Memory Chips Just Fell Into a Bear Market. Micron Is Down 30% From Its High Even as AI Demand Booms.

The market is treating this as a cycle-peak call, but the more important mechanism is that AI memory has become a bottlenecked input rather than a generic commodity. That changes the winner set: the strongest balance sheets with HBM exposure can keep converting pricing power into free cash flow, while slower-moving rivals are forced to chase capacity at the worst point in the cycle. The second-order beneficiary is the equipment layer, because sustained supernormal margins incentivize capex even if equity investors are already de-rating the end market.

The risk is timing, not the current quarter. If spot DRAM/HBM prices stay firm for another 1-2 quarters, the bear case gets pushed out and the stock can re-rate quickly; if pricing rolls over, equity holders will discount the next 6-12 months before the income statement shows it. Watch for any sign that Samsung or SK Hynix is adding capacity faster than demand, because that is the classic trigger for gross margin compression from the 80s back toward normalized levels. The thesis is falsified sooner if management guides gross margin below the low-80s or if order momentum weakens on the next print.

Contrarianly, consensus may be underestimating how sticky AI memory demand is versus prior handset/PC cycles, because AI server builds are less elastic and HBM supply is harder to flood. That argues the selloff may be overdone in the near term even if the long-duration valuation is still vulnerable. The best expression is relative value, not an outright hero trade: own the highest-quality memory exposure while fading the weakest links in the global memory complex. If the market is right and this is the top, the short side will work fast; if it is wrong, the upside in the leader should compound through the next two quarters.