Dan Loeb Called Nvidia Undervalued Then Sold Everything. Here's What Actually Happened

In June 2026, Dan Loeb told investors that Nvidia was "absolutely undervalued" despite its $5 trillion market cap. He compared it to Google in 2015 and Amazon in 2018 — companies that looked too big to grow, then broke out. He said hedge funds had made Nvidia a default short, and called that a "narrative trap".

Three months later, Loeb's hedge fund Third Point sold every single one of its 190,000 Nvidia shares. Along with Broadcom, KLA, Lam Research, and a semiconductor ETF, he wiped the entire semiconductor portion of his portfolio clean.

The headline is irresistible. Billionaire says Nvidia is cheap, then sells everything. But the actual story is much less dramatic — and much less useful as a signal — than the headline implies. What happened isn't that Loeb changed his mind about Nvidia. It's that you have to understand who he is and how his money moves before his public statements tell you anything about what to do with your own.

How Loeb actually invests

Third Point is not a buy-and-hold fund. Loeb manages roughly $30 billion across hedge fund, private credit, CLO, and insurance platforms. On the equity side, the average holding period for his top-20 stocks was just nine months as of June 30. His own words describe him as a "swing trader", not a multi-year compounder.

That holding period is the missing link between the undervalued comment and the full exit.

A nine-month average means Loeb's fund rotates through positions on a calendar that has nothing to do with the two-to-three year earnings horizon he was describing in that June interview. When he said Nvidia is undervalued, he was giving a long-conviction view. When his fund sold, it was executing a portfolio rebalance on his actual operating cycle. The two statements aren't contradictory if you understand that Loeb thinks in multiple time horizons and his fund acts on the shortest one.

The Q2 13F filing, released August 14 for positions as of June 30, shows the broader pattern. Third Point didn't just exit Nvidia — it exited the entire semiconductor trade. At the same time, Loeb built a $533 million position in Warner Bros. Discovery, increasing his Alphabet stake by 489%, and added positions in Block, Keysight, Norfolk Southern, and Capital One. This reads less like a bearish Nvidia call and more like a fund that had grown comfortable in its tech winners and rotated toward media, financials, and industrials.

Profit-taking after a long run is the most straightforward explanation. Nvidia rose roughly 15% from the start of the year through June. The stock had run much further over the prior two years. In a fund where the average position lasts nine months, selling at the end of a long winning streak is exactly what you'd expect.

What Nvidia actually did after Loeb left

The timing of the exit makes the contrast sharper. Third Point went flat on Nvidia by June 30. The stock was around $200.

Two and a half months later, on August 26, Nvidia reported fiscal Q2 earnings that would be the defining quarter of its year. Revenue hit $96.2 billion — well above the $92.3 billion Wall Street expected. Data center revenue alone was $89 billion, up 117% year over year. The company reported 75% gross margins, held across both GAAP and non-GAAP measures.

Then came the guidance, which is where Nvidia's operating momentum really showed through. Management guided Q3 revenue to $108 billion, plus or minus 2%. That's another 12% sequential increase — sustained acceleration at a scale that almost no other company on earth has achieved. Jensen Huang called it a golden age, saying AI had reached the point where "compute is revenue" and that the infrastructure build-out was at full steam.

The Vera Rubin platform — Nvidia's next-generation architecture after Blackwell — is now in full production, running at CoreWeave, Google Cloud, Microsoft Azure, Oracle, and Nebius. Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to mobilize over $500 billion of third-party capital for AI infrastructure.

The stock closed at $236 by mid-August, then pulled back. Today it sits around $227, roughly 13% higher than where Loeb's fund last held it.

None of this means Loeb was wrong to sell. He made money on the trade, and he doesn't owe Nvidia shareholders a position. But it does mean that his exit told you nothing about Nvidia's future and everything about his fund's rotation cycle.

The real question isn't what Loeb sold

The question that matters for anyone considering Nvidia now is separate from any hedge fund's quarterly rebalance. It's about the return curve.

Nvidia trades at a trailing P/E of 28 on earnings that grew 106% year over year in the last quarter. Free cash flow over the trailing twelve months is $127 billion, with a 47% free cash flow margin. Return on invested capital sits at 87%. These aren't reasonable ratios on a reasonable company — they're reasonable ratios on an extraordinary one that just printed more revenue in a single quarter than most countries collect in GDP.

The tension is between the scale of what Nvidia is delivering and the question of whether the acceleration can hold. Q3 guidance of $108 billion implies continued growth. If the company hits that run rate and sustains 74-75% margins through fiscal 2027, the trailing P/E compresses further even if the stock doesn't move. The market is pricing in continued strong growth, but the execution keeps beating that pricing.

The competitive risks are real — hyperscalers are building custom chips, China remains closed as a market, and the inference workloads that dominate long-term AI economics favor efficiency over the brute-force training advantage where Nvidia's CUDA moat is strongest. But none of these risks has touched the quarterly revenue or margins yet, and Nvidia's product architecture keeps moving forward faster than the alternatives can close the gap.

Loeb's exit is interesting as a case study in how to read hedge fund filings — loud public opinions don't always match quiet portfolio mechanics, and the holding period matters as much as the direction. But for Nvidia, the material facts are in Nvidia's own earnings and guidance, not in whether a $30 billion fund with a nine-month average holding period decided to rotate out.

The debate isn't about whether Nvidia stays important. It's about whether the return from here — after a year-to-date gain of roughly 22%, with $96 billion quarterly revenue already behind you — is still the most attractive deployment of capital in the AI trade. That's a question Loeb's sale doesn't help answer. Nvidia's next earnings report will.