• ■

    Alibaba and China Telecom launched a data center in China with 10,000 of Alibaba’s proprietary AI chips for training and inference workloads

  • ■

    The facility represents China’s largest deployment of domestically-produced AI accelerators as the country races to achieve semiconductor independence amid U.S. export restrictions

  • ■

    Alibaba’s custom chips compete directly with Nvidia GPUs that have been increasingly restricted from Chinese markets since 2022

  • ■

    The partnership showcases how Chinese tech giants are vertically integrating AI infrastructure to power the country’s rapidly growing AI industry

Alibaba just made China’s biggest bet on homegrown AI chips yet. The tech giant partnered with China Telecom to launch a massive data center packed with 10,000 of its proprietary AI chips, designed specifically for training and running large language models. The move signals China’s accelerating push to build AI infrastructure independent of Western chip suppliers, as U.S. export controls continue to restrict access to cutting-edge semiconductors from Nvidia and others.

Alibaba is throwing down the gauntlet in the global AI chip race. The company just unveiled a new data center built in partnership with state-owned telecom giant China Telecom, housing 10,000 of its own AI accelerator chips designed to handle everything from training massive neural networks to running inference at scale.

The facility marks the largest known deployment of Chinese-developed AI chips to date, arriving at a critical moment as Beijing doubles down on semiconductor self-sufficiency. With U.S. export controls cutting off access to Nvidia H100 and A100 GPUs since late 2022, Chinese companies have scrambled to develop alternatives that can power their AI ambitions without relying on American technology.

Alibaba’s chips, likely its Yitian or Hanguang series processors developed by its semiconductor division T-Head, are purpose-built for the computational demands of modern AI. Training large language models requires massive parallel processing power and memory bandwidth – capabilities that typically made Nvidia’s data center GPUs the gold standard. But China’s tech giants couldn’t sit idle while competitors like OpenAI, Google, and Microsoft raced ahead with unrestricted access to cutting-edge silicon.

The China Telecom partnership is strategic. As one of the country’s three major state-owned carriers, China Telecom operates nationwide infrastructure and has deep pockets for capital-intensive AI buildouts. The collaboration suggests this won’t be a one-off project – expect more facilities as China’s AI industry scales up. The telecom giant already operates extensive cloud services and can immediately monetize the compute capacity by offering AI training and inference to enterprise customers across China.