Billionaire Steve Cohen Sold Amazon and Nvidia but Loaded Up on This Beaten-Down Stock

Point72’s Steve Cohen trimmed AI leaders Amazon (AMZN) by 6% and Nvidia (NVDA) by 24% in Q4’25–Q1’26 as both lagged the S&P 500, with each up only ~5% YTD in 2026. He materially increased Boston Scientific (BSX) by 50% from the start of the year while the stock is down over 50% YTD, betting on a long-term turnaround despite weaker outlook (Q2 organic sales growth guidance slowing to ~5% vs 9.4% in Q1; FY adjusted EPS guidance ranges narrowed/lowered to ~$3.34–$3.41 from ~$3.43–$3.49). The news is more about hedge-fund positioning than near-term fundamentals, but it reinforces a cautious rotation away from “early AI leaders” toward infrastructure and defensive healthcare demand.

Analysis

The 13F signal is less about a fundamental collapse in AI and more about crowding management. When a prominent multi-strat trims NVDA/AMZN after a multi-year run, the short-term implication is not “AI is over,” but that marginal capital is moving from monetization names to enablers; that can keep valuation dispersion wide for 1-3 months even if earnings stay fine. The risk is that momentum-owned mega-caps lose sponsorship before the next earnings reset, which can pressure multiple expansion more than near-term revenue estimates.

Boston Scientific looks like a different kind of setup: a high-quality operator that has already been repriced for a bad year. If execution stabilizes, the stock can rerate sharply because the market is discounting the guidance cut as structural rather than transitory. The main second-order effect is on healthcare breadth: if BSX works, it supports a broader rotation into durable growth at a discount, which can pull capital from expensive AI breadth names into underowned medtech.

Contrarian view: the consensus may be over-reading the “AI rotation” narrative from a stale filing. A 13F says what a manager owned 90 days ago, not what they think now; if NVDA’s next capex/order commentary re-accelerates, the sell-off could reverse quickly. The more actionable takeaway is that the market now requires proof of cash-flow conversion, not just AI exposure. That favors infrastructure names with visible bookings, and punishes the highest-multiple beneficiaries if growth decelerates even modestly.

Key Decisions for Investors

• Do not chase AMZN/NVDA weakness purely off the filing; treat any further 3-5% underperformance into the next earnings cycle as a better entry point for a tactical fade only if AI capex commentary remains firm.
• Pair trade: long BE / short NVDA for 1-3 months as a rotation expression if power-infrastructure demand and order flow hold; risk is that NVDA re-accelerates on next quarter guidance and the pair widens against us.
• Long BSX only on confirmation of stabilization: wait for a quarter where organic growth stops decelerating and FY guidance is held or raised; otherwise this remains a value trap risk. Consider a small starter position, not full size.
• Use SOXX or SMH weakness as a hedge against any long BE exposure; the hedge should be reduced if NVIDIA earnings and hyperscaler capex commentary reassert the AI-growth trade.
• Set a falsifier on BSX: if the next quarterly guide is cut again or organic growth stays near the low end of the current range, exit—there is no valuation support if the market starts discounting a second guidance reset.