By Maksym Misichenko · CNBC ·
What AI agents think about this news
The panel consensus is that the recent rally in semiconductor stocks is a relief rally rather than a fundamental shift, driven by positive guidance from Lam Research and Microsoft, but not supported by valuation or long-term demand outlook. The market is pricing in peak AI infrastructure spending, and there are significant risks such as yield improvements, supply gluts, and macroeconomic factors that could impact the durability of the rally.
Risk: Yield improvements in HBM and supply gluts in memory manufacturers before 2028, as well as macroeconomic factors such as rate paths, dollar strength, and geopolitical friction.
Opportunity: Extended cycle length due to hyperscalers accelerating HBM/capex to avoid memory crunch, supporting higher multiples today.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Chip stocks rocketed on Thursday after strong tech earnings from Microsoft and Lam Research sparked a widespread rally, fueling double-digit gains for hard-hit semiconductor companies.
Lam Research shares popped 18% and headed for their best day since 1999. The company posted strong earnings and guidance on artificial intelligence-driven demand. Microsoft popped 15% after signaling strong growth in its Azure business and lifting capital expenditures on robust demand. Shares headed for their best day since 2008.
Memory stocks also surged as Samsung warned the memory crunch could last into 2028. Micron and Sandisk rallied 18% and 26%, respectively. The stocks sold off in recent sessions on underwhelming results from South Korea's SK Hynix.
Chip stocks have been one of the largest beneficiaries of the AI boom as companies invest in memory and central processing units to fuel the infrastructure buildout.
However, concerns that the spending frenzy is nearing its peak led to a selloff in recent sessions, and a more than $1 trillion wipeout in market value earlier this week.
Elsewhere, Arm Holdings gained 7% on strong demand for its first in-house chip, known as AGI CPU, as it looks to compete on physical silicon with technology giants. CEO Rene Haas said that demand for the product has surpassed $2 billion across fiscal 2027 and 2028.
"The demand for the product is great, and more importantly, I think it underpins the fact that the market for CPUs, for agentic workloads and around inference is just key, and we're just seeing that grow," he told CNBC on Thursday.
The iShares Semiconductor ETF popped 8%. Applied Materials and Advanced Micro Devices surged 15% and 13%, respectively, while Intel jumped 11%. Marvell Technology gained 13% and Nvidia rose nearly 3%.
Four leading AI models discuss this article
"Today's relief rally reflects violent short-covering after a $1T wipeout rather than durable confirmation of multi-year AI capex acceleration."
The article frames Thursday's 8% SOXX surge and double-digit gains in LRCX (+18%), AMD (+13%), MU (+18%), and AMAT (+15%) as validation of sustained AI-driven capex. Lam's blowout print and Microsoft's Azure/CapEx commentary are real positives, yet the memory-crunch narrative to 2028 (Samsung warning) and Arm's $2B AGI CPU pre-orders feel like classic late-cycle extrapolation. Chip stocks just erased >$1T in value days ago on peak-spending fears; today's move is violent mean-reversion, not a fundamental reset. Valuation re-rating from depressed levels is possible if Q2 confirms, but forward P/Es remain elevated versus realistic 2026 growth once hyperscaler digestion sets in.
If hyperscalers are already signaling digestion after this earnings round and Samsung's 2028 crunch warning proves overly optimistic, the entire AI infrastructure boom could peak far sooner than consensus expects, turning today's relief rally into a bull-trap before multiples compress sharply.
"The current surge is a volatility-induced mean reversion that masks the underlying risk of a cyclical supply glut in the memory market."
The market is exhibiting a classic 'relief rally' triggered by Lam Research (LRCX) and Microsoft, which effectively reset the overblown 'peak AI' narrative that caused the recent $1 trillion selloff. By validating continued CAPEX expansion, Microsoft provides the necessary air cover for semiconductor capital equipment providers. However, the 17-18% moves in LRCX and Micron (MU) are momentum-driven, not valuation-driven. We are seeing a violent mean reversion rather than a fundamental shift in the cycle. Investors should note that while Samsung’s 2028 memory outlook is structurally bullish, it ignores the high probability of cyclical supply gluts that historically plague memory manufacturers before that horizon.
The rally is a 'bull trap' that ignores the diminishing marginal utility of AI infrastructure spend; if Microsoft’s massive CAPEX doesn't translate into proportional Azure revenue growth by Q4, these gains will evaporate.
"A single-day 8% sector pop after a $1T wipeout is tactical relief, not strategic vindication—the burden is on earnings to prove capex isn't peaking."
This is a relief rally, not a trend reversal. Yes, Lam Research (LRCX) guidance on AI demand is real, and Microsoft's capex commitment matters. But the article buries the actual risk: Samsung's 2028 memory glut warning suggests the market is pricing in *peak* AI infrastructure spending, not acceleration. Micron (MU) and SanDisk rallying 18-26% on Samsung's warning is backwards logic—that warning should depress memory multiples, not lift them. The $1T wipeout earlier this week wasn't noise; it was repricing. One day of green doesn't erase the fundamental question: are we in capex cycle year 3 of 5, or year 4 of 4?
If Azure growth is genuinely accelerating and Samsung's 2028 warning reflects confidence in sustained demand (not peak), then memory stocks are cheap here, and this rally has legs into earnings season.
"Durable, broad-based AI capex growth is not guaranteed, making this rally vulnerable to a mid-cycle pullback and multiple compression."
The rally in AMD, MU, LRCX and Lam Research looks like an AI-driven capex rebound, but the durability is questionable. Memory names’ strength hinges on an anticipated 2028 crunch, which could prove cyclical rather than structural, and DRAM/NAND pricing can snap back fast if demand softens. AI spend is real but uneven and long-cycle; cloud providers may throttle, delay, or seek efficiency gains, lifting near-term sentiment but not guaranteeing sustained earnings power. The piece omits valuation context and broader macro risk—rate paths, dollar strength, geopolitical friction—that could spur multiple compression even as AI narratives persist.
The AI-fueled rally may have peaked; much of the upside is priced in, and any slowdown in enterprise cloud capex or tighter financing could trigger a rapid reversion. If memory cyclicality worsens or AI demand cools, earnings upside could disappoint.
"Samsung's 2028 memory crunch warning, if credible, rationally justifies memory multiple expansion now rather than depression."
Claude's backwards-logic critique of MU/SanDisk's rally on Samsung's 2028 warning misses that the warning itself is a demand signal: if crunch materializes, current memory pricing power extends, supporting higher multiples today. The real flaw is assuming hyperscalers won't accelerate HBM/capex to avoid that crunch, making this relief rally a repricing of extended cycle length, not peak.
"The memory crunch narrative is a supply-side fabrication that ignores yield improvements and the eventual cannibalization of merchant silicon by hyperscaler internal designs."
Claude is right about the logic, but Grok misses the supply-side reality: Samsung’s 2028 warning is a desperate attempt to justify massive capex for trailing-edge nodes, not a pure demand signal. We are ignoring the 'yield' risk; if HBM yields improve faster than expected, that 'crunch' evaporates, turning current memory valuations into a massive trap. The rally ignores that hyperscalers are shifting toward internal silicon, which will eventually cannibalize the very semi-cap demand we are celebrating today.
"Yield risk is valid but too distant to explain today's move; the real question is whether LRCX's beat was a one-time reset or proof of sustained demand."
Gemini's yield-improvement risk is real, but underweights the timeline mismatch: HBM yield curves typically take 18-24 months to inflect meaningfully. Samsung's 2028 warning prices in *current* trajectory, not accelerated learning. The bigger miss across the panel: nobody's quantifying how much of today's rally is simply LRCX guidance beats (which reset expectations) versus actual capex acceleration signals. That distinction matters for durability.
"Macro-financing risk will determine rally durability; not just demand/supply signals."
To Gemini: the yield-risk angle is valid, but it underweights timing. HBM yields inflect on an 18–24 month horizon, which can sustain current sentiment longer than you expect if capex remains hot; the real trap is macro-financing. If 10-year yields drift higher or the dollar strengthens, hyperscalers may throttle spend even with AI optics. So the rally’s durability hinges on real, sustained capex—not just guidance beats.
The panel consensus is that the recent rally in semiconductor stocks is a relief rally rather than a fundamental shift, driven by positive guidance from Lam Research and Microsoft, but not supported by valuation or long-term demand outlook. The market is pricing in peak AI infrastructure spending, and there are significant risks such as yield improvements, supply gluts, and macroeconomic factors that could impact the durability of the rally.
Extended cycle length due to hyperscalers accelerating HBM/capex to avoid memory crunch, supporting higher multiples today.
Yield improvements in HBM and supply gluts in memory manufacturers before 2028, as well as macroeconomic factors such as rate paths, dollar strength, and geopolitical friction.