Less than three weeks ago, the most important data release of the first quarter occurred -- and you might have missed it. Amid a flurry of earnings reports and economic data releases, Feb. 14 marked the deadline for institutional investors with at least $100 million in assets under management (AUM) to file Form 13F with the Securities and Exchange Commission.

A 13F provides investors with an over-the-shoulder look of which stocks Wall Street's top money managers have been buying and selling. These snapshots can be particularly telling when it comes to next-big-thing trends, such as artificial intelligence (AI).

Image source: Getty Images.

One billionaire asset manager who hasn't been shy about purchasing and dumping AI stocks is Duquesne Family Office's Stanley Druckenmiller. Druckenmiller's fund closed out 2024 with more than $3.7 billion in AUM spread across 78 holdings. It's also an active fund, with an average holding time of less than seven months across all 78 positions.

This propensity for turnover was readily apparent last year, with Druckenmiller sending shares of AI leaders Nvidia (NVDA 1.69%) and Palantir Technologies (PLTR 1.18%) to the chopping block. But while Duquesne Family Office's chief was dumping these highfliers, he was building a stake in what's become his new favorite artificial intelligence stock.

Druckenmiller bids adieu to Nvidia and notably slashes Duquesne's stake in Palantir

When June 2023 came to a close, Druckenmiller's stake in graphics processing unit goliath Nvidia peaked at a split-adjusted 9,500,750 shares. This "split-adjusted" aspect has to do with Nvidia completing a historic 10-for-1 forward split in June 2024. In the 12-month stretch ending June 30, 2024, Druckenmiller dumped this entire position.

Meanwhile, Druckenmiller's fund held 769,965 shares of Palantir at the end of March 2024. As of Dec. 31, just 41,710 shares remained, which means 95% of this position had been sold.

While both companies are industry leaders with well-defined competitive advantages, there are a couple of catalysts that explain this aggressive selling activity.

The first and most-obvious is that Druckenmiller is simply cashing in his chips. Nvidia and Palantir have both soared in value, and Duquesne is a relatively active hedge fund. The worry is that Druckenmiller was a seller for more than just benign reasons.

Druckenmiller's actions might also reflect concern about an AI bubble brewing. Since (and including) the advent of the internet three decades, there hasn't been a next-big-thing innovation that's avoided a bubble-bursting event. All new technologies and innovations need adequate time to mature, and we're not at a point where businesses have a good handle on how to optimize AI solutions or even generate a positive return on their AI investments.

In a CNBC interview in May 2024, Druckenmiller compared the early days of the internet to the current AI revolution. Specifically, he pointed to the internet being bigger 20 years later than anyone thought it would be in 1999, but it took time for this technology to evolve. Said Druckenmiller, "AI might be a little over-hyped now, but under-hyped long-term."

The other issue for Nvidia and Palantir is their historically pricey valuations. Nvidia tipped the scales at a price-to-sales (P/S) ratio of 42 last summer, while Palantir's recent run-up to $125 per share put its P/S ratio at a peak of 99!

Companies that were on the leading edge of the dot-com bubble saw their P/S ratios peak in the 30 to 40 range before losing 80% to 90% of their value. While it's unlikely Nvidia and Palantir would suffer losses this great, the historic writing is on the wall that valuation premiums of this magnitude are never warranted or sustainable.

Image source: Amazon.

Billionaire Stanley Druckenmiller has a new favorite AI stock

While Duquesne's CEO has been busy slashing his fund's stakes in Nvidia and Palantir, he's been piling into his new favorite artificial intelligence stock, Amazon (AMZN -0.59%). Duquesne Family Office's latest 13F shows that 328,400 shares of Amazon were purchased in the December-ended quarter, making it the fund's 11th-largest holding by market value -- and top AI stock.

Most people become acquainted with Amazon by visiting its world-leading online marketplace. According to estimates from eMarketer, Amazon is expected to account for 40.9% of U.S. online retail sales this year.

Although e-commerce accounts for a substantial percentage of Amazon's net sales, this is a reasonably low-margin operating segment that doesn't play a key role in cash-flow generation or net income. The segment that makes Amazon tick -- and likely encouraged Stanley Druckenmiller to be a buyer -- is Amazon Web Services (AWS).

AWS is the world's top cloud infrastructure services provider, with a 33% share of cloud-service spending during the fourth quarter, based on estimates from Canalys. Companies are still very early in their cloud spending ramp, and the incorporation of AI solutions, including generative AI, into AWS should encourage businesses to boost their spending.

As of the end of 2024, AWS was pacing more than $115 billion in annual run-rate revenue, with its growth rate on a constant-currency basis reaccelerating to nearly 20%. Even though AWS accounted for less than 17% of the company's net sales last year, it was responsible for 58% of Amazon's operating income. The juicy margins associated with AWS should allow Amazon's cash flow growth to handily outpace its sales growth for the foreseeable future.

Druckenmiller may be looking beyond AWS, too. For instance, Amazon has nabbed the exclusive rights to stream Thursday Night Football, along with a package of NBA games. As its content library grows to include sporting events, so should its subscription (i.e., Prime) and ad-pricing power. Amazon's subscription and advertising services segments have respectively maintained a double-digit, constant-currency growth rate.

The final lure for Stanley Druckenmiller might just be Amazon's valuation. While it's not cheap by traditional fundamental standards, the company's price-to-cash-flow multiple is historically inexpensive. Throughout the 2010s, investors regularly purchased Amazon stock for a median multiple of 30 times year-end cash flow. Shares can be purchased right now for less than 13 times forecast cash flow in 2026.