3 Chip Giants Still Worth Buying Despite Massive Gains

The article argues that Micron, Western Digital, and Taiwan Semiconductor remain attractive despite sharp rallies, citing strong AI-driven demand, sold-out capacity, and still-reasonable valuations. Micron is up 297% YTD and 831% over 12 months with a forward P/E of 10; Western Digital is up 333% YTD and 1,160% over 12 months with a forward P/E of 39 and PEG of 0.66; TSMC is up 54% YTD and 116% over 12 months with a forward P/E of 29. The piece is broadly constructive on AI chip and storage stocks, though it is advisory rather than event-driven.

Analysis

The market is rewarding the AI supply chain where capacity is scarce, not just where demand is visible. That matters because the economic moat is shifting from model-layer excitement to bottleneck ownership: memory, storage, and foundry capacity now have the best pricing power, and the next leg of earnings is likely to come from contract repricing rather than unit growth alone. The fact that these names are already up massively is less important than the duration of their backlog and the evidence that customers are signing multi-year supply agreements to avoid being shut out.

The second-order effect is that the usual AI winners may be capped by physics while these infrastructure names still have operating leverage. If memory and storage remain sold out into the out-years, downstream GPU/system vendors face a hidden tax: they need to keep deploying capital just to preserve performance, but their gross margins may get squeezed by rising input costs. That creates relative value inside semis: the best risk-adjusted exposure is not necessarily the obvious compute leaders, but the picks-and-shovels vendors with contracted demand and visible price discipline.

The main risk is not secular demand decay; it is cycle normalization and capital misallocation. These stocks are vulnerable if supply comes on faster than expected, if customers start design-to-cost substitutions, or if hyperscalers stretch replacement cycles after the current buildout wave. The time horizon is months to years: near-term momentum can continue, but any guidance miss will be punished because valuation now embeds a multi-year scarcity premium rather than just current growth.

Contrarianly, the move may still be underdone in relative terms because investors often anchor on headline multiples instead of earnings revision velocity. If forward estimates continue to ratchet up, today’s seemingly rich valuations can compress quickly. The cleaner trade is to own the bottleneck beneficiaries while fading parts of the AI stack where supply is abundant and price competition is more likely to reappear first.

Key Decisions for Investors

• Maintain a tactical long MU / WDC basket for the next 3-6 months; the setup favors continued estimate revisions as long as backlog visibility remains intact. Use a 15-20% trailing stop because any sign of supply normalization would hit these names hard.
• Prefer TSM over the higher-beta memory names for a 6-12 month hold; it offers AI exposure with better earnings durability and lower execution risk. Expect lower upside in a straight-line rally, but materially better downside capture if the AI trade de-risks.
• Pair trade: long MU, short a weaker broad hardware/semicap proxy if you want to isolate bottleneck pricing power from the rest of tech. The thesis is that scarcity rents accrue to constrained components faster than to more competitive system integrators.
• Consider buying WDC call spreads 3-6 months out rather than stock for defined-risk participation in momentum continuation. This is appropriate only if you believe guidance revisions will keep outpacing multiple compression.
• Avoid chasing the obvious compute beneficiaries here; rotate capital toward the supply-constrained enablers until pricing power clearly inflects lower. If customer concentration rises or lead times shorten, that is the signal to cut exposure.