This article first appeared on GuruFocus.

Nvidia's (NASDAQ:NVDA) AI server systems by roughly 15% is not a demand warning, according to Dan Ives. He sees it as another sign that AI infrastructure remains severely supply constrained, with hyperscalers and enterprises still willing to absorb higher costs rather than slow spending.

Demand and supply for chips is upwards of 15 to one, Ives told Bloomberg Television, arguing that the current memory shortage is forcing Nvidia and others to pass rising component costs through the supply chain.

Warning! GuruFocus has detected 5 Warning Signs with INTC.

Is GOOGL fairly valued? Test your thesis with our free DCF calculator.

The pressure is coming largely from memory, where Ives sees a continuing memory super cycle. He does not expect the industry to reach meaningful supply-demand equilibrium until the middle or latter part of 2028.

That makes the reported Nvidia price increase particularly important ahead of earnings. Microsoft (NASDAQ:MSFT), Alphabet (NASDAQ:GOOGL) and Oracle (NYSE:ORCL) are reportedly among customers facing higher data-center system costs as memory prices rise.

Yet Ives sees little evidence that those costs are causing customers to retreat.

You cannot take the foot off the gas right now because of the demand cycle, he said.

Instead, he expects hyperscalers to continue raising capital and increasing capital expenditures. Ives described the AI buildout as still being in the third inning, with enterprise adoption accelerating rather than slowing.

That matters for Nvidia because the stock's next leg increasingly depends on earnings growth rather than simple multiple expansion. Shares have risen about 15% this year after gains of 38% last year and triple-digit returns in prior years, making investors more demanding about future upside.

Nvidia's earnings this week become the immediate test of Ives' thesis. Investors should watch data-center growth, gross margins, Vera Rubin demand and management's commentary on memory costs.

If Nvidia can raise system prices without slowing orders, the move would reinforce its pricing power and suggest hyperscaler AI budgets remain resilient.

The risk is that higher infrastructure costs eventually force customers toward internal chips, cheaper models or slower deployments. For now, however, Ives' conclusion is clear: From a demand perspective, it's accelerated, not decelerated.