David Tepper Sold 41% of His Micron Shares and It Is Still His Second-Biggest Holding

Appaloosa Management cut its Micron stake by 41% (sold 690,000 shares), but still held 975,000 shares worth about $1.13B at end-June (~15% of the fund), after MU rose ~242% in the quarter. The article highlights Micron’s cycle-driven surge: fiscal Q3’26 revenue $41.5B (from $9.3B YoY) and net income $28.2B (about 15x), with AI data-center demand boosting cloud memory revenue to $13.8B. Despite the strong run, the write-up flags decelerating guidance step-ups and cyclical risk—leading to a “hold, don’t chase” stance.

Analysis

The important read-through is not that a large holder trimmed, but that memory has moved from “under-owned recovery” to “crowded earnings compounding.” When a stock rerates this fast, the first-order winners are the vendors with the tightest allocation discipline and the cleanest exposure to AI server builds; the losers are lower-quality NAND-heavy names and downstream hardware customers that can’t fully pass through cost inflation. In that setup, MU is the better house in a strong neighborhood, while SNDK is more vulnerable if pricing normalizes even modestly.

The market is likely underpricing how quickly cyclicals can stop working once the guide stops accelerating. The next 1-3 months matter more than the 6-18 month story: if sequential revenue growth decelerates again and the next guide only “meets” instead of “beats,” the multiple can compress hard even with still-strong earnings. The falsifier for the bullish cycle view is not an earnings miss today; it is evidence that HBM/DRAM tightness is easing, inventory days are rising, or competitor capex is finally converting into supply faster than expected.

Contrarian take: the 13F trim may actually be supportive because it suggests disciplined position sizing rather than a fundamental exit. Consensus is treating MU like a secular AI beneficiary, but the cash flow is still highly path-dependent and the stock is already pricing in a long runway. That makes the best asymmetry a relative-value expression, not an outright chase: own the best quality memory exposure, but hedge the beta and avoid paying peak-cycle multiples for the wrong part of the stack.

Key Decisions for Investors

• Maintain/accumulate MU only on pullbacks over the next 2-6 weeks; avoid adding after gap-up days. Risk/reward is favorable only if next quarter guide still implies strong sequential growth; trim if the company stops beating revenue by a wide margin or gross margin rolls over.
• Pair trade: long MU / short SNDK for 1-3 months. MU has cleaner AI-linked earnings power and better pricing leverage; SNDK is more exposed to a faster NAND normalization. Target relative outperformance if memory pricing stays tight; cover if NAND spot prices inflect higher.
• For existing MU longs, collar 3-6 month exposure: buy 10-15% OTM puts financed with 20-25% OTM calls. This protects against a peak-cycle multiple reset while keeping most of the upside if AI memory demand remains intact.
• Set an alert on the next MU guidance update and HBM allocation commentary. If sequential revenue growth decelerates materially or management hints at capacity easing, expect a 15-25% drawdown risk in the stock even without an earnings miss.
• If looking for indirect AI exposure with lower cycle risk, prefer NVDA over added MU at these levels. MU can outperform in the short run on tight supply, but NVDA is less exposed to a sharp memory pricing reversal.