Dan Ives is back in front of a CNBC camera, making the same argument he has made for most of this year: that investors continue to underrate the scale and scope of what NVIDIA is doing. The pitch matters this week because NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) reports fiscal second-quarter results after the market close on Wednesday, August 26, 2026.
Ives, now at Yorkville Ives, told CNBC that his skepticism runs in the other direction from the crowd. He believes analyst models for the next year and a half sit meaningfully below what the company will actually deliver.
Shares closed at $216.85, up 23.79% over the past year and 16.41% year to date, which is why a bullish analyst call must identify something the market has genuinely missed.
Ives centered his argument on two things: Asia and hyperscaler exposure, which have near-term evidence, and physical AI, which does not. Both deserve consideration, although not equally.
Where Ives Sees Underestimation #
The scale claim has real evidence. NVIDIA’s last quarter produced $75 billion in data center revenue, up 92% year over year, with sovereign revenue up more than 80% year over year and infrastructure deployed across nearly 40 countries.
Hyperscaler activity matches this pace. Colette Kress told analysts that AWS plans to deploy more than 1 million Blackwell and Rubin GPUs starting this year, and Microsoft’s Fairwater data center is live ahead of schedule, powered by hundreds of thousands of Blackwell GPUs.
Prediction markets are already pricing this in. Polymarket traders assign a 0.966 probability that NVIDIA beats consensus non-GAAP EPS next week, and a 0.845 probability of a new all-time high by year-end.
Analyst consensus is similarly skewed. Forty-eight analysts rate the stock a Buy, ten Strong Buy, two Hold, and one Sell, with an average target of $302.83. Calling something underestimated when the sell side is broadly bullish requires unpacking what Ives actually means.
Physical AI as a Duration Bet #
Ives said the piece the market gives NVIDIA the least credit for is physical AI. “NVIDIA right now, you’re almost giving them minimal credit for physical AI. And what’s really going to be the next evolution of AI.”
He is directionally right that the market underweights this line today. NVIDIA disclosed that physical AI revenue exceeded $9 billion over the last 12 months, small relative to data center but growing.
The strongest version of the argument is that physical AI changes the addressable market once inference moves out of the data center and into machines operating in the world. That is a different demand curve than training clusters and plausibly extends the growth story past the current hyperscaler capex cycle.
The weakness is timing. Robotics and autonomous systems have taken longer to commercialize than advocates projected in every prior cycle, and paying today for revenue arriving in 2030 is a duration bet rather than an earnings bet.
15% to 20% Framing and What Would Confirm It #
Ives put a number on it. “I think it really comes down to it’s them giving expectations that numbers would be underestimated by 15 to 20% for the next 12 to 18 months.”
That framing deserves skepticism. A company beating consensus by that margin for six consecutive quarters would say as much about analyst conservatism as about the business, and analyst conservatism differs from durable earnings power.
Last quarter’s Q2 guide of $91 billion, plus or minus 2%, explicitly excluded China data center compute revenue, so any restart of H200 shipments to China would land as pure upside to the current quarter rather than as validation of a re-rating.
What the report on Wednesday needs to show for the Ives thesis to hold is a Q3 guide that clearly steps up on Blackwell and Rubin, confirmation that Vera Rubin production shipments begin in Q3, and enough color on sovereign and ACIE demand to support the $1 trillion in Blackwell and Rubin revenue visibility through calendar 2027 that management staked out in the Q1 release.
Contact [email protected] for any questions or corrections.