President Trump Sells Micron Stock and Buys an AI Stock Up 1,340% Since 2023

U.S. Office of Government Ethics disclosures show Trump’s managed accounts net sold Micron, with net sales of about $90,000–$116,000 YTD through May, while net buying Nvidia of about $246,000–$3.7 million. Despite Micron’s Q2 FY2026 surge (revenue +345% to $41.4B; non-GAAP EPS +1,200% to $25.11) and 16 multiyear supply agreements, Wall Street expects memory pricing to soften as soon as 2028. Nvidia meanwhile reported Q1 FY2027 strength (revenue +85% to $81.6B; non-GAAP EPS +140% to $1.87) and highlighted AI infrastructure catalysts including Vera Rubin volume production and the RTX Spark PC superchip launch.

Analysis

The useful signal here is not the political-account flow; it is the market’s preference for duration and platform control over pure component leverage. NVDA’s stack integration keeps pulling more economics from the AI buildout, which should continue to squeeze AMD/INTC on performance-per-watt and leave OEMs such as DELL/HPE/MSFT with less bargaining power but better attach rates on premium systems. That matters over 1-3 months as product launches hit the channel, and over 6-18 months as AI PCs become a refresh cycle rather than a one-off pilot.

MU is the cleaner cyclical-to-structural transition story. Long-term supply agreements should dampen near-term earnings volatility and support a higher multiple, but they do not eliminate the core risk that memory pricing remains the most reflexive leg of the AI supply chain once capacity catches demand. In practice, MU now looks more like a quality cyclical with a floor than a pure mean-reversion trade; that is bullish for valuation, but also means upside is more sensitive to HBM mix and contract pricing than to headline unit growth.

Contrarian view: consensus is likely underestimating how crowded NVDA already is as an obvious AI winner, while underappreciating how much MU’s visibility has improved. The better near-term trade may be relative, not absolute: own the enabler with recurring platform share, but avoid paying peak enthusiasm for the most crowded beta to AI capex. Falsifiers: NVDA thesis breaks if data-center growth or gross margin decelerates on any 1-2 quarter basis; MU thesis breaks if DRAM/HBM pricing rolls over before contract floors meaningfully offset it.

Key Decisions for Investors

• Long NVDA on any 3-5% pullback over the next 2-4 weeks; use a 3-6 month horizon. Risk/reward favors continued multiple support if Vera Rubin ramps on schedule, but trim if data-center growth or gross margin cracks in the next earnings print.
• Short AMD vs long NVDA as a relative-value pair into the AI PC and systems-launch cycle. The trade works if NVDA’s stack integration widens share while AMD remains a closer substitute; cover if AMD closes the performance gap or wins materially more OEM design wins.
• Accumulate MU only on weakness, not strength, over the next 1-3 months. The contract story reduces downside volatility, but the upside case depends on HBM tightness persisting; invalidate if memory spot pricing starts to roll over or management guides to easing supply conditions.
• If looking for a cleaner hedge, pair long NVDA with short INTC on a 6-12 month view. INTC is most exposed to NVDA’s move into CPU-adjacent workloads and AI PC attach, with less room to absorb share loss.