Semiconductors wobble as investors react to Broadcom’s AI “durability” warning, hitting South Korea’s memory firms and raising China-area competitive questions for Nvidia

A warning from Broadcom about the “durability” of artificial intelligence spending triggered a sharp risk-off move in semiconductor stocks on Tuesday, according to market reporting carried by Yahoo Finance. The selloff spread beyond AI logic chips to the memory complex, where South Korean suppliers were described as among the hardest hit.

The market reaction reflected a growing focus on whether recent surges in AI-related capex, including spending on data-center infrastructure, will remain as strong as many investors have priced in. In the same thread, the concerns were framed not as a pause in AI demand but as uncertainty over how reliably spending will sustain over time.

SK Hynix and Samsung Electronics were specifically cited in the reporting as seeing notable declines in Korea, underscoring the sensitivity of memory prices and order expectations to shifts in perceived data-center buildout pace. Memory is a key input to AI systems because training and inference workloads require high-bandwidth, high-capacity DRAM and NAND, so any announcement that AI infrastructure growth could cool tends to reverberate quickly.

Nvidia, a central supplier of AI compute, was also pulled into the debate, though in a different way. The report highlighted additional concern about Nvidia’s share of AI chips in China, stating that Huawei is closing in. The framing suggested the market was looking not only at overall spending levels, but also at the competitive landscape for AI accelerators and adjacent components in a constrained and tightly regulated market.

From a business standpoint, Broadcom’s comments matter to the semiconductor supply chain because AI buildouts depend on more than just GPUs. Broadcom supplies key networking and switching infrastructure used in data centers, and investors often treat “durability” language from such a supplier as an early read-through on whether hyperscalers and other customers will keep scaling AI clusters at the same rate.

For Nvidia, the China angle is a separate stressor. China is both a large end market and a market shaped by export controls and local vendor alternatives. Huawei’s progress, as described in the market report, would likely factor into how investors model future demand for Nvidia hardware or for compatible systems sold through regional channels.

Still, the available reporting provides limited detail on what Broadcom specifically meant by “durability,” and it does not lay out confirmed changes to customer orders, contract revisions, or guidance from any of the cited companies. Similarly, the report’s description of Nvidia’s China chip share falling and Huawei “closing in” is presented as market commentary rather than backed by company disclosures in the text available here.

What to watch next is whether the AI spending debate shifts from general sentiment to concrete indicators, such as revised company forecasts, memory suppliers’ commentary on order momentum, and any follow-up from Broadcom or hyperscale customers that clarifies whether the earlier pace of AI infrastructure investment is expected to hold.