David Bahnsen, Chief Investment Officer at The Bahnsen Group, appeared on CNBC’s Squawk Box this morning with a specific warning. His tell is more clinical. “I think a bubble gets ready to pop is when you start seeing companies have the good news everyone’s been talking about and cheerleading and then the stocks don’t respond,” he told CNBC.
The companies are still delivering. The buyers have stopped showing up at the price.
Price action in the AI bellwethers backs him up.
The Palantir and NVIDIA disconnect
Palantir (NASDAQ:PLTR | PLTR Price Prediction) just printed one of the cleanest quarters any large-cap software company has produced. Revenue grew 84.71% year over year to $1.63 billion, U.S. commercial revenue jumped 133%, and CEO Alex Karp announced a Rule of 40 score of 145%, a number he claimed only NVIDIA (NASDAQ:NVDA), Micron (NASDAQ:MU), and SK Hynix can match. Management raised the full-year guide to roughly 71% growth. You can read the press release on the SEC’s site.
The stock filed at $144.45 on May 4 and has traded down to $116.86. Year to date PLTR is down 36.47%. Bahnsen on CNBC: “Palantir just had the two greatest quarters it’s ever going to have and the stock is down 50%. Nvidia has not moved in nine months.”
NVIDIA reported Q1 FY27 revenue of $81.6 billion, up 85.2% year over year, with Data Center at $75.25 billion and a Q2 guide of $91.0 billion. Yet over the past nine months, from September 29, 2025 through June 26, 2026, NVDA is up only 6.01%. For the most important compute platform on earth posting free cash flow growth of 74.51%, that is something close to a stall.
Broadcom (NASDAQ:AVGO) follows the same pattern. AI semiconductor revenue grew 143% to $10.8 billion, with management guiding to $16.0 billion next quarter, up over 200% year over year. The stock filed at $495 on June 3, traded to $419.60 within an hour, and now sits near $373.90. Three companies, three blowouts, three weak responses.
The 50x-forever problem
Bahnsen separates the chip from the multiple. “You don’t have to be a doubter in AI. You don’t have to be a doubter in even some of these individual companies.” Rather, you “just have to doubt that it’s worth 50 times earnings forever and ever,” he said on CNBC. PLTR currently trades at a P/E near 159x. That sets a high bar for the next decade of execution.
His complaint about Wall Street’s S&P 500 targets near 7,800 applies the same logic at the index level. Those targets assume roughly 20x multiples and flawless earnings. “I think the forward earnings are not realistic. And I also think the assumptions of continued growth to rationalize even a lower multiple are very likely not realistic,” he told the show. The S&P is already up 8% year to date, much of that carried by the same names now stalling.
What Bahnsen actually predicts
His call diverges from the usual perma-bear script. “I think you will have a moment in which things reprice. We saw it with dot com and you had five companies survive and you had 99% go to the graveyard,” he said, then walked it back from a timing call. “I’m not talking about timing anything. I’m just simply saying the investor psychology always 100% of the time gets ahead of itself.”
His base case for the next couple of years is a grind. “What I would say is a better prediction is that you wake up in a couple of years and you have a very muted return that along the way there’s rallies but there’s sell offs.” Margin debt at record levels gives him pause, though the composition differs from the 1999 retail mania.
Strong fundamentals do not mechanically beat starting valuations. If you own these names, the question is whether the next leg of earnings growth has already been priced in. Retail is starting to wonder out loud.
Reddit data shows Palantir sentiment flipping from a bullish 72 in early June to a very bearish 18 by June 3, triggered partly by a viral Michael Burry post calling the stock “a sand castle supported only by AI applications narrative.” Bahnsen would put it less colorfully. The good news stopped working, and that is the data point worth tracking.