Cathie Wood’s ARK sells AMD stock, buys CoreWeave and Kratos By Investing.com

ARK disclosed a mixed set of daily trades for May 8, 2026, selling 46,034 AMD shares (~$18.8M) while buying 113,076 CoreWeave shares (~$14.6M), 168,441 Kratos shares (~$9.6M), and 394,156 Kodiak AI shares (~$3.6M). It also added to Cloudflare, X-Energy, and Intellia, while trimming Rocket Lab (187,612 shares, ~$14.7M), Teradyne, and Twist Bioscience. The article is primarily a portfolio-flow update and reflects ongoing rotation rather than a broad market catalyst.

Analysis

The flow signal matters more than the names: this reads like a rotation away from high-beta “story” semis and space toward businesses with more defensible revenue visibility and tighter links to current capex cycles. That usually helps the second tier of beneficiaries—cloud networking, defense autonomy, and select biotech platforms—because passive and quant sellers tend to amplify the same factor unwind once momentum stalls.

The most important second-order effect is on crowded positioning. When a high-profile allocator trims a name like AMD after a strong tape, it can create a short-lived air pocket not because fundamentals broke, but because marginal buyers disappear and systematic de-risking kicks in. Conversely, repeated adds to CRWV/KTOS-type exposures can extend the life of those trades for several weeks, especially if they sit in themes where investors are still under-owned relative to AI infrastructure and defense spend.

The contrarian take is that this is less a “risk-on pivot” than a quality filter inside risk-on. If index highs are driven by narrowing participation, names that benefit from bespoke buying should outperform on a 1-3 month horizon even if the broader tape pauses. The biggest reversal risk is that any disappointment in AI capex or defense budget headlines would hit the newer buys first, because they trade on narrative premium rather than mature cash flow.

For time horizon, the negative signal in RKLB/TER is more actionable over days to weeks than months: these are the kinds of holdings where ETF selling can depress sentiment disproportionately versus fundamentals. In contrast, KTOS and NET can keep compounding if the market continues rewarding infrastructure and national-security exposure, but they are vulnerable if rates back up and long-duration growth gets repriced.

Key Decisions for Investors

• Short AMD for 2-4 weeks or use put spreads: better risk/reward than chasing the index because the name is vulnerable to marginal-flow exhaustion; cover if semis regain leadership breadth.
• Pair long KTOS / short RKLB for 1-2 months: both sit in adjacent defense/space baskets, but KTOS has a clearer catalyst path and should absorb the same thematic bid with less execution risk.
• Add a tactical long in NET on any intraday weakness over the next 1-3 weeks: this is a cleaner beneficiary of AI infrastructure spend than more speculative hardware names, with lower downside if the market rotates within tech.
• Avoid chasing CRWV after a strong sponsor-led bid; if you want exposure, wait for a post-flow consolidation and buy the next pullback rather than paying up into momentum.
• Use TER as a hedge against the AI-capex trade: shorting it for 1-2 months is attractive if the market starts questioning the breadth of semiconductor equipment demand outside the strongest AI nodes.