Last week’s rout in chip stocks is producing some cautious optimism on Wall Street that now could be a good time to reenter the sector. The VanEck Semiconductor ETF (SMH) dropped 8.9% last week, its biggest weekly tumble since April 2025 when it fell 15%. The Philadelphia SE Semiconductor Index (SOX) fell nearly 10%. Despite concerns about new frontier models forcing some consolidation at the application level, analysts see fundamentals at the hardware and infrastructure levels as essentially intact, with plenty of runway in capital expenditures and little near-term relief for the memory chip bottleneck. “We believe there is a lot of gas left in the tank as AI capex … and [gigawatt] power installations ramp well beyond 2028-29E … all driving semiconductor demand with a wide supply gap,” Vijay Rakesh at Mizuho wrote to clients in a Saturday note. Joseph Moore at Morgan Stanley wrote Monday that the recent sell-off in U.S. memory stocks “creates a compelling entry point.” “We think that the best value in the market comes from the compute names, notably NVDA and AVGO , but memory is catching up quickly given this deceleration, and we think this should provide a good entry point for the stocks,” he said. .SOX 1M mountain Philadelphia SE Semiconductor Index performance over the past month Nvidia and Broadcom shares have pulled back 4% and nearly 10%, over the past month, respectively. “Semiconductors should soon start to find a bid,” Mislav Matejka at JP Morgan wrote on Monday. “Our view is that fundamentals will likely remain constructive, as meaningful supply additions are not due before 2028, so it would be too early at present to price in an inflection.” Analysts at Evercore were slightly more cautious over the weekend. While they described the 20% dip in the SOX over the past four weeks as a “mid-cycle correction” – implying the fundamentals are still good – they think it will sink more before it bounces. “[It] could take another 2-to-3 weeks and another 10%-15% to finish playing out,” Mark Lipacis at Evercore wrote on Sunday. “[The] median post mid-cycle correction SOX bounce is 36% over 20 weeks.” Last week, Chinese company Moonshot AI released Kimi K3, a frontier AI model with some publicly available – or “open-weight” – settings that sent a chill down the spine of the U.S. technology sector. OpenAI’s head of strategic futures Dean Ball talked about it in the context of “full AI communism” in a social media post on Friday. However, almost everyone on Wall Street is anticipating consolidation at the level of AI software, which should produce just a few dominant players or company configurations, and they don’t think it will necessarily come at the detriment of the hardware level. “There’ll be an interesting, almost thermonuclear battle to provide the [computing power] to support the demand that’s emerging for open weight models, given that they are delivering them at a significantly lower cost … and higher performance,” Peter Fenton, general partner at venture capital firm Benchmark, told CNBC over the weekend.
