Cathie Wood’s ARK Invest bought Nvidia and Taiwan Semiconductor Manufacturing Co. stock in the trading days after Meta Platforms’ disappointing earnings report, while trimming part of its AMD stake, according to the firm’s daily trading disclosures, reported by the Globe and Mail on August 9. The purchases extend a pattern that has held for months: adding to the two companies at the center of the AI computing supply chain, and reducing exposure to Advanced Micro Devices.
The timing is telling. Meta reported second-quarter earnings on July 31 that missed estimates, with costs up 55 percent from a year earlier and revenue guidance below Wall Street’s forecast; the stock fell nearly 8 percent in a single session. The morning after, ARK bought 26,509 Meta shares across three ETFs for about $14.3 million at $539.03 a share, then added to its Nvidia and TSMC positions over the following days. In late July, before the Meta report, ARK had already bought roughly $15.5 million of Nvidia and about $20 million of TSMC, while adding a small SpaceX position.
Wood’s logic has been consistent through the AI cycle. Her funds buy the companies that execute on the compute chain, the chip designers and foundries whose products every AI model depends on, and sell the names whose growth is already priced in. The AMD trimming fits that frame: analysts note AMD holds only a modest slice of the AI accelerator market, squeezed between Nvidia’s dominance and the custom silicon that hyperscalers are designing themselves. In the same stretch, ARK sold about $18.2 million of AMD and $2.7 million of CrowdStrike, while buying roughly $20.6 million of Broadcom and about $13 million of Cerebras, the wafer-scale chip startup, a rotation that sharpens the firm’s bet on AI infrastructure over conventional security software.
The transactions are small by ARK’s standards, and that is part of the story. ARK’s assets under management have shrunk from a peak of roughly $60 billion in early 2021, and the flagship ARK Innovation ETF trades far below its February 2021 high. Wood’s purchases no longer move markets the way they did in the pandemic-era boom, when her daily disclosures were treated as a buy signal by a large retail following. The Meta buy on July 31 drew attention mostly because of its contrarian timing, not its size.
What the pattern does show is conviction. Wood has repeatedly said AI infrastructure spending is a near-term cost for the companies paying it but a long-duration revenue driver for the suppliers, and her funds have acted on that view through the cycle, buying Nvidia dips in late July and again in August. The position contrasts with well-known bears, including Michael Burry, whose funds have bet against Nvidia at various points this year. For ARK, the trade is a bet that the compute build-out outlasts the earnings volatility that punctuates it.
The Nvidia and TSMC purchases also reflect a view about where the AI market’s bottlenecks sit. Nvidia holds the dominant share of AI accelerators and most of the advanced packaging capacity that AI chips require; TSMC manufactures nearly all of it. Whatever happens to individual AI applications or model companies, the reasoning goes, the suppliers of silicon and manufacturing capacity collect a toll. Wood has described the pair as the most direct way to own the AI build-out.
Wood’s history with Nvidia is itself a study in timing. ARK sold most of its Nvidia stake in early 2023, before the AI boom pushed the stock to successive records, and Wood has acknowledged the firm misread the pace of the rally. The recent purchases are, in part, an attempt to rebuild a position the firm exited too early, and they carry the weight of that earlier decision. Analysts who follow ARK note the firm’s public arguments have shifted accordingly: Wood now describes Nvidia as a toll collector on AI infrastructure, a role she once assigned to a broader basket of innovative software companies.
Skeptics note that ARK’s fund flows have been negative for years, meaning the firm’s buying has not translated into performance that attracts capital. The disclosures remain closely watched by retail investors and by the trade-tracking services that monitor Wood’s moves, but their market impact has faded. Even so, the rotation into Nvidia and TSMC is a clear signal of where the firm sees value in a market dominated by earnings misses and infrastructure spending.
For TSMC, ARK’s interest is part of a broader institutional embrace: the foundry’s stock has been a favorite of growth funds betting on AI demand. For Nvidia, the purchases add to a base of shareholders who have treated every pullback as a buying opportunity. Whether that patience is rewarded depends on whether AI revenue keeps compounding at the pace the bulls expect. Wood’s record this cycle is mixed; her flagship fund remains far below its peak even as Nvidia and TSMC have soared. But the direction of her trades is unambiguous: when the market flinches, she buys the two companies that stand to gain from whatever happens next.
