You might not realize it, but arguably the most important data dump of the first quarter occurred on Friday, Feb. 14.
While some folks were wisely focused on making their significant other feel special for Valentine’s Day, institutional investors with at least $100 million in assets under management (AUM) were filing Form 13F with the Securities and Exchange Commission. A 13F provides a concise snapshot of which stocks Wall Street’s smartest money managers bought and sold in the latest quarter.
Though 13Fs have their drawbacks — since they’re filed up to 45 calendar days after a quarter ends, the data might be stale for an active hedge fund — they can still clue investors into which stocks, industries, sectors, and trends are piquing the interest of Wall Street’s billionaire money managers.
Image source: Getty Images.
One billionaire investor Wall Street pays close attention to during 13F season is Philippe Laffont of Coatue Management. Laffont, who closed out 2024 overseeing almost $30 billion in AUM spread across 68 holdings, favors high-growth tech and healthcare stocks and has been a big investor in the rise of artificial intelligence (AI).
But what you might be surprised to learn is that Laffont has been an aggressive seller of the face of the AI revolution, semiconductor colossus Nvidia (NVDA 2.63%), while piling into another trillion-dollar stock that’s critical to AI infrastructure.
Philippe Laffont has sent nearly 39.8 million shares of Nvidia to the chopping block
At the end of March 2023, Coatue Management held a split-adjusted 49,802,020 shares of Nvidia stock. This “adjustment” factors in that Nvidia completed a historic 10-for-1 forward stock split in June 2024. Since March 31, 2023, Laffont has been a seller for seven consecutive quarters, to the tune of almost 39.8 million shares, or 80% of Coatue’s peak position in Nvidia.
One logical reason behind this persistent selling activity from billionaire Philippe Laffont is simple profit-taking. Nvidia’s stock has nearly 10X’d since 2023 began, which has provided plenty of opportunity for investors to ring the register.
Nvidia’s otherworldly returns have been fueled by the dominance of its AI-data center hardware. Its Hopper (H100) graphics processing unit (GPU) and next-generation Blackwell GPU architecture collectively hold the lion’s share of the GPU market in enterprise AI-accelerated data centers.
However, there may be more to Laffont’s continued selling of Nvidia stock than meets the eye.
For instance, competition coming at Nvidia from all angles is likely to have a deleterious effect on its pricing power and margins. While most investors are focused on direct competitors bringing their own GPUs to market, the potentially bigger threat is that many of Nvidia’s top customers, which are members of the “Magnificent Seven,” are internally developing AI chips to use in their data centers. Even though these chips are unlikely to offer comparable computing speed to Nvidia’s Hopper or Blackwell, they’ll be notably cheaper and more readily accessible.
Nvidia has relied on AI-GPU scarcity and the overwhelming demand for its hardware to meaningfully increase the price of its Hopper and Blackwell chips and substantially boost its gross margin. But as its top customers fill their data centers with internally developed AI chips, AI-GPU scarcity will wane. Ultimately, this is bad news for Nvidia’s pricing power and gross margin.
Laffont might also be relying on history as a cue to ring the register. Since the advent of the internet some three decades ago, every next-big-thing innovation has worked its way through an early expansion bubble. This is to say that investors always overestimate the initial adoption rate and utility of game-changing technologies.
The recent DeepSeek scare for Nvidia serves as a stark reminder that most businesses lack a clear understanding of how they’ll optimize their hardware or generate a positive return on their AI investments. It’s going to take years for this technology to mature, which effectively confirms that investors have, once again, set the bar too high for a next-big-thing technology.
If an AI bubble forms and bursts, Nvidia would be among the hardest-hit companies.
Image source: Getty Images.
Coatue Management is piling into this key AI infrastructure stock
On the other side of the aisle, the growth-focused Laffont can’t seem to get enough of world-leading chip fabrication company, and one of the newest members of the trillion-dollar club, Taiwan Semiconductor Manufacturing (TSM 1.03%). Since the start of 2023, Laffont has purchased 9,962,102 shares of Taiwan Semi, making it Coatue’s third-largest holding by market value.
Taiwan Semiconductor is relied on by many top-tier AI companies (including Nvidia) to fabricate the hardware that makes AI-data centers tick. In particular, the company’s chip-on-wafer-on-substrate (CoWoS) capacity is of the utmost importance. CoWoS is a packaging technology for high-bandwidth memory, which is a necessity for high-performance computing and AI solutions. Taiwan Semi is aiming to have monthly CoWoS capacity of 135,000 units in 2026, up from just 35,000 units in 2024.
Similar to Nvidia, Taiwan Semi is benefiting from the incredible backlog of orders that’s accompanied the AI revolution. This has aided its pricing power and is providing transparency to its operating cash flow that few market-leading businesses can match.
It’s important to recognize that Taiwan Semiconductor Manufacturing is about more than just AI infrastructure. Although AI is its primary growth driver, Taiwan Semi also provides traditional central processing units and custom chips for some of the world’s largest companies, such as Apple. This operating diversity should allow Taiwan Semi to better navigate an AI bubble-bursting event, should one occur.
Despite these catalysts, Taiwan Semi may find it challenging to remain a member of the trillion-dollar club. For example, the return of Donald Trump to the White House complicates trade matters and its pricing power. Since 80% to 90% of the company’s production capacity is located in Taiwan, the potential for import tariffs and/or a trade war taking shape looms large.
Taiwan Semiconductor Manufacturing isn’t the fundamental bargain it once was, either. Though its trailing-12-month (TTM) price-to-earnings (P/E) ratio of 30 isn’t egregiously high for a company that’s critical to AI infrastructure, this does represent a 32% premium to its average TTM P/E ratio over the last five years.
Given the historic priciness of the stock market, companies that trade a premium are likely to take it on the proverbial chin when the next correction arrives.
Financial Market Newsflash
No financial news published today. Check back later.
