By Sophie Lindqvist

China wants its technology firms to stop buying Nvidia chips. A Financial Times analysis concludes that this effort faces an uphill battle. The report frames the problem plainly. Chinese developers still want the hardware they know.

That is the core finding. Everything else is context, and the context matters.

Why does Beijing want firms off Nvidia?

Washington has restricted Nvidia's most advanced AI accelerators from reaching Chinese customers. Beijing has responded by pressing domestic firms toward domestic chips. Two policies, pulling in the same direction for different reasons.

The Financial Times does not pretend the transition is going smoothly. Its headline says "uphill battle," and that phrase carries the analytical weight of the piece. Substitution is easier to mandate than to execute.

What holds Chinese firms back?

The report identifies demand for Nvidia products as the sticking point. Chinese technology firms have built software stacks, model training pipelines and engineering habits around Nvidia hardware. Switching costs are real, even when the political pressure to switch is intense.

A caveat belongs here. The Financial Times headline and framing do not, by themselves, quantify the gap. The report does not state how many units Chinese firms still acquire, through which channels, or at what premium. Readers should treat the direction of the finding as solid and the magnitude as unmeasured in the summary now circulating.

This is a familiar pattern in export-control coverage. The policy is public. The leak rates, gray-market volumes and stockpile depths are not.

What does this mean for Nvidia?

For Nvidia, the finding cuts both ways. Chinese demand for its chips persists, which confirms the strength of its platform. Yet that demand cannot convert into reported revenue as easily as it once did, because US rules restrict what Nvidia may sell into China.

The Financial Times framing implies Chinese buyers remain attached to Nvidia's ecosystem rather than any specific price point. That attachment is the asset Washington is trying to tax and Beijing is trying to break. Neither government has fully succeeded, on the evidence presented.

Again, a caution. The report as summarized gives no shipment figures, no revenue estimates and no named Chinese customers. Any precise market-share claims built on this piece would overreach it.

What is not said?

Three silences stand out.

• No timeline. The report does not estimate when Chinese AI chips could match Nvidia's offerings in the workloads that matter to buyers.
• No vendor ledger. It does not rank domestic alternatives or their claimed performance.
• No enforcement picture. It does not detail how effectively export controls or Beijing's substitution push are being policed.

Absences are not evidence. But in semiconductor journalism, they often mark where the next story will surface.

What comes next?

The competitive question now is whether Chinese chipmakers can convert political protection into genuine engineering preference, because mandates expire and software ecosystems do not. On the Financial Times account, that conversion has not happened yet, and every quarter of continued Nvidia demand inside China measures the distance still left to travel.