Take-or-Pay and the Fab Gap: Why Micron’s Demand Surge Can’t Stop the Stock Slide

The memory chip industry has run headlong into a problem that no amount of long-term contracts can solve overnight: factories take years to build, and the clean rooms needed for high-bandwidth memory (HBM) are already running flat out. Micron’s customers — hyperscalers and automakers — are signing multi-year take-or-pay agreements, but the supply response remains stuck until at least late 2026. That tension between locked-in revenue and physical capacity limits sits at the heart of a stock that has fallen nearly 13 percent in a single week, even as the business itself reports some of the strongest fundamentals in its history.

Shares of Micron now trade around €795.40, some 28 percent below the 52-week high of €1,103.80 reached on June 25. The weekly slide accelerated after Samsung posted a near-20-fold jump in second-quarter operating profit, only to see its own stock drop as investors concluded the memory cycle may be peaking. That same mood washed over Micron, whose stock has shed about a quarter of its value from the record. The decline has caught the attention of high-profile bears: investor Michael Burry disclosed a short position initiated at €1,051.87, a bet that the artificial-intelligence-driven frenzy for memory chips is approaching its zenith.

Yet the fundamental picture tells a different story. Micron shares have still gained 649 percent over the past twelve months and remain more than 777 percent above the 52-week low of €90.64. The divergence between price action and earnings power has widened into what analysts describe as a reassessment of an entire industry’s valuation.

The real bottleneck is not demand — it is capacity. Every wafer that Micron, Samsung, and SK Hynix allocate to HBM stacks is one less for commodity DRAM or NAND flash. The hyperscalers’ appetite for HBM is so voracious that even NAND producers are converting lines to DRAM to chase higher margins. This structural reallocation has pushed pricing power into territory that analysts at TrendForce say could last at least through 2027. New fabrication plants from Micron, Samsung, Kioxia, and others will not reach meaningful output levels until late 2026 or 2027; Micron’s own ID1 fab in the United States is not expected online before 2027. Money alone cannot shorten the lead time needed to install lithography tools and qualify chips.

Should investors sell immediately? Or is it worth buying Micron?

Micron has tried to insulate itself from the traditional boom-bust cycle by locking in customers. The company has signed 16 long-term supply agreements, including a recently announced deal with Ford for automotive-grade DRAM and NAND. Those contracts typically include take-or-pay clauses, price floors, and advance payments that already run into the billions of euros. The idea is to transform a volatile commodity business into a predictable infrastructure partnership with the giants of artificial intelligence. For the bull case, these agreements prove that the current demand is structural, not cyclical. For the bears, they merely postpone the reckoning when oversupply eventually returns.

The chart suggests a market still trying to find its footing. The relative strength index has cooled to 46.1, well below the euphoric levels of June. At €795, the stock sits within 5 percent of its 50-day moving average of €786.40, but remains more than 105 percent above the 200-day average of €401.94. That gap underscores how rapidly the market has repriced Micron’s earnings power over the past year. The annualized 30-day volatility of 111 percent confirms that confidence remains shallow: a single earnings miss or capacity announcement could swing the stock violently.

Analysts remain broadly optimistic. The average 12-month price target stands at €1,298.99, implying upside of roughly 63 percent from current levels. That consensus has held despite the recent slide, suggesting the Street has not abandoned the supercycle thesis. Yet with the RSI in neutral territory and the stock stuck between technical support and resistance, investors are effectively betting on whether the next quarterly report — due in September — will validate the take-or-pay model or expose its limits.

What makes this cycle different from earlier booms is the source of demand. Previous peaks were driven by PC replacement cycles or smartphone upgrades; this one is built on hyperscaler spending on artificial intelligence infrastructure that shows no sign of easing. The question is whether that spending is as structural as its proponents claim. For a stock that has already multiplied its value nearly eight times from the low, the wager now is whether new fabs can arrive fast enough to satisfy demand without crashing the pricing environment. The answer will not come from quarterly filings alone — it will take years of factory construction and contract performance to settle.