Some of Wall Street’s smartest billionaire fund managers have sent shares of Nvidia packing in favor of a business that’s critical to enterprise AI data centers.
Data is Wall Street’s currency, and there’s an abundance of it to go around. Between earnings season — the six-week stretch each quarter where most S&P 500 companies report their operating results — and economic data releases, there’s an almost overwhelming amount of information for investors to absorb. Occasionally, something important can fall through the cracks.
For example, May 15 marked the deadline for institutional investors overseeing at least $100 million in assets under management to file Form 13F with the Securities and Exchange Commission — and you might have missed it.
This filing, which is due no later than 45 calendar days following the end to a quarter, details which stocks, exchange-traded funds (ETFs), and (select) stock options Wall Street’s brightest money managers purchased and sold in the latest quarter. It can offer big-time clues as to which influential stocks are garnering interest or falling out of favor.
While Berkshire Hathaway’s Warren Buffett is the stock market’s most-followed money manager, he’s far from the only billionaire investor known to move markets.
Image source: Getty Images.
For instance, Duquesne Family Office’s Stanley Druckenmiller and Lone Pine Capital’s Stephen Mandel have exemplary investment track records of their own, along with billions of dollars in assets under management.
What’s particularly noteworthy about the first-quarter 13Fs from Druckenmiller’s and Mandel’s respective funds has been their approach to the artificial intelligence (AI) revolution. Both billionaires have dumped the preeminent AI stock on Wall Street in favor of a company that’s critical to enterprise AI data centers.
Billionaires Druckenmiller and Mandel have completely dumped their Nvidia stock
With the analysts at PwC pegging the addressable market for artificial intelligence at $15.7 trillion by 2030, there’s room for hundreds of businesses to get their piece of the pie. However, no company has been a more direct beneficiary of the evolution of AI than Nvidia (NVDA 1.69%). It’s also the stock billionaires Stanley Druckenmiller and Stephen Mandel sent packing.
Accounting for Nvidia’s 10-for-1 forward split in June 2024, Duquesne Family Office held 9,500,750 shares in the June-ended quarter of 2023. Meanwhile, Lone Pine Capital possessed 6,416,490 shares of Nvidia, also at the end of June 2023.
But over the following 15 months for Druckenmiller and 12 months for Mandel, both billionaires would oversee the complete purge of their respective fund’s Nvidia holdings.
While there’s no denying that Nvidia’s Hopper (H100) graphics processing unit (GPU) and Blackwell GPU architecture are the preferred options in AI-accelerated data centers, and it’s pretty clear that no other external competitors are particularly close to challenging Nvidia’s hardware in terms of compute abilities, there are still viable reasons for Druckenmiller and Mandel to have cashed in their chips.
One obvious reason to sell is simple profit-taking. Nvidia stock skyrocketed from early 2023 into late 2024, which increased its valuation by more than $3 trillion. We’ve never witnessed a megacap business add $3 trillion in market cap so quickly before, which may have encouraged these two billionaires to lock in their gains.
But there might be more than just profit-taking behind this selling activity.
For instance, it’s only logical to expect competitive pressures to mount in the hardware arena. Even though Hopper and Blackwell hold most of the AI-GPU market share in high-compute data centers, external competitors are ramping up production of existing chips and bringing more energy-efficient hardware to market.
What’s more, many of Nvidia’s top customers by net sales are developing AI-GPUs and AI solutions of their own. Even though these chips aren’t going to be as fast as the Hopper or Blackwell, they’re expected to be considerably cheaper and they won’t be backlogged like Nvidia’s hardware. This is a direct threat to the AI-GPU scarcity that’s afforded Nvidia superior pricing power for its GPUs.
History isn’t exactly in Nvidia’s corner, either. Despite AI supporting a lofty addressable market, every next-big-thing technology and innovation for more than three decades has endured an early stage bubble-bursting event. In short, investors have a historically strong tendency to overestimate how quickly a new innovation will gain utility and be adopted on a mainstream basis.
With artificial intelligence likely needing time to mature as a technology, it’s the most-direct beneficiary, Nvidia, which could feel the pain if a bubble forms and bursts.
Image source: Getty Images.
This is the new AI apple of Druckenmiller’s and Mandel’s eye
While Duquesne’s and Lone Pine’s billionaire chiefs pared down the number of stocks they’re holding amid a volatile first quarter, there’s one artificial intelligence stock both have been buying — and it plays a vital role in the expansion of AI-accelerated data centers.
The new AI apple of Druckenmiller’s and Mandel’s eye is none other than leading chip fabrication company Taiwan Semiconductor Manufacturing (TSM 0.80%), which is more commonly known as “TSMC.” Duquesne more than quintupled its existing stake by adding 491,265 shares of TSMC during the March-ended quarter, while Lone Pine’s 13F shows that 104,937 shares of TSMC were purchased in the first quarter of 2025.
Most AI-GPU companies rely on Taiwan Semi’s fabrication services, including industry leader Nvidia and key rival Advanced Micro Devices. TSMC is in the process of rapidly expanding its monthly chip-on-wafer-on-substrate (CoWoS) capacity from approximately 35,000 units in 2024 to an estimated 135,000 units monthly by 2026. CoWoS is a technology used to package high-bandwidth memory, which is necessary for high-compute data centers where software and systems are making split-second decisions.
With demand for AI-GPUs overwhelming their supply over the last two years, TSMC has enjoyed a significant backlog for its chip fabrication services and has seen more its net sales skew toward high-performance computing, which can yield higher margins for the company. On a year-over-year basis, TSMC’s net sales from high-performance computing surged from 46% to 59%, as of the March-ended quarter.
Although the possible bursting of an AI bubble would be a concern for Taiwan Semiconductor Manufacturing, the company’s order backlog and revenue diversification offers some semblance of protection.
For instance, 28% of net sales in the first quarter derived from advanced chips used in smartphones. Apple prominently leans on TSMC for the chips used in the iPhone. The great thing about smartphones and wireless service access is they’ve both evolved into basic necessities for most Americans. Even though demand for smartphone chips isn’t growing as quickly as it once did, the cash flow from this segment tends to be highly predictable for TSMC.
Taiwan Semi has a long runway of opportunity in Internet of Things and automotive, as well. As homes and vehicles become more technology-dependent, companies like TSMC will be relied on to manufacture these advanced chips.
Lastly, Druckenmiller and Mandel may have been encouraged by the dip in Taiwan Semiconductor’s stock in the first quarter. Though TSMC stock isn’t (currently) historically inexpensive, its shares did drop to a forward price-to-earnings ratio of nearly 15 during tail-end of the March quarter. This makes for an attractive multiple, when compared to Nvidia.
GIPHY App Key not set. Please check settings