Semiconductor investors faced a volatile start to the week as calls for a slower pace of artificial intelligence development triggered a sharp sell-off in US chip stocks, while Korean memory makers staged a partial rebound in Seoul.
On Monday, the PHLX Semiconductor Index slid about 5%, and the iShares Semiconductor ETF tumbled nearly 6%. Nvidia, Micron and AMD fell between 3% and 5% as traders reassessed whether a slower rollout of advanced AI models would weaken demand for chips, memory and data-centre infrastructure. The concerns followed warnings from industry figures including Anthropic CEO Dario Amodei about the need for more time to manage AI risks.
By Tuesday, however, Samsung Electronics and SK Hynix defied the US rout. Samsung gained 0.7% and SK Hynix advanced 1.82% by late morning in Seoul, while the KOSPI was little changed. The divergence partly reflected timing: Samsung had already dropped 4.05% on Monday and SK Hynix fell 6.35%, helping drag the KOSPI down 3.26%. Seoul therefore priced in the AI scare before Wall Street’s semiconductor sell-off.
Analysts said the bearish narrative is not straightforward. TrendForce reported that DRAM industry revenue jumped 59.5% quarter-on-quarter in the second quarter to nearly $154.73 billion, driven by AI server demand for high-bandwidth memory and high-capacity server memory. The research firm expects server DRAM shortages and elevated pricing to extend into 2027, while some Chinese AI-chip suppliers have already raised prices by roughly 20% to 50% because of higher HBM costs. Ben Barringer of Quilter Cheviot said inference capacity remains constrained and "demand still far outstrips supply," even if training and model rollouts slow.
Still, the debate is expanding beyond immediate supply and demand. AI-related capital expenditure is expected to approach $800 billion in 2026, and AllianceBernstein expects leading hyperscalers to increase nominal capital expenditure to more than $1 trillion next year. Christopher Wood of Jefferies warned the biggest risk is that markets conclude companies will be unable to "monetise this capex," which would eventually hit GPUs, servers and memory alike. Others argued that slower AI development does not automatically mean lower infrastructure spending.
For crypto markets, the news does not directly involve digital assets, but broader risk sentiment from AI-driven equity volatility remains worth monitoring.
