ByteDance, Alibaba and Tencent order $35 billion in Nvidia chips - Latest news from Azerbaijan

AI & Compute

ByteDance, Alibaba and Tencent Order $35 Billion in Nvidia Chips

ByteDance, Alibaba and Tencent have ordered $35 billion in Nvidia chips, one of the largest AI silicon commitments ever reported — pending U.S. export rules.

By
Rebecca Stone
Filed
Channel
AI & Compute
Read
3 min read

ByteDance, Alibaba and Tencent — three of China's largest technology companies — have ordered $35 billion worth of Nvidia chips, according to a report carried by News from Azerbaijan. The figure dwarfs typical single-quarter AI accelerator purchases and signals how hard Chinese buyers are pushing to lock in supply.

The order lands at a moment when Nvidia's China business sits on a geopolitical fault line. Washington has tightened export controls on advanced AI silicon for years, and every large Chinese order now carries the same question: can the chips actually ship?

Who is buying, and how much?

The reported $35 billion breaks across three buyers:

  • ByteDance, the parent of TikTok and Doubao, China's most-used AI chatbot
  • Alibaba, whose cloud unit has pledged major spending on AI infrastructure
  • Tencent, which builds AI services across its cloud, gaming and WeChat ecosystems

All three are racing to expand inference and training capacity for large language models serving hundreds of millions of users. They are also Nvidia's most important remaining customers in China.

Why the size matters

A $35 billion commitment is not a routine procurement. For context, such a sum would rank among the largest single AI chip orders ever reported, and it would represent a substantial share of Nvidia's total data-center revenue in any given quarter.

It also tells you what Chinese hyperscalers believe about demand. Companies don't commit tens of billions to silicon unless they expect the models running on it to earn that money back — through cloud rentals, advertising, enterprise AI services or consumer products.

What stands between order and delivery

Export rules are the obvious risk. Nvidia cannot legally ship its most advanced data-center GPUs to Chinese customers without U.S. government approval, and the rules have tightened repeatedly. Each new restriction has forced Nvidia to design derivative parts for the China market, and each derivative has faced fresh scrutiny.

So a reported order is not the same as reported shipments. Whether this $35 billion translates into delivered silicon depends on:

  • Which specific chips the order covers
  • Whether those parts fall under current U.S. export licensing requirements
  • Whether Washington's posture toward China AI exports shifts again before delivery

None of those details appear in the initial report.

The commercial picture

n For Nvidia, the order underscores a uncomfortable reality: China's appetite for AI compute keeps growing even as the regulatory window for serving it keeps narrowing. CEO Jensen Huang has repeatedly argued that losing the China market would hand opportunities to domestic rivals — and China's own accelerator makers, led by Huawei, are pushing hard to fill any gap Nvidia is forced to leave.

For Beijing, the order cuts both ways. Chinese giants buying American chips at this scale shows they still can't get equivalent capacity at home. But it also strengthens the case, already official policy, to accelerate domestic AI silicon.

For the buyers themselves, the calculus is simple. Compute is the bottleneck. Whoever secures the most accelerators can train bigger models, serve more users and price more aggressively. A $35 billion bet is the price of staying in that race.

If the shipments clear regulatory hurdles, expect China's three AI giants to convert that silicon into expanded model training and cloud inference capacity over the coming quarters — and expect Washington's next export-policy move to determine how much of the $35 billion Nvidia actually books.

Source: Google News: chip export controls

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

670 articles

Related articles

« Previous article