Nvidia's China Share Slips to 34% as Funding Doubts and Export Curbs Bite

Nvidia's investment case now rests on three separate pillars — what it can sell, who can afford to buy it, and where it is allowed to ship it. Recent disclosures have moved all three at once, and not in the same direction.

The most concrete shift comes from China. Data from market researcher IDC, reported by Reuters and published on 13 October, puts Nvidia's share of the Chinese market for AI data-centre chips at 34% in the first half of 2026. The company shipped 678,000 AI accelerators over that period, a 44% decline year on year. US export restrictions and share gains by Chinese rivals were cited as the drivers. Access and competition, in other words, not the willingness of buyers to pay.

That distinction matters because the wider debate around AI spending is a different animal. On Friday, Reuters reported fresh concerns about debt-financed AI infrastructure investment weighing on technology stocks, with rising oil prices adding pressure on Wall Street. Chip shares had already come under strain on Thursday amid reports of higher inflation and the prospect of climbing financing costs, with worries about heavily indebted AI infrastructure operator CoreWeave intensifying the squeeze. Nvidia's stock fell 1.7% on Friday — a move that fits the cautious mood but says nothing on its own about demand for any specific product.

Research Commitments Cut Both Ways

Alongside the market-access story, Nvidia has been building out the scientific groundwork for its platform. On Wednesday the company pledged computing power worth $250 million to British research projects over five years, directed at the safety of agent-based AI. According to company statements, $50 million in compute access will be made available annually. Intended partners include the Oxford-led BOLD project, the UCL-led SOFAIR project and Imperial College London. The commitment is a donation of compute, not an announced cash payment, and no immediate revenue can be derived from it.

It builds on a broader research push: on 8 October Nvidia announced support commitments worth $1 billion for science and research in the United States, also spread over five years, covering fields including quantum computing, healthcare and energy security. The two programmes differ in substance — Britain's centres on the safety and protection of agent-based AI, while the American effort spans multiple scientific applications.

The platform itself is spreading further into industry. Nvidia said on Wednesday that semiconductor companies are using its CUDA-X libraries for chip development and manufacturing, and introduced cuPotential, a library for AI-driven simulation of atoms and molecules. Early tests with cuPotential achieved a two- to three-fold speed-up, according to the company, while Lam Research used NanoVDB to run a chip-fab simulation 30 times faster. Those figures describe specific tests and applications rather than a general performance guarantee.

The Consumer and Automotive Layer

On the product side, RTX Spark notebooks from several manufacturers went on sale Friday. The same day, Nvidia and Mercedes-Benz expanded their strategic partnership for driver assistance and automated driving, following Volvo Cars' confirmation on Thursday that it would broaden its own collaboration with Nvidia. These steps demonstrate additional uses for the technology, though they offer no direct answer to Chinese export barriers or the higher cost of financing AI infrastructure build-outs.

Weighing the stock therefore requires keeping three things apart: the availability of new products, the funding behind their deployment, and access to key sales markets. The latest news paints no uniform picture. Nvidia is widening its partnerships and its research base while financing conditions and regional competition have the potential to cap the economic outlook.