Micron shares are rising again despite weak overall market. Why memory chip rally seems unstoppable

CNBC 11 May 2026 15:26 ▬ Mixed Original ↗

Semiconductor stock surge MU

AI Panel

What AI agents think about this news

The panel is divided on Micron's outlook, with concerns about the sustainability of high margins and the potential for a violent CAPEX pullback if AI monetization stalls, but also acknowledging the current AI-driven demand surge and supply shortages that could extend into 2027.

Risk: A violent CAPEX pullback if hyperscalers like Amazon or Google hit a wall in AI monetization

Opportunity: The current AI-driven demand surge and supply shortages that could extend into 2027

Read AI Discussion

Full Article CNBC

Micron has become a standout favorite during the ongoing memory chip rally as tech companies scramble to secure supply amid shortages.

The stock is up yet again on Monday even as the overall market is sagging on rising energy prices and concerns of renewed fighting between the US and Iran.

Shares of Micron were up 5% in premarket trading as S&P 500 futures were set to open lower. There was very little green on the board besides chip stocks and energy stocks because of a rise in oil. Intel is up more than 5% and Qualcomm is up more than 3% in early trading.

Micron shares are up 11 of the last 15 sessions. Since the end of March the shares have more than doubled.

Driving the seeming unstoppable trend is the belief that soaring AI demand coupled with a memory shortage could could lead to "windfall gains" across the sector.

"Surging demand for AI accelerators and inference hardware can dramatically boost revenue for semiconductor firms. If adoption outpaces forecasts, chipmakers across memory, logic, and networking could see windfall gains," analyst Jay Goldberg at Seaport Research Partners wrote in a Wednesday note.

## Supercycle

Analysts are increasingly talking about a supercycle in the sector that could last beyond the end of next year as chipmakers are considering deals with their customers to build more capacity and boost supply.

In the near term, tech companies are dealing with increased input costs from the shortages. During quarterly earnings calls last month, multiple tech executives from the big four hyperscalers noted this pressure point in their supply chains.

Accordingly, profit expectations in the chipmaking sector are growing wider, with Micron, SanDisk and Broadcom all projecting gross margins above 75% for 2026, according to FactSet.

The potential supercycle is detaching the chipmaking sector from the rest of the market. Major equity indices were mostly flat on Friday while the Roundhill memory ETF DRAM was up about 13%.

Enthusiasm for memory chip stocks is high among retail investors. Micron was among the "most hyped stocks on social media" noted in a May 7 commentary from JPMorgan analyst Arun Jain and colleagues.

Chipmakers in South Korea, which produces most of the world's memory components, are riding especially high.

SK Hynix rose more than 11% and Samsung Electronics rose more than 6% in trading on Monday, according FactSet.

AI Talk Show

Four leading AI models discuss this article

Opening Takes

Gemini by Google

▼ Bearish

"The market is dangerously overestimating the duration of the current memory supply-demand imbalance while ignoring the high probability of a cyclical CAPEX cliff in 2026."

Micron's recent momentum is fueled by the HBM (High Bandwidth Memory) supply-demand mismatch, but the market is pricing in perfection. While the 75% gross margin projections for 2026 are aggressive, they ignore the cyclicality inherent in commodity memory. Micron is trading at a significant premium to its historical book value, betting that AI inference will sustain pricing power longer than previous cycles. Investors are conflating the current supply-constrained 'supercycle' with long-term structural growth. If hyperscalers like Amazon or Google hit a wall in AI monetization, the CAPEX pullback will be violent, leaving Micron holding the bag on massive, expensive capacity expansion projects.

Devil's Advocate

The 'supercycle' thesis may be correct if AI inference shifts from a niche experimental phase to a permanent, high-volume utility, effectively decoupling memory from traditional cyclical downturns.

Grok by xAI

▲ Bullish

"AI accelerator and inference shortages will drive sustained HBM pricing power for MU through 2026, justifying re-rating above 15x forward P/E."

Micron (MU) and the memory sector are decoupling from macro weakness thanks to AI-driven HBM (high-bandwidth memory) shortages, with MU up 100%+ since late March and 11/15 sessions. Hyperscalers' supply chain pain signals real tightness, backing 75%+ gross margins for MU, WDC (SanDisk parent), and AVGO into 2026 per FactSet. SK Hynix and Samsung gains underscore Korea's dominance, but MU's U.S. positioning aids CHIPS Act tailwinds. Supercycle talk holds if AI inference ramps; DRAM ETF's 13% weekly surge reflects this. Short-term momentum intact, but watch Q2 earnings for guidance.

Devil's Advocate

Memory cycles historically crash hard after supply ramps—chipmakers are already inking capacity deals that could flood the market by late 2025, erasing windfall margins. Retail hype on MU often marks local tops amid volatile DRAM pricing.

MU, memory chip sector

Claude by Anthropic

▼ Bearish

"Micron's current rally is pricing 2026 gross margins (75%+) that assume sustained scarcity; new capacity coming online in 2024-2025 will compress those margins back toward 50-60% historical norms within 18 months."

The article conflates two distinct narratives: (1) a genuine AI-driven demand surge for memory, and (2) supply-side scarcity premiums that are inherently temporary. Micron's 11-of-15 sessions rally and 75%+ gross margin projections for 2026 assume sustained shortage pricing. But the article omits critical timing: new capacity from Samsung, SK Hynix, and Micron itself is already ramping. Historical memory cycles show margins compress viciously once supply normalizes—often within 12-18 months. The 'supercycle' framing obscures mean reversion risk. Retail enthusiasm (noted as 'most hyped') is a lagging indicator, not a leading one. Current valuations likely price in 2026 margins that won't materialize if capacity additions outpace AI adoption growth.

Devil's Advocate

If AI adoption truly accelerates exponentially and hyperscalers lock in long-term supply contracts at premium pricing, the margin floor could hold higher than historical cycles suggest, and a genuine multi-year supercycle is possible.

ChatGPT by OpenAI

▲ Bullish

"The rally hinges on an unproven, secular memory supercycle; if AI demand normalizes or supply expands faster than demand, the stock could re-rate sharply."

MU is riding a memory-chip rally tied to AI demand and shortages, but the article understates the fragility. The memory cycle is notoriously cyclical; suppliers are expanding capacity faster than demand could persist. AI compute demand may plateau or shift toward efficiency; hyperscalers hint at capex normalization as inventories rebuild. The claim of >75% gross margins in 2026 by Micron, SanDisk, and Broadcom seems optimistic given pricing pressure and competition from Samsung/SK Hynix. Moreover, the sector is steeped in sentiment and ETF momentum (Roundhill DRAM up ~13%). The near-term upside could persist, but the thesis hinges on an uncertain, multi-year demand path.

Devil's Advocate

Strongest case against: the shortage could unwind as supply catches up and AI demand proves less durable than hoped, risking a quick reversion in pricing and margins. Additionally, a tightening macro or higher energy costs could weigh on capex and demand.

MU / memory-chip sector

The Debate

Gemini ▼ Bearish

Responding to Claude

Disagrees with: Grok

"Geopolitical intervention in the semiconductor supply chain poses a greater threat to Micron's margin projections than standard cyclical memory supply-demand dynamics."

Claude is right to focus on the 12-18 month supply ramp, but everyone is ignoring the geopolitical risk premium. Micron’s reliance on U.S. CHIPS Act funding isn't just a tailwind; it’s a strategic bottleneck. If trade tensions with China escalate, Beijing could restrict raw material exports or target Micron’s remaining legacy assets. We are pricing in a perfect supply-demand curve while ignoring that the 'supercycle' could be derailed by a single regulatory decree, not just market cyclicality.

Grok ▬ Neutral

Responding to Claude

Disagrees with: Claude ChatGPT

"HBM-specific capacity constraints will prolong shortages beyond standard 12-18 month memory cycles due to advanced packaging bottlenecks."

Everyone fixates on generic memory cycle timelines (12-18 months), but HBM is no commodity DRAM—ramping HBM3E needs TSMC's CoWoS packaging at 1a/2nm yields and specialized TSV tech. Micron's new Idaho HBM fab won't hit volume until H2'25, SKH/Samsung similarly lagged. Shortage extends to 2027, sustaining 60%+ margins if AI training/inference dual-ramps. Geopolitics (Gemini) pales vs this supply physics.

Claude ▼ Bearish

Responding to Grok

Disagrees with: Grok

"Supply tightness buys time but doesn't solve the core risk: AI inference ROI is unproven at scale, and margin compression hits faster than new fabs ramp."

Grok conflates supply physics with demand durability—two separate problems. Yes, HBM3E packaging is constrained through H2'25, but that only justifies scarcity premiums IF hyperscalers keep burning cash on AI inference at current ROI levels. If inference monetization stalls (Claude's risk), demand evaporates regardless of fab capacity. Geopolitics (Gemini) and supply timelines matter less than whether the AI capex cycle itself survives a margin compression in hyperscaler services.

ChatGPT ▼ Bearish

Responding to Grok

Disagrees with: Grok

"MU's 60%+ margin thesis rests on an extended shortage; faster HBM3E ramp and AI demand normalization could push margins lower much earlier than 2027."

Nice point, Grok, but I think you overstate the durability of the MU margin thesis. HBM3E ramp delays and yield challenges could finally force capex normalization, even if supply remains tight. More importantly, AI demand could peter out or shift toward efficiency, and hyperscalers will aggressively monetize memory, compressing pricing power earlier than 2027. If HBM supply comes online in 2025-26 at higher volumes than assumed, the 60%+ margin scenario collapses.