Investing Legend John Templeton Has a Warning for Micron and SK Hynix Investors

The market is treating memory like a quasi-infrastructure annuity, but the business model still behaves like a commodity with a long lead-time supply response. That creates a classic second-order trap: the more confidently capital is poured into new capacity, the more the eventual margin peak is pulled forward, and the more depreciation/underutilization becomes the next earnings headwind for MU and SKHY. The best near-term beneficiaries are not the memory vendors themselves but the AI system stack names with pricing power — especially NVDA — and, farther out, hyperscalers that get component-cost relief if memory normalizes.

The key catalyst path is time-dependent. Over the next 1-3 months, this can stay bid if HBM pricing and allocation remain tight; the bearish setup only works if you see inventory builds, a pause in hyperscaler capex, or any sign that long-term supply agreements are no longer enough to defend pricing. Over 6-18 months, the structural risk is capacity coming online into a demand growth deceleration: that is when revenue growth can remain decent while margins and FCF collapse, and valuation usually compresses before the earnings inflect.

The contrarian point is that the consensus may be underestimating how long qualification bottlenecks and advanced packaging constraints can keep the cycle tight. In other words, the thesis is not that memory is broken now; it is that current valuations already discount a very durable scarcity regime. That argues for expressing the view with optionality or relative value rather than a naked short, and for watching next-quarter pricing commentary as the main falsifier.