Cathie Wood Just Went Bargain Hunting. Here Are the 3 AI Stocks She Bought.
Cathie Wood doesn't care what you think.
The ARK Invest founder has spent the better part of 2026 watching her flagship fund trail the S&P 500. The ARK Innovation ETF (ARKK) is up just 11.8% year to date. The broader market? It's sitting at 13%.
Not a disaster. But not the kind of performance that made Wood a household name during the pandemic boom.
Yet here she is. Buying. Not selling. Not hiding in cash. Buying.
In late August, Wood directed her firm to scoop up shares of three AI companies. The total tab? Roughly $90 million. Spread across Nvidia, Cerebras Systems, and Broadcom.
This isn't random. It's a rotation. A bet. A statement.
Let's look at what she bought and why it matters.
The State of ARK Invest in 2026
Let's be honest about where things stand.
ARKK isn't crushing it this year. Only one of ARK's seven ETFs, the Genomic Revolution fund, has outperformed the market. Wood's flagship is lagging.
But here's the thing about Cathie Wood. She doesn't trade for the next quarter. She trades for the next decade. Her research projects AI spending will triple from $500 billion in 2025 to roughly $1.5 trillion by 2030.
That's a ten-year horizon. And Wood is positioning for it now.
She's also operating on a thesis she calls the "SaaSpocalypse", the idea that traditional software-as-a-service models are getting disrupted by AI. One-size-fits-all software is dying. Custom AI infrastructure is the future.
That thesis explains everything she did in August.
The 3 AI Stocks Cathie Wood Just Bought
1. Nvidia (NVDA) – The $55 Million Bet
Nvidia is the elephant in the room. Always has been. Always will be.
On August 28, ARK Invest bought 243,707 shares of Nvidia across five of its ETFs. The purchase was worth roughly $55 million.
That made Nvidia ARK's largest purchase of the month. By a lot.
Why Nvidia? The numbers tell the story. The company just reported 106% year-over-year revenue growth, hitting $96 billion. Management is guiding for 70% growth in fiscal 2028.
Those are absurd numbers. But they reflect a simple reality: AI runs on Nvidia.
The company's GPUs power essentially every major AI model in existence. Sovereign large language models across more than 80 countries use Nvidia chips. That's not a monopoly. It's closer to a utility.
Competition is heating up. Google's Tensor Processing Units and Amazon's Trainium chips can rival Nvidia's performance for certain workloads. Custom chips are expected to gain share in the coming years.
Wood seems to think the upside outweighs the risk.
Nvidia is working with major financial firms, Apollo, Blackstone, BlackRock, Goldman Sachs, KKR, to arrange more than $500 billion in financing for AI infrastructure projects. That's not small. That's world-changing.
Wood bought the dip. On August 10, Nvidia stock fell 2.86% following reports of this financing arrangement. Wood added shares. Then she added more on August 28.
The pattern is clear. Wood sees Nvidia as the backbone of the AI revolution. She's buying every time the market flinches.
2. Cerebras Systems (CBRS) – The Underdog AI Chip Play
Nvidia gets the headlines. Cerebras gets the interesting story.
On August 25, ARK Invest bought 93,290 shares of Cerebras Systems across multiple ETFs. The purchase was worth roughly $17.2 million.
Cerebras does something different. Instead of building traditional chips, the company makes wafer-scale processors. The entire silicon wafer becomes one giant chip. It's a radically different approach to AI computing.
Wall Street sees upside. Analysts project Cerebras stock has 59% upside from current levels.
Wood has been adding to this position throughout August. On August 26, ARK bought another 69,585 shares of Cerebras.
Why the conviction?
Cerebras represents the challenger thesis. Nvidia dominates the AI chip market, but that doesn't mean there's no room for alternatives. Cerebras's wafer-scale approach offers advantages for certain AI workloads, particularly training massive models that require enormous parallel processing power.
The stock has struggled to regain its footing. That's exactly when Wood likes to buy. She's not chasing momentum. She's betting on the long-term trajectory of AI infrastructure.
Cerebras is a smaller position than Nvidia. But it's a meaningful bet. Wood is diversifying her AI chip exposure beyond the obvious choice.
3. Broadcom (AVGO) – The Custom Silicon Play
Broadcom is the quiet winner in the AI chip race.
On August 26, ARK Invest purchased 57,705 shares of Broadcom worth roughly $20.5 million. The firm added another 55,131 shares on August 28, bringing the total to about $20.5 million across three ETFs.
Broadcom isn't competing directly with Nvidia. It's playing a different game.
Broadcom makes custom silicon, application-specific integrated circuits (ASICs) designed for specific AI workloads. Hyperscalers like Google, Amazon, and Microsoft need custom chips for their unique infrastructure requirements. Broadcom designs and builds them.
The stock had dropped 27% heading into its September earnings report. Wood bought the dip. Aggressively.
Broadcom's Q3 earnings were approaching when Wood made these purchases. She wasn't waiting for the results. She was placing her bet before the numbers came out.
The thesis is simple. AI infrastructure spending is going to explode. Hyperscalers are going to need custom silicon. Broadcom is the best-positioned company to provide it.
Wall Street agrees. Analysts point to Broadcom's AI-powered growth trajectory as a key driver of their bullish view.
Wood is betting that the AI spending cycle will continue flowing toward computing capacity, networking, and infrastructure. Broadcom sits at the intersection of all three.
The Rotation Strategy – What Wood Is Selling
Here's where it gets interesting.
Wood isn't just buying. She's selling. And what she's selling tells you everything about her strategy.
On August 28, the same day she bought Nvidia, ARK sold $74.5 million of AMD stock. The same day.
AMD is a chipmaker. But it's not winning the AI race the way Nvidia is. Wood is consolidating her AI chip exposure into the companies she believes will dominate.
She also sold Palantir. On August 31, ARK sold 139,456 Palantir shares worth roughly $26 million. Earlier in August, she sold $27.2 million of Palantir to fund a SpaceX purchase. Those sales came after Palantir's sharp rally to record highs.
Palantir is an AI software company. But it's trading at more than 150 times earnings. Wood took profits.
She also sold Deere. In mid-August, ARK disposed of approximately $60 million of Deere stock. The pattern has continued, with smaller sales over several weeks.
The message is clear. Wood is rotating out of traditional industrials and overvalued AI software plays. She's rotating into AI infrastructure, the chips and computing power that will enable the next decade of AI growth.
The Deere sale is worth watching. If industrial exposure keeps falling while AI purchases expand, it reinforces that Wood is making an increasingly concentrated bet on the durability of the AI capital cycle.
What This Means for Investors
Should you follow Cathie Wood into these stocks?
That depends on your timeline.
Wood is a long-term investor. She's betting on a ten-year trend. Her funds have trailed the market this year, but she's not adjusting her strategy based on short-term performance.
If you're a long-term investor who believes in the AI infrastructure buildout, Wood's moves are worth studying. She's putting capital where she sees the most upside.
But there are risks.
Valuations across parts of the technology market remain elevated. Nvidia is priced for perfection. Any slowdown in data-center budgets could quickly challenge premium valuations.
Competition is real. Google's TPUs and Amazon's Trainium chips are gaining share. The AI chip market won't be a monopoly forever.
And Wood isn't always right. ARKK is up just 11.8% this year while the S&P 500 is at 13%. Her funds have underperformed.
But Wood's purchases shouldn't automatically be read as a signal that these stocks are cheap. Instead, investors should watch whether their earnings can justify ARK's rising exposure.
For Nvidia and Broadcom, hyperscaler capital spending and AI infrastructure financing remain critical. For Cerebras, the key is whether its wafer-scale approach can gain meaningful market share against Nvidia's dominance.
The bottom line? Wood is making a bet. A big one. She believes the AI infrastructure buildout is just getting started. She's putting her money where her mouth is.
You don't have to agree with her. But you should understand what she's doing and why.
Cathie Wood just went bargain hunting.
She bought Nvidia, Cerebras Systems, and Broadcom. Three AI chip companies. Three very different bets on the same thesis.
Nvidia is the safe choice, the market leader with dominant market share and explosive growth.
Cerebras is the contrarian play, the underdog with a radically different approach to AI computing.
Broadcom is the infrastructure bet, the company enabling the custom silicon that hyperscalers need.
Wood funded these purchases by selling Palantir, AMD, and Deere. She's rotating out of overvalued AI software and traditional industrials. She's rotating into AI infrastructure.
The ARK Innovation ETF might be trailing the market this year. But Wood isn't trading for the next quarter. She's trading for the next decade.
AI spending is projected to triple from $500 billion to $1.5 trillion by 2030. That's the opportunity Wood sees. That's the bet she's making.
She could be right. She could be wrong. But one thing is certain, Cathie Wood isn't sitting on the sidelines.