Why We Sold NVIDIA & Palantir
NVIDIA and Palantir are two tremendous technology companies, with enviable competitive advantages.
No company designs and develops high-performance GPUs and AI computing platforms better than NVIDIA, and if there’s a company that develops data integration and analytics software better than Palantir, we don’t know what it is.
Many asset managers trimmed their positions in NVDA beginning in early 2024 and following the release of DeepSeek’s R1 AI model in late January 2025, which apparently cost only $6 million to train.
DeepSeek's innovative engineering techniques implied to some market participants that future AI models may depend less on premium chips than previously expected, and NVDA sold off sharply on the news.
Two weeks before this news came out, we had already begun to trim NVDA shares.
Our thesis was not about the threat posed by DeepSeek, though that is a consideration today. It was more about our estimation that the AI infrastructure cycle is nearing its peak, which we think could challenge NVDA’s ability to deliver the extraordinary growth rates investors had become accustomed to.
As mentioned, PLTR is a market leader in the world of big data analysis for enterprises and government. There is really no company in the marketplace that can better help organizations—both in government and the private sector—integrate, analyze, and visualize large and complex datasets to make informed decisions.
It’s fair to assume that the stock price should increase as a reflection of the business’s strengths and competitive advantages. How much it should go up is a different story.
When we sold PLTR, it was trading at 128x and 40x 2025 earnings-per-share (EPS) and revenue, respectively. From a risk management perspective, this valuation simply strayed too far from the premium we were willing to pay for PLTR’s future earnings. As we write—which follows PLTR’s roughly -15% decline in two days, the stock still trades at 195x and around 156x 2025 and 2026 EPS, and roughly 62x and 49x revenue. These are significantly high multiples, which even with very bullish growth assumptions, we think indicate that the stock may be in nosebleed territory.
Finally, with closer analysis, we also determined that PLTR had an excessive amount of retail backing, i.e., popular with individual investors and on discussion forums. This level of exposure raises the likelihood of volatile swings in the price, in our view, which can be driven more by shifts in sentiment than underlying earnings and fundamentals.
The Upshot:
With cash proceeds from these and other sales, we were able to position into new companies and industries we’re excited about in the new year, which we’ll address in next month’s letter.