The US chip export ban has hurt China’s AI startups, not the giants yet

Before Washington banned Nvidia’s export of high-performance graphics processing units to China, the country’s tech giants kept them in anticipation of an escalating tech war between the two countries. .

Baidu, one of the tech companies building China’s counterpart to OpenAI, has acquired enough AI chips to continue training its ChatGPT equivalent Ernie Bot for “the next year or two,” the CEO of company Robin Li said in a income call this week.

“Also, inference requires less powerful chips, and we believe that our chip reserves, as well as other alternatives, are sufficient to support many AI-native apps for end users,” he said. “And in the long run, the difficulties in obtaining the most advanced chips will inevitably affect the pace of AI development in China. Therefore, we are actively looking for alternatives.”

Other deep-pocketed Chinese technology companies are also taking proactive steps in response to US export controls. Baidu, ByteDance, Tencent and Alibaba collectively ordered about 100,000 units of Nvidia’s A800 processors to be shipped this year, worth $4 billion, the Financial Times reported. reported in August. They also bought $1 billion worth of GPUs scheduled to ship in 2024.

Such a heavy investment could easily prevent many startups from entering the LLM race. There are exceptions when the young business is able to secure handsome investments quickly. 01.AI, which was founded in late March by prominent investor Kai-Fu Lee, acquired a large number of high-performance inference chips through loans and has already paid off its debt after raising its capital cost. $1 billion.

With the reserve of these GPUs, Baidu recently launched Ernie Bot 4, which Li claims is “not inferior to any part of GPT-4.”

Rating LLMs is tricky thanks to the sheer complexity of these AI models. Many Chinese AI companies have resorted to increasing the ranking by diligently complying with the criteria of the LLM charts, but the effectiveness of these models when used in real-life applications is still awaiting judgment.

Smaller AI players, who lack the cash flow to hoard chips, will have to settle for less powerful processors that are not subject to US export controls. Alternatively, they may wait for potential acquisition opportunities. Li expects that with a confluence of factors, including the scarcity of advanced chips, high demand for data and AI talent, and large upfront investments, the industry will soon move to a “consolidation that stage.”

Leave a comment