TL;DR: Two developments in China unsettled AI markets last week: memory-chip maker CXMT’s Shanghai listing, which rose 466% to 3.3 trillion yuan (£365 billion), and reports that Chinese engineers have built their own deep-ultraviolet lithography tools. Analysts quoted by The Guardian argue the sell-off outran the underlying threat.

The distinction that got lost is what CXMT actually makes. It produces DRAM — the memory that stores data for other chips to work on — not the graphics processors that do the computing. Those are complementary products, not competing ones, which makes CXMT a boost to the AI supply chain rather than a challenge to Nvidia. It does compete with SK Hynix and Micron, but Forrester analyst Alvin Nguyen called the sell-off in those shares an overreaction, noting the memory shortage is expected to persist until 2030.

The lithography question is the serious one

Lithography is the harder problem. These machines etch the finest lines in the world onto silicon, and ASML has been the only company able to build them. If the reports hold, Beijing could in principle produce processors rivalling Nvidia’s — which would matter enormously to Wall Street.

The caveat is time. Fabrication plants take years to bring up, Nguyen said. Mark Boost, chief executive of UK cloud provider Civo, called building a handful of such machines a major symbolic win but not a commercial substitute for ASML, arguing the Dutch firm’s position outside mainland China stays structurally safe until Chinese tools match Western yield and reliability.

What the reaction revealed

Chris Beauchamp, chief market analyst at trading platform IG, expects Chinese chipmakers to do what Chinese manufacturers did in steel and cars — undercut and outcompete on price. Given US export controls, developing domestic capability was the only path available.

The more uncomfortable point is what the volatility exposed rather than what caused it. A further factor in Nvidia’s slide was a Wall Street Journal report that it was weighing a $250 billion (£186 billion) backstop for an OpenAI datacentre project, roughly six months after a $100 billion arrangement between them collapsed. Nvidia has become, in effect, the central bank of AI, propping up large parts of the market through arrangements few investors can see clearly.

Looking forward

For UK investors, the practical read is that the AI trade’s fragility is structural rather than a response to any single Chinese announcement. As Nguyen put it, at some point Nvidia will no longer be among the world’s most valuable companies — and even then it might be worth $2 trillion, which is still a considerable business.