EXCLUSIVE: Watch These 3 AI Bottlenecks Ahead of Nvidia Earnings, Yorkville CEO Says

EXCLUSIVE: Watch These 3 AI Bottlenecks Ahead of Nvidia Earnings, Yorkville CEO Says

Daragh Thomas

Tue, August 25, 2026 at 4:41 PM GMT+3 3 min read

• NVDA +2.33%

Benzinga and Yahoo Finance LLC may earn commission or revenue on some items through the links below.

Nvidia Corp. (NASDAQ:NVDA) reports earnings Wednesday in what may be the clearest test yet of whether Big Tech's enormous AI infrastructure buildout is still accelerating. The largest technology companies are set to spend more than $700 billion on capex this year.

But Steve Neamtz, CEO of Yorkville America, says some of the most important constraints are emerging beyond the chips themselves.

"The $700 billion number is real, and it matters, but I think the more interesting question is where that money actually has to go," Neamtz told Benzinga.

AI Spending Hits Physical Constraints

Neamtz points to three bottlenecks further down the supply chain. The most immediate is memory, storage and the power needed to run it all. "Compute has scaled faster than that layer can keep up," he said.

That concern already looks timely. Some of Nvidia's customers have reportedly been told that prices for some AI servers will rise more than 15%, largely because of soaring memory costs.

The second is critical minerals. By 2030, AI data centers could require the equivalent of about 3% of today's global rare-earth demand, according to estimates cited by Sprott.

The third is labor. The construction industry needs an estimated 349,000 additional U.S. workers in 2026, according to Associated Builders and Contractors, a figure covering the whole industry rather than data centers alone.

"It really isn't a capital problem — it's a physical one," Neamtz said.

The scale of the spending isn't what worries investors, in his view. What matters is whether management can point to a genuine constraint and explain what the capex unlocks. Spending that looks reactive is more likely to be treated as a drain on cash.

"It's less about the size of the number and more about whether there's a coherent story behind it," he said.

Nvidia Gives Investors the Next Read

Wall Street expects Nvidia to report revenue of about $92.2 billion Wednesday, nearly double last year's figure, and Reuters describes the results as a test of the AI spending boom's sustainability.

Prediction market traders expect the physical buildout to keep growing. Kalshi traders currently see a 66% chance the U.S. ends 2026 with at least 5,300 data centers, a 52% chance of at least 5,400 and a 46% chance of reaching 5,500. That compares with 4,313 U.S. data centers as of June 1, 2026, according to Kalshi.

Neamtz will be listening for signs that demand is still running ahead of supply across the AI infrastructure chain. He'll watch whether companies describe lead times and capacity shortages as easing or tightening.

"If it's still tightening, that tells you the buildout hasn't hit its ceiling yet," he said.

If supply begins catching up with demand, Neamtz sees the investment opportunity entering a different phase.

"The opportunity shifts from infrastructure scarcity to whoever can actually put that capacity to work most efficiently," he said.

Image: Shutterstock

Kalshi and Benzinga have an existing data collaboration agreement.

Read Also:Anthropic Asks Recruits How They'd Feel if Its Stock Hit Zero. Traders See 68% Chance of $2 Trillion This Year

This article EXCLUSIVE: Watch These 3 AI Bottlenecks Ahead of Nvidia Earnings, Yorkville CEO Says originally appeared on Benzinga.com

© 2026 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.