Nvidia Corp. is once again setting the tone for the artificial intelligence industry, with the company’s earnings highlighting both the insatiable demand for graphics processing units and the growing concentration of power among a handful of hyperscaler cloud providers. Despite geopolitical pressures and emerging rivals, the company’s strong financial performance underscores its role in the development of modern AI.

Beyond the numbers, the story is about how infrastructure itself is being reshaped. Hyperscalers are pouring billions into AI factories, betting that massive GPU investments will drive efficiency gains and new competitive advantages. Enterprises are moving more slowly, but networking advances, hybrid data strategies and shifting vendor ecosystems suggest the industry is racing toward a different future for compute.

In an exclusive analysis on this week’s theCUBE Pod, theCUBE Research’s John Furrier (pictured, left), executive analyst, and Dave Vellante (right), chief analyst, dig into Nvidia earnings and what they reveal about broader market dynamics. From VMware Explore to Broadcom Inc.’s expanding role in data centers, and from enterprise adoption challenges to ethical concerns around AI deployment, their conversation highlights both the promise and the challenges shaping today’s technology landscape.

“Nvidia continues to dominate even with the shortfall with H20,” Furrier said. “No sales to China in Q2 and no sales expected in Q3, yet revenues are up. [The] Wall Street Journal and CNBC both reporting what we’ve been speculating on these big GPU sales is that Nvidia has literally a handful of customers that count for almost 40% of their sales.”

Nvidia earnings highlight hyperscaler momentum

At the center of the discussion is the dependency Nvidia now has on a small group of massive customers. Hyperscalers such as Microsoft Corp., Google LLC and Amazon Web Services Inc. account for nearly 40% of the company’s sales, reflecting both the opportunities and risks of concentrated demand. Analysts noted that even without access to the Chinese market, the company continues to soar as hyperscalers scale up their AI factories.

“This company’s growing at 50%. They did 47 million, a billion on the top line. They’re growing 50%,” Vellante said. “To me the quarter was just a validation of everything that we’ve known is that the AI momentum continues.”

This heavy reliance comes as hyperscalers double down on the “spend more, save more” dynamic, where greater GPU investment translates into improved operational efficiency. It’s a dynamic shaping everything from model training to deployment strategies, according to Vellante.

“Spend more, save more actually works for the hyperscalers. I think that’s what GPT-5 was all about,” Vellante added. “GPT-5 … didn’t impress in the benchmarks; it’s just going to barely eke out the lead. But that was an economic announcement where they’re basically using better, more efficient infrastructure and dropping dollars to the bottom line. As long as that dynamic holds up and these guys are building data centers, by spending more with Nvidia, they actually will make more money.”

VMware Explore spotlights VMware and Broadcom’s evolving role

Another major topic was VMware Explore, where the post-acquisition direction under Broadcom drew significant attention. The narrative around VMware LLC has shifted dramatically — from speculation about its decline to recognition that product delivery under Broadcom is surprising many in the industry, Furrier explained.

“I think they’re really shocking the world with the fact that it went from, it’s a dead company narrative, the war against the company because of the price increases,” he said. “Hock Tan [has] delivered products, so we’re going to see that.”

Broadcom’s presence, or non-presence, at both VMware Explore and technical events such as Hot Chips highlighted its growing importance in AI-driven data center architectures. Its networking capabilities — particularly in switching and scaling — are increasingly viewed as critical complements to Nvidia’s GPU leadership. The integration of VMware’s software with Broadcom’s hardware ecosystem points to a strategic alignment aimed at competing in hybrid and on-prem AI solutions.

“Hot Chips … was a very much an insider technical event here in Palo Alto at Stanford University,” Furrier said. “Nvidia was there in force. All the top chip semiconductor folks. I did see some chip folks at VMware Explore, of course owned by Broadcom now.”

Enterprise adoption and evolving infrastructure

While hyperscalers accelerate, enterprises remain slower to commit. Proof-of-concept pipelines are growing, but many organizations are still grappling with security concerns, regulatory requirements and the complexities of integrating AI into existing workflows. Dell Technologies Inc.’s recent results, for instance, showed a dip in storage revenue, reflecting the uneven transition from cloud-centric growth to hybrid and on-prem AI deployments, according to Furrier.

“Dell’s a bellwether for the enterprise,” he said. “I think that supports our narrative that we’ve been reporting and commenting on is that there’s demand, there’s POC backlog, enterprises are a lot slower to adopt, there’s a high security bar, and that there was a trough of disillusionment in the past couple quarters on the enterprise. You’re starting to see that pick up.”

At the same time, vendors such as Broadcom are stepping up with new networking technologies designed to reduce bottlenecks in data centers. Innovations such as Tomahawk and Jericho switches promise to enhance utilization, creating more value out of expensive GPU deployments. This evolution underscores how the AI factory is no longer just about chips, but about a full stack of hardware, networking and governance tools.

“We’ve got Tomahawk 6, they got Tomahawk Ultra and Jericho. Those three chips essentially do scale up, scale out and scale across,” Furrier explained. “They’re way ahead of Nvidia on the optics, the co-packaged optics. I’m looking very carefully because I think there’s an open ecosystem approach that could be a threat to Nvidia.”

Watch the full podcast below to find out why these industry pros were mentioned:

Taylor Swift, American singer, songwriter
Jensen Huang, founder and CEO of Nvidia
Colette Kress, executive vice president and CFO of Nvidia
Ben Bajarin, principal analyst and CEO of Creative Strategies
Sarbjeet Johal, founder and CEO of Stackpane
David Floyer, analyst emeritus at theCUBE Research
Ram Velaga, SVP and GM at Broadcom
Paul Nashawaty, principal analyst at theCUBE Research
Ali Ghodsi, co-founder and CEO of Databricks
Rob Hof, editor-in-chief at SiliconANGLE Media
Elon Musk, CEO of Tesla
George Kurtz, CEO of CrowdStrike
Joe Tucci, chairman and co-founder at Big Growth Partners
Andre Agassi, former American tennis player

Here’s the full episode of this week’s theCUBE Pod:

Don’t miss out on the latest episodes of “theCUBE Pod.” Join us by subscribing to our RSS feed. You can also listen to us on Apple Podcasts or on Spotify. And for those who prefer to watch, check out our YouTube playlist. Tune in now, and be part of the ongoing conversation.

Photo: SiliconANGLE