Micron and the "Supercycle": When a Boring Component Becomes AI's Main Character

Micron Technology, the DRAM and NAND memory maker that for years traded as a low-margin cyclical stock most of the market ignored, has surged 268% so far in 2026, and its market capitalization has reached $1.22 trillion, placing it among the 15 most valuable companies in the world. The driver isn't a consumer product story — it's a structural memory shortage caused by the massive buildout of AI data centers, and it's exactly the kind of story an investor shouldn't dismiss just because "memory chips" sound boring.

Context: from the 2025 shortage to an S&P 500 standout

High-bandwidth memory (HBM), used in Nvidia's AI GPUs and other accelerators, and conventional DRAM share the same fabs and the same cleanroom space. When Samsung, SK Hynix and Micron began diverting capacity toward HBM to serve hyperscalers (Microsoft, Google, Meta, Amazon), production of conventional DRAM for PCs, smartphones and traditional servers got squeezed — a classic supply shock, not an ordinary demand increase.

Back in January 2026, AI memory was already described as "sold out," with unprecedented price surges. The shortage has only intensified since: SK Hynix's CEO warned in July that demand will outstrip supply "well into the next decade," and new fab capacity from Micron and SK Hynix won't reach volume production before 2027.

The concrete, verified data

• Micron (MU) stock price, the full trajectory: 52-week low of $155.18 in January → a plunge to $322 on March 30, following earnings that actually beat expectations (see below) → a recovery to $751 by May 22 → an all-time closing high of $1,213.37 on June 25 → $1,085.02 on September 27. The stock has already pulled back 10.6% from its all-time high, even though it's still up 268% year-to-date — the path here has been far more volatile than a simple "up all year" headline suggests.
• The March scare, explained: on March 18, Micron reported fiscal Q2 2026 results well above expectations (revenue of $23.86 billion, adjusted EPS of $12.20 versus a consensus of $9.16). Even so, the stock fell as much as 30% in the following days: on March 24, Google Research published TurboQuant, a compression algorithm that cuts the KV cache models need by up to 6x with no accuracy loss — for a few days, the market read that as meaning less memory demand per AI model going forward.
• Market capitalization: roughly $1.22 trillion, the 13th most valuable company in the world; Micron crossed the $1 trillion mark on May 26, with shares jumping 19% that single day after UBS tripled its price target from $535 to $1,625.
• DRAM prices: up 80%-90% quarter-over-quarter from Q4 2025 into Q1 2026 across most segments; in April alone, prices rose 57% versus the quarterly average.
• Forward pricing guidance: industry suppliers have told customers to plan for additional price increases of 10% to 20% per month through the end of 2026.
• Contracts: Micron has already locked in price-and-volume agreements for its entire calendar-2026 HBM supply, with customers paying upfront (Supply and Capacity Agreements) — a sign the industry itself is pricing in the shortage as a given, not just market noise.

Here's how Micron's stock has moved through this entire cycle, scare included:

The chart shows something a "up 268%" headline doesn't: there was a drawdown of more than 30% in just 12 days after the TurboQuant scare (a 30.31% maximum drawdown from the high right before those earnings), and anyone who panic-sold at that point missed a nearly 4x recovery (from $322 to $1,213.37) over the following three months. Volatility, not just direction, is part of the data.

The data that actually backs the story: quarterly revenue

Beyond the stock price — which can overreact in either direction, as shown above — what has genuinely changed is the business itself. The acceleration in Micron's quarterly revenue during this cycle has no recent precedent at the company:

From $11.32 billion in the quarter right before the cycle began (fiscal Q4 2025) to $41.20 billion in fiscal Q3 2026 — revenue has nearly quadrupled in three quarters, with DRAM alone generating $31.3 billion in fiscal Q3 2026 (up 343% year-over-year) and NAND contributing another $9.9 billion (+361% year-over-year). This is what separates this episode from a purely speculative stock rally: the underlying business is growing almost as fast as the stock.

Asset-by-asset analysis: who wins and who loses from the memory shortage

Micron and its direct competitors (SK Hynix, Samsung): benefit directly from pricing power — they're selling a scarce product to buyers with no short-term alternative, with Micron's pricing already locked in years out via long-term contracts.

Hyperscalers and server buyers (Microsoft, Google, Meta, Amazon): face the other side of the same coin — memory (especially HBM) is one of the fastest-rising cost components of building an AI data center, pressuring the return on investment of AI projects right when the market is watching that return closely. March's TurboQuant episode also shows that any software advance that reduces the memory needed per model is, paradoxically, a direct threat to this thesis — physical supply isn't the only variable; demand can shift through software too.

PC and smartphone makers: are likely the least able to pass this cost increase through in the short term due to competitive pressure, which compresses their margins — a price-transmission mechanism that could eventually show up in durable-goods inflation if it persists.

The memory sector itself as an asset class: this is the risk that gets discussed the least. Memory makers have historically traded at low multiples (4-5x earnings at past cycle peaks) precisely because the market knows these cycles are cyclical, not structural — this rally, by contrast, is pricing Micron as if it had a structural, AI-linked growth story. If the market is wrong about that premise, the multiple correction could be sharp, as the historical parallel below shows.

Historical parallel: the lesson of the 2015-2018 memory supercycle

There was an earlier memory "supercycle," and it carries a pattern every investor in this sector should know — and unlike the chart above, this one starts by showing the cycle bottom, not just the climb. Between 2015 and 2018, a combination of bigger smartphone storage, the early cloud buildout, and a supply constraint (manufacturers diverting capacity to 3D NAND) first dragged revenue down, then sent it soaring:

Micron went from $16.19 billion in 2015 down to a low of $12.40 billion in 2016 — the "bottom" before the takeoff — then up to $20.30 billion in 2017 (+64%) and $30.40 billion in 2018 (+50%), with gross margin peaking at 58.9% and operating income of $15.0 billion in 2018. But Micron's stock had already topped out earlier, around $64 (at a P/E of just 4-5x), and from there it fell 56% while earnings were still climbing. That's the classic pattern in memory cycles: the stock price front-runs the cycle turn long before it shows up in results — the market starts pricing in the next glut while profits still look unbeatable.

What to watch from here

• Micron's fiscal fourth-quarter results, September 30, 2026: the next concrete date on the calendar — whether quarterly revenue keeps accelerating at the pace shown above, or starts to level off, will be the most direct signal of where this cycle is headed.
• New AI compression/efficiency breakthroughs, TurboQuant-style: March already proved the memory-shortage thesis can take a real hit from the software side, not just from factory supply — any similar announcement deserves immediate attention, not just physical-supply data.
• New fab capacity (2027): once Micron's and SK Hynix's new capacity starts reaching production, it will be the first real supply-relief signal — until then, the market keeps operating on the assumption of scarcity.
• Whether the 10-20% monthly price increases hold or moderate: any moderation would be the first crack in the structural-shortage narrative.
• Micron's distance from its all-time high: the current 10.6% pullback from June is, for now, just a correction within an uptrend — but the 2018 parallel is the reference point to keep in mind if that gap keeps widening while earnings are still at record highs.