Cathie Wood's Ark Invest increases positions in Nvidia and TSMC as Meta's massive AI spending surge validates multi-year capex thesis.
Cathie Wood's Ark Invest executed a rapid-fire, two-day buying spree in the heart of the artificial intelligence supply chain, snapping up roughly $15 million of Nvidia stock on July 28 and $14.7 million of Taiwan Semiconductor Manufacturing Co. shares on July 29. The purchases, disclosed in daily trading data, landed squarely in the window of Meta Platforms' second-quarter earnings report, a filing that underscored an accelerating, multi-hundred-billion-dollar race to build out AI infrastructure.
On July 28, Ark spread its Nvidia purchases across five of its exchange-traded funds, with the flagship ARK Innovation ETF accounting for $8.1 million of the total. The very next day, Wood's team extended the bet downstream, buying TSMC across four Ark funds. The concentrated buying reflects Wood's conviction that AI demand is far from peaking, and that the biggest near-term profits will flow to the enablers—the companies that design and manufacture the infrastructure of the AI revolution.
Meta's latest financial report provided the fundamental backdrop. The company raised its 2026 capital expenditure guidance to $130 billion to $145 billion, tightening the lower end upward from a prior forecast of $125 billion to $145 billion. In the second quarter alone, Meta spent $31.1 billion in capital outlays. Management stressed that the company is "supply-constrained" and actively prioritizing near-term compute capacity to train AI models, power AI agents, and expand its global data center footprint.
For Wood, that narrative translates into a direct line to Nvidia's revenue. Nvidia commands a dominant share of the GPU market for AI workloads. A sustained and still-growing capex budget at a hyperscaler the size of Meta is widely expected to convert into fresh orders for Nvidia's silicon. By purchasing Nvidia on the heels of Meta's report, Ark is wagering that the social-media giant's infrastructure spending will continue to be a primary catalyst for Nvidia's top-line growth.
But Wood's simultaneous purchase of TSMC reveals a deliberate, full-stack approach. Nvidia designs the world's most sought-after AI chips but fabricates none of them. That manufacturing is outsourced almost entirely to TSMC, the world's largest contract chipmaker. An increase in Nvidia GPU shipments mechanically drives higher wafer starts at TSMC's advanced fabrication plants, linking the two companies at the production level. By purchasing both stocks, Wood is capturing the entire AI hardware value chain—from architecture design to the manufacturing profit generated when hyperscalers like Meta, Microsoft, and Amazon write ever-larger checks for Nvidia-based computing clusters.
TSMC's financial performance underscores this coupling. In the second quarter, the company posted revenue of $40.2 billion, a 36 percent year-over-year surge, with net income catapulting 77.4 percent.
The scale and timing of the trades mark a notable moment for Ark, which has historically been more associated with high-growth, software-centric disruptors. The decision to channel nearly $30 million into two of the world's largest semiconductor names over two consecutive days suggests a tactical pivot toward the physical infrastructure that underpins AI, rather than just the applications that run on top of it.
The valuation backdrop may have made the decision easier. Both Nvidia and TSMC currently trade at approximately 25 times forward earnings. That multiple is well below the elevated levels witnessed during earlier, more speculative phases of the AI investment cycle. For investors who share Wood's view that hyperscaler capital expenditure budgets will remain elevated for several more years, the current valuation profile appears more grounded than it did when AI hype was at its peak.
Analysts tracking the semiconductor space have been raising price targets for companies exposed to data center spending, citing visibility into multi-year buildout cycles. The logic underpinning Ark's trades—that hyperscalers' capex is sticky and will continue to flow to Nvidia and TSMC—is increasingly becoming consensus rather than a contrarian call.
The trades are not without risk, however. A slowdown in Meta's advertising revenue, a shift in AI model efficiency that reduces the need for brute-force computing, or geopolitical disruptions affecting TSMC's manufacturing base in Taiwan could all pressure the thesis. While Wood's conviction is clear, her flagship ARKK fund has experienced sharp drawdowns in the past when growth-oriented narratives have faltered.
On August 7, roughly a week and a half after loading up on the two chip stocks, Ark sold more than 1.5 million Roblox shares, worth an estimated $57.6 million, while building a new, roughly $32.4 million position in Cloudflare. That trade followed Cloudflare's August 6 report of $696 million in quarterly revenue, up 36 percent year over year, and a raised full-year outlook. The rotation out of a consumer-facing gaming platform and into enterprise infrastructure adds to the broader picture of Ark repositioning around AI-linked buildout themes.
For retail investors watching Wood's moves, the signal concerns the direction of capital flows as much as the specific stocks. After years of the hyperscalers themselves—Meta, Microsoft, Alphabet, Amazon—dominating AI-related returns, capital may be rotating toward picks-and-shovels providers. Wood's latest bets are a high-profile endorsement of that shift.
The coming quarters will test the thesis. Meta's capex trajectory, Nvidia's ability to maintain its GPU dominance, and TSMC's execution on next-generation process nodes will all be key variables. Nvidia's earnings call, scheduled for August 26, will be the first results the company reports since Ark's purchases, offering an early read on whether hyperscaler order flow is keeping pace with Wood's bet.