A handful of big tech companies operate a large, global network of massive AI data centers.
Those companies, chief among them Meta, Amazon, Google, Microsoft, and Oracle, are often referred to as hyperscalers. As the top AI chipmaker, Nvidia provides the hardware for these hyperscalers, and in turn, the hyperscalers are the chip giant’s biggest customers.
During the rise of the AI hype era, any investment made into AI infrastructure by these hyperscalers was met with market fervor, and any good news for these hyperscalers meant good news for Nvidia.
But recently that dynamic has started to change. The hyperscalers’ financial commitments for the year topped $725 billion, a figure that has doubled since a year ago. Meanwhile, this AI commitment is still showing limited returns, causing investors to grow wary of this record AI spending and wonder if it is warranted or signifies a bubble. Earlier this year, Evercore analysts warned that the investment commitments could turn these tech giants’ cash flow negative, a fate that, if it came to pass, would certainly hit Nvidia’s profits as well. It seems to be a major reason why some experts think Nvidia’s revenue growth may be peaking.
This pressure may have been the driver of a major change in the way Nvidia reports its most valuable financial metric, data center revenue. In an attempt to prove revenue diversification to investors, Nvidia announced on Wednesday that, going forward, it will break down data center revenue into two: hyperscalers and “ACIE,” which stands for AI Clouds, Industrial, and Enterprise, aka a catch-all category that encompasses pretty much everything else.
“It’s really about the fact that our business has now evolved and grown to such a large scale, it’s helpful to segment it, so that you have a better understanding of how our business works,” Nvidia CEO Jensen Huang said at the company’s earnings call on Wednesday evening.
On the call, Huang spent considerable time hammering in the point that Nvidia’s revenue is not as dependent on the hyperscalers as once thought, even though in the past quarter hyperscalers accounted for half of all data center revenue.
Huang assured investors that while hyperscaler growth came first and both categories will “grow incredibly fast,” ultimately the second category would “be larger over time.”
“Hyperscale developed AI first for a lot of reasons, you know, they have great computer science, they have excellent data center capability, and they also focus largely on consumer applications, which, if not perfect, is not the end of the world,” Huang said. “For many of the other applications…until the AI is very capable and does really productive work, and does it safely, and it could do it in a way that can actually generate impact and income, it doesn’t really get used.”
Nvidia could back up those growth promises with its results so far. Hyperscaler data center revenue grew 12% quarter over quarter, while the second category grew 31%.
Huang has spent the past few months trying to battle the negative investor outlook on the hyperscaler capex commitments to save Nvidia’s reputation, but this new reporting method signals a change in strategy. In the previous earnings report, he spent much of the call assuring investors that the spending commitments were warranted, offering an alternative way to measure revenue by correlating it directly with compute power instead.
“Compute equals revenue,” Huang proclaimed at that earnings call repeatedly.
But since then, his new mantra has done little to dissipate the fears, while the hyperscalers upped AI spending commitments even further in the latest round of big tech earnings in April. Now, it seems he’s hopping ship and focusing on proving the revenue growth that everyone else will surely bring to Nvidia instead.
Read the full article here
