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

China's three largest internet companies have placed orders for Nvidia chips worth more than $35 billion, according to a Reuters exclusive. ByteDance, Alibaba and Tencent are the buyers, and the combined value of their commitments makes this one of the largest single-country chip procurement stories of the current cycle.

The figure deserves attention because of who is spending, not just how much. These three firms operate the biggest cloud and AI compute fleets in China, and their purchasing decisions effectively set demand for the entire domestic AI infrastructure market. When they commit tens of billions of dollars to a single supplier, that order signals where they expect the compute bottleneck to sit over the next several years.

Why does the $35 billion figure matter?

The size of the order tells us that Chinese AI builders still treat Nvidia silicon as the default engine for training and serving large models. Despite two years of United States export controls aimed at cutting off China's access to top-end AI accelerators, the country's largest spenders are putting money behind Nvidia products on a scale that rivals the capital plans of major cloud providers elsewhere.

The mechanism here is straightforward. Large language models require enormous clusters of GPUs. ByteDance builds recommendation and generative AI systems for Douyin and its overseas apps. Alibaba runs cloud training for its Qwen model family. Tencent serves AI features across WeChat and its cloud business. Each of those workloads scales with GPU count, and each company appears willing to pay Nvidia prices to secure supply rather than wait for domestic alternatives to mature.

In plain terms: the biggest AI buyers in China have voted with their wallets, and they voted for Nvidia.

What does this say about export controls?

The order sits inside a complicated trade environment. Washington has restricted sales of Nvidia's most advanced data-center chips to China, which forced the company to develop modified products that comply with the rules. Beijing has, in turn, discouraged Chinese firms from buying those compliant chips and has pushed domestic accelerators from Huawei and others.

The $35 billion commitment shows the limits of both pressures. United States rules shaped which Nvidia products Chinese firms can buy, but they did not eliminate the demand. Chinese government preferences for local hardware have not yet replaced Nvidia's software ecosystem and performance in the budgets of the companies that actually run the workloads.

The commercial picture, then, is a market pulled in two directions at once, with buyers resolving the tension in favor of supply security. Takeaway: regulation defines the menu, but the customers still choose the fastest dish on it.

How does this affect Nvidia and its competitors?

For Nvidia, the orders reinforce China's weight in its revenue base, even after earlier restrictions cut into its ability to sell flagship parts there. An order book above $35 billion from three customers provides visibility that most chipmakers cannot match, and it comes at a moment when Nvidia's AI accelerators are already supply-constrained globally.

For domestic Chinese accelerator vendors, the news is a sober benchmark. If the country's three largest AI spenders are directing this much money to Nvidia rather than to local alternatives, those alternatives still face a gap in performance, software maturity, or reliable volume supply — likely some combination of all three.

For the wider supply chain, the order implies continued pressure on advanced packaging capacity at TSMC and memory supply at SK Hynix and Samsung, since every Nvidia GPU shipment pulls those components through the pipeline with it. Demand of this scale does not stay contained within one vendor's order book.

What comes next?

The open question is fulfillment. Export rules can change, approvals can slow shipments, and Chinese policy on foreign chips can tighten again. Reuters' report establishes the orders themselves; how much of the $35 billion converts into delivered silicon will depend on regulators in Washington and Beijing as much as on Nvidia's factories. If the orders hold, China's AI build-out will keep running substantially on Nvidia hardware for the foreseeable future.