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

By Tom Whitfield

China's three biggest technology buyers — ByteDance, Alibaba and Tencent — have ordered $35 billion worth of Nvidia chips, Reuters reports. The figure covers orders placed by the trio for Nvidia processors, the silicon at the centre of the current build-out of artificial intelligence infrastructure.

The scale matters more than the timing. Three companies, one supplier, $35 billion. That concentration tells you who actually consumes advanced AI accelerators in China, and it tells you how dependent that demand remains on a single American vendor.

What does the deal consist of?

The reported orders span Nvidia chips destined for the buyers' data centre operations. ByteDance runs the TikTok and Douyin recommendation engines. Alibaba and Tencent operate the two largest Chinese public clouds, both of which rent GPU capacity to third-party developers.

Each of the three has its own AI model programme and its own need for inference capacity at scale. A joint commitment of this size suggests they expect compute demand to keep climbing, not plateau.

In plain terms: China's largest AI operators are betting billions that they will need more Nvidia silicon than they currently hold.

Why does one supplier dominate?

Nvidia remains the default choice for training and running large AI models, largely because its CUDA software ecosystem locks in developers. Chinese buyers have domestic alternatives — Huawei's Ascend line being the most prominent — but those alternatives trail in both supply and software maturity.

So the money flows to Nvidia even as Washington tightens export rules on advanced accelerators to China. Every round of restrictions has pushed Nvidia to design reduced-specification parts for the Chinese market, and Chinese buyers have kept ordering them because the alternatives cost more in engineering time than they save in procurement.

The mechanism is simple. Export controls set the ceiling on what Nvidia may ship; demand sets the floor on what Chinese firms will buy; and the gap between the two keeps widening as models grow.

What does this mean for the supply chain?

An order book of $35 billion from three customers gives Nvidia visibility well into future quarters and ties its revenue more closely to Chinese demand than any policy statement in Washington would suggest. It also puts the three buyers in a familiar bind: heavy reliance on a supplier whose government can restrict the product line by decree.

For China's domestic accelerator programme, the order is both a rebuke and a deadline. The rebuke: the biggest Chinese buyers still prefer American chips when they can get them. The deadline: every billion spent abroad is a billion of guaranteed demand that local vendors must capture if the country is to reduce that dependence.

Takeaway in one line — Chinese cloud money and American chip design remain joined at the wallet, whatever either government intends.

What comes next?

The orders now sit in Nvidia's pipeline, and their fulfilment depends on what US export rules allow to ship and when. How much of the $35 billion converts into delivered silicon, and how much Huawei and its peers capture as policy tightens, will shape the balance of the AI compute market on both sides of the Pacific.