Nvidia Q4 earnings test limits of AI boom as growth expectations soar

Nvidia was preparing to report its fourth-quarter results on a Wednesday evening in late February, and Wall Street had arranged itself into a posture of almost religious anticipation. The company had become the indispensable machinery of the artificial intelligence boom—the maker of the expensive graphics processors that every major tech company needed to build and train the large language models that had captured the world's attention. Since the end of 2022, Nvidia's stock had climbed nearly fivefold. The company's market value had swollen to $1.72 trillion, briefly eclipsing Amazon and Alphabet in total worth.

The numbers analysts were expecting told the story of a company riding something extraordinary. Revenue was forecast to jump 240 percent from the same quarter a year earlier, landing at $20.6 billion. Of that, $17.06 billion would come from data center sales—the division that sold the H100 chips and other processors to companies building AI systems. Net income was projected to surge more than sevenfold to $10.5 billion. For the current quarter ahead, Wall Street was already penciling in another 208 percent growth rate, pushing sales to roughly $22.17 billion. These were not the numbers of a mature business. These were the numbers of a company that had found itself at the center of a technological transformation.

But beneath the euphoria lay a question that no amount of record revenue could quite answer: How long could this actually last? Nvidia's customers were not scattered startups or academic institutions. They were Microsoft, Amazon, Meta, and Google—the companies with the deepest pockets and the most to gain from AI dominance. In their own recent earnings reports, these giants had signaled they would keep buying. Yet they had also used a particular phrase that made some analysts pause. They described their GPU purchases as "flexible" and "demand driven," language that suggested they could dial back spending if the moment demanded it. One analyst at D.A. Davidson noted that while the current hype cycle still seemed robust, there were already "possible early signs" that the long-term picture might not sustain these growth rates indefinitely.

Jensen Huang, Nvidia's chief executive, would need to address this directly when he spoke to investors after the earnings announcement. What was the company's own view of how long these stratospheric growth rates could continue? The question mattered because it cut to the heart of whether the AI boom was a genuine, multi-year shift in how technology companies would operate, or whether it was a more compressed cycle of investment that would eventually plateau.

There was also the matter of Nvidia's next-generation chip. The company was planning to begin shipping a new highest-end server GPU called the B100 in 2024. The timing of that transition could reshape the company's growth trajectory in ways that were difficult to predict. Would customers rush to upgrade, creating a new surge in demand? Or would the arrival of a more advanced chip cause some to pause their current purchasing, waiting for the newer technology to become available?

Nvidia's other businesses—gaming chips for personal computers, automotive processors—existed in the background of this conversation. But they barely registered. Data center GPUs accounted for more than 80 percent of the company's sales. For analysts watching the earnings call, there was really only one metric that would matter: the data center revenue number, and whatever Huang had to say about how broadly AI adoption would spread across the industry in the months ahead. Everything else was noise.