One Huge Fund Just Bought Taiwan Semiconductor Manufacturing and Dumped Micron Stock. Which Stock Is the Superior Choice?

Appaloosa Management (David Tepper) cut its Micron (MU) stake by over 40% in Q2 while adding nearly 25% to Taiwan Semiconductor (TSMC), taking TSMC to ~11% and its third-largest position. The article argues memory chips (Micron) are benefiting from near-term supply shortages with higher potential upside, but expects normalization as soon as 2028, while TSMC’s logic-chip exposure positions it to capture the AI build-out over a 3–5 year horizon. Overall takeaway is a time-horizon-dependent view rather than a clear bullish/bearish call.

Analysis

The market is likely underestimating the difference between a bottleneck business and a cyclical shortage business. TSM can compound through multiple AI cycles because it sits at a structural choke point; that supports a persistent quality premium versus the broader semiconductor tape, while also transmitting demand to toolmakers like ASML, AMAT, LRCX and KLAC. By contrast, MU’s upside is mostly a function of near-term pricing discipline, which tends to attract supply response and compress forward returns once capex turns less rational.

In the next 1-3 months, the cleanest read-through is on earnings revisions: if memory pricing and HBM availability stay tight, MU can keep outperforming on margin surprises, but that upside is brittle. The risk is that the whole sector starts to spend aggressively to relieve shortages, creating a 6-18 month oversupply set-up that usually hits memory names first and then drags on storage-heavy end markets such as PCs, smartphones and enterprise hardware. That would also pressure gross margins at OEMs and hyperscale buyers if input costs remain elevated.

The contrarian point is that the crowd may be too comfortable calling TSM the "expensive" one and MU the "cheap" one without adjusting for duration. TSM’s premium is justified if AI capex stays concentrated in advanced nodes, while MU is only cheap if the shortage persists longer than the market expects and capital allocation remains disciplined. The key falsifiers are a sharp normalization in DRAM/HBM pricing or a TSM guide-down on utilization/margin, either of which would reverse the relative trade quickly.

Key Decisions for Investors

• Favor a 6-12 month relative-value long TSM / short MU pair on strength in MU and dips in TSM; thesis is quality-duration premium versus cyclical shortage beta, with the pair exposed to memory oversupply normalization. Falsify if MU keeps raising pricing/guide for multiple quarters and TSM margin expansion stalls.
• If you want outright exposure, buy TSM on pullbacks rather than chasing momentum; it is the cleaner way to express AI infrastructure compounding without relying on spot commodity pricing. Best entry is after any post-earnings consolidation when implied growth expectations reset.
• Treat MU as a tactical trade, not a core compounder: only add on selloffs if DRAM/HBM pricing data remains tight and channel inventories stay lean. The risk/reward is better for a 1-2 quarter hold than a 1-2 year hold.
• Use AMAT/LRCX/KLAC as a lower-beta way to own both outcomes; they benefit if either TSM keeps investing or MU/supply peers add capacity. This is the cleaner long if you want semiconductor capex exposure without picking the winner in the product cycle.
• Watch the next two quarterly prints for HBM commentary, utilization, and capex guidance; if memory producers signal aggressive expansion, reduce MU exposure first and expect relative multiple compression to begin within 3-6 months.