economy.”
While China can still access some NVIDIA products, and also has a own industry, both options are far from what is expected to train AI models. According to one of the sources consulted by the Chinese newspaper, its supercomputing devices are still trying to catch up with the performance of the NVIDIA V100 released about six years ago.
China’s fledgling semiconductor industry, which is somewhat dependent on US tooling, licenses and software, is also suffering from trade restrictions. For example, the Biden administration has tried to persuade ASML not to sell it the extreme ultraviolet (EVU) photolithography equipment needed to make advanced chips.
Images: 偉宗勞 | Owen Winkel
