This year has begun in much in the same way as the previous two: with the bulls firmly in charge. Last week, the widely-followed S&P 500 officially erased a 5% pullback and blasted its way to another record-closing high.
Catalysts haven’t been hard to come by on Wall Street, with better-than-expected corporate earnings, a tamer prevailing rate of inflation, and Donald Trump’s November victory -- all three major stock indexes soared during Trump’s first term in the White House -- lighting a fire under equities. But at the top of the pecking order in terms of catalysts has been the ascent of artificial intelligence (AI).
Image source: Getty Images.
In Sizing the Prize, the analysts at PwC pegged the addressable market for AI at $15.7 trillion by 2030. This stratospheric figure hasn’t been lost on investors or Wall Street’s leading money managers.
Although most billionaire asset managers have benefited from the rise of AI, the outlook for the perceived-to-be key players in the artificial intelligence arena is mixed. For instance, billionaire Stanley Druckenmiller of Duquesne Family Office, who oversees close to $3 billion in assets under management (as of Sept. 30), has been a decisive seller of two of Wall Street’s hottest AI stocks, and a clear buyer of two other foundational AI companies.
Nvidia and Palantir get the heave-ho from Druckenmiller
Even though most AI stocks have risen on the prospect of robust long-term growth prospects, few have captured the attention of investors quite like semiconductor giant Nvidia (NVDA -3.12%) and data-mining specialist Palantir Technologies (PLTR).
Nvidia has added more than $3.1 trillion in market value since the start of 2023. The company’s graphics processing units (GPUs), which include the ultra-popular Hopper (H100) chip and next-generation Blackwell GPU architecture, are the brains powering split-second decision-making and large language model training in high-compute data centers.
Meanwhile, Palantir’s stock has rallied more than 1,100% since 2023 began. Palantir’s AI-fueled Gotham platform, which gathers and analyzes copious amounts of data for federal governments and helps with military mission planning and execution, is its primary profit driver. More importantly, Gotham and Foundry (Palantir’s enterprise-focused segment) aren’t replicable at scale, which gives the company a relatively rock-solid moat.
Despite these well-defined competitive advantages, Druckenmiller’s fund has predominantly shown these two AI leaders to the door. In the six-month period between Mar. 31, 2024 and Sept. 30, 2024, Duquesne’s Form 13F filings show that all 175,943 shares of Nvidia stock were sold, while 95% of the fund’s shares in Palantir were sent packing.
Aside from simple profit-taking, the prevailing concern with market leaders like Nvidia and Palantir is the prospect of an AI bubble forming. There hasn’t been a next-big-thing investment trend in three decades that’s managed to avoid an early-innings bubble-bursting event. Though use cases exist for AI, most businesses lack a clear game plan to deploy this technology optimally and generate a positive return on their investment. AI will need time to mature, and that’s potentially bad news for Nvidia and Palantir Technologies.
Whereas Palantir is well-insulated from competitive pressures, this isn’t necessarily the case for Nvidia, which has competitors coming at it from all angles. What’s arguably most worrisome is that many of Nvidia’s largest customers by net sales are developing AI-GPUs to use in their data centers. Even if these chips lack the computing speed of Nvidia’s H100 and Blackwell, they’ll be notably cheaper and easier to access. In other words, it’ll minimize the AI-GPU scarcity that’s powered Nvidia’s margins and premier pricing power.
Lastly, Nvidia’s and Palantir’s valuations are concerning. Historically, companies on the leading edge of next-big-thing innovations have topped out around 40 times trailing-12-month (TTM) sales. Nvidia peaked at north of 42 times TTM sales last summer, while Palantir is currently valued at a staggering 72 times TTM sales. This level of valuation premium isn’t sustainable.
Image source: Getty Images.
Stanley Druckenmiller is piling into two other vital AI stocks
But just because Duquesne Family Office’s chief has been dumping shares of Nvidia and Palantir over six months, it doesn’t mean he hasn’t found perceived value in the AI space. Based on 13F filings over this comparable six-month stretch, Druckenmiller has purchased 239,980 shares of AI networking solutions kingpin Broadcom (AVGO 1.84%), as well as 21,610 shares of intellectual-property (IP)-propelled semiconductor stock Arm Holdings (ARM -2.43%).
Although Broadcom doesn’t have a stranglehold on GPUs quite like Nvidia, it is a preferred networking solutions option for businesses running generative AI solutions and building/training large language models. When Broadcom introduced its Jericho3-AI fabric nearly two years ago, it gave businesses a way to connect up to 32,000 GPUs to maximize their computing potential and minimize tail latency. Reducing response times is critical to the success of AI software and systems.
Beyond networking solutions, Broadcom CEO Hock Tan anticipates that custom application-specific integrated circuit orders will yield $60 billion to $90 billion in aggregate AI revenue from the company’s three leading hyperscale customers over the next three years.
The other important attribute to remember about Broadcom is that it’s more than just an AI company. Even though AI is its unquestioned top growth driver, it’s also a top provider of wireless chips and accessories for smartphones, a seller of cybersecurity software, and a leading manufacturer of optical components. If an AI bubble forms and bursts, Broadcom would offer some semblance of downside protection because of its diversified operations.
What makes Arm such an attractive stock is its unique positioning within the AI revolution. Instead of making the hardware that’s powering AI data centers, Arm is responsible for the IP that companies like Nvidia use to design their GPUs. The more AI proliferates, the more potential for Arm to collect royalties and licensing fees.
Where Arm Holdings can be especially helpful is the evolution of AI chips going into data centers and smartphones. While improving computing speed is always important, Arm’s value is in making GPUs and central processing units (CPUs) more energy efficient.
Similar to Broadcom, Arm benefits from its diversified operations. Even with AI being the company’s leading growth driver, it’s generating royalties and licensing revenue from CPUs, IP systems, and physical IP, as well. In the event that an AI bubble forms and bursts, Arm’s relatively minimal overhead expenses and juicy margins may somewhat buoy its stock.
Perhaps the one worry with Arm Holdings is its valuation. In spite of sustained sales growth of more than 20% per year and gross margin consistently in the mid-90% range, a forward price-to-earnings (P/E) ratio of 80 is a tough pill to swallow.
