Cathie Wood’s ARK sells Meta and Nvidia stocks, buys Tempus AI

Key trades: ARK sold 76,622 shares of Meta (META) for $45,581,661 and 155,441 shares of Nvidia (NVDA) for $27,774,197, while purchasing 60,973 shares of Tempus AI for $2,851,097. Other notable sells included AMD $8.42M, Deere $5.46M, Block $5.18M and Roku $7.20M; Archer reduction was $2.43M. The activity signals ARK is trimming large-cap tech and industrial positions and reallocating into smaller AI/health-related names. These moves are likely to modestly affect individual stock flows but are unlikely to move broad markets.

Analysis

ARK’s rebalancing should be read less as a sector repudiation and more as a liquidity-driven rotation: large passive/ETF sellers in mega-cap AI names remove a portion of the short-term marginal bid, raising dispersion and creating intra-sector opportunities for hardware and niche AI plays that sit downstream of GPU deployment. That creates a two-week to two-month window where supply/demand technicals matter more than fundamentals — expect higher realized volatility in flagship semis and greater bid for smaller-cap, fast-growth AI stocks as active managers and quant funds hunt for idiosyncratic alpha.

Second-order winners are vendors of data-center infrastructure and OEM server vendors that capture incremental spend when enterprises scale GPUs (SMCI) and monetizers of AI applications with clear revenue hooks (APP-style monetization). Conversely, index-heavy liquidity outflows amplify borrowing costs and worsen intraday pricing for the largest caps, which can temporarily compress multiples even if growth remains intact. Key catalysts that will reverse these rotational moves are earnings guidance on GPU bookings (1–3 months) and any surprise upgrades to cloud capex roadmaps (2–6 months); macro shocks or a resurgence in risk-off flows can flip sentiment inside days.

The consensus interpretation — that ETF trimming equals secular weakness — is likely overdone. ETF flows are cyclical and often concentrate selling into market leaders, leaving durable demand signals (orders, ASPs, OEM backlogs) as the true arbiter of 6–24 month returns. A disciplined approach that isolates flow-driven dislocations from demand-driven structural wins will produce asymmetric returns over the next 3–9 months.

Key Decisions for Investors

• Long SMCI 3–6 month call spread (buy 6-month ATM call, sell 6-month +30% call). Entry: on a pullback to the 10-day MA or a 5% intraday drop. Target: 40–60% upside on spread if server bookings tick up; max loss = premium paid (~100% of premium).
• Buy APP Jan 2027 LEAPS (or 12–18 month calls) sized as a growth allocation. Rationale: exposure to ad/monetization reacceleration and AI-driven ARPU improvements. Risk: binary if user metrics disappoint; hedge with a 20% position size cap.
• Tactical pair: long TEM (small-cap AI healthcare) vs short a synthetic NVDA exposure (e.g., short-dated vertical puts on an NVDA ETF or reduce outright NVDA exposure). Timeframe: 1–6 months to capture rotation into niche AI names; reward skewed if TEM secures commercial contracts, loss capped by option premiums on the short leg.
• If NVDA gaps down >8% on flow headlines, use that as an opportunistic add for a 3–9 month hold — size conservatively and consider selling 1/3 covered calls after the first 30% move to monetize elevated implieds.