Americans have big concerns about artificial intelligence, but that hasn’t stopped OpenAI Inc., Nvidia Corp., IBM Corp. and others from charging ahead.

Last week saw shares of Nvidia, CoreWeave Inc., Oracle Corp. and Nebius Group N.V. decline amid reports that OpenAI’s annualized net revenue was approximately $50 billion, roughly $20 billion below previously reported figures. Despite the ongoing public pushback against AI, John Furrier (pictured, right), research executive analyst at theCUBE Research, sees the uproar as temporary.

“This is an AI freakout moment,” he said. “It’s funny to watch knowing what’s happening on the inside. We talk to all those suppliers, whether it’s Nvidia, the chip side … the neoclouds or the AI infrastructure clouds. This new AI layer is a system; the data center as a system is now the computer. Rack scale is changing. Companies like Marvell, Broadcom and others are creating chips to create more efficiencies.”

On the latest episode of theCUBE Pod, Furrier and Dave Vellante (pictured, left), chief analyst at theCUBE Research, discussed takeaways from the recent IBM Analyst Council. They also explored chipmaker Marvell Technology Inc.’s growth and the industry shift toward inference.

IBM bets on commodity models

Big Blue’s recent meeting provided insights into the legacy company’s position in the AI era. As Vellante sees it, customers are increasingly spending on the infrastructure and models needed to produce intelligence before turning to IBM to put that intelligence to work, presenting a challenge for the company as the AI market matures.

“The alpha is going to be in your data, your processes and your operating model,” Vellante said. “This is [IBM’s] opportunity. That AI operating model is going to be the new source of competitive advantage. It’s not going to come from Palantir; it’s not going to come from models. It’s going to come from your own proprietary data, how you apply that data, how you streamline your processes and how fast you can get products and services to market.”

IBM has promoted a four-part architecture composed of agents, data, automation and hybrid infrastructure, with an orchestration layer that can integrate agents from different vendors. Rather than relying exclusively on proprietary models, IBM is positioning itself as a provider of hybrid AI operating infrastructure for enterprise customers, building on Red Hat’s open-source foundation.

“Their bet is the commoditization of the models,” Furrier said. “Having a nice model management capability will be good for IBM; but when you have a commoditization bet, you better be right. IBM doesn’t need its own proprietary operating system [because] … they have a valuable abstraction control layer around open infrastructure through Red Hat. Red Hat gets paid a lot of money to manage that open abstraction. That’s an opportunity: IBM could run all the agents.”

The question remains whether IBM’s bet on commodity models and open-source solutions will pay off. Customers trust IBM. but they don’t see the company as an innovator in the current AI landscape, according to Vellante. However, if IBM can establish itself as a provider of the operational infrastructure around AI models and agents, then it could own “the scarcity around the abundance.”

“They’ve got to play in both camps,” Vellante added. “In other words, they’ve got to have the IBM version of Oracle’s Red Stack. Last year, we asked [IBM’s CCO] Rob Thomas, what is the platform? And the answer came back: Red Hat. I agree. Let’s build a solution around Red Hat to simplify the adoption of AI. It’s unclear to me what that solution is. It’s clear to me they got all the piece parts.”

AI market enters the inference stage

In the chip world, Marvell Technology Inc. is drawing attention with an ambitious revenue target of $70 billion to $90 billion for fiscal 2031. Earlier this year, the company, which makes custom AI processors and optical networking technology, announced a partnership with Nvidia’s NVLink Fusion.

“What impressed me about Marvell is that they had a 10-year vision,” Furrier said. “Their mission statement that they put up from 10 years ago … they didn’t have factories. It wasn’t in the dialogue. But back then … they made the argument that AI research is becoming a connectivity problem as much as a compute problem. And they’re not wrong. You’ve got a semiconductor company talking about economics. They’re talking about real value.” 

Furrier doesn’t believe the AI market is approaching a bubble, while Vellante sees the breaking point as being farther away than many might expect. As the enterprise world shifts its focus from training to inference, companies such as Marvell and the neoclouds stand to benefit financially as spending on AI infrastructure increases.

“A year ago, 90% of the workloads — we saw this at CoreWeave — were training,” Vellante said. “Today, it’s like 60-40 training-inference. The expectation is that it’s going to flip to 10-90. Why is that? Because they’re driving revenue; they’re making money off of inference. That’s where you talk about this all the time. You go to college for training, and then you go make money and you continually learn.”

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

Ray Dalio, chief investment officer of Bridgewater Associates
Matt Calkins, CEO of Appian
David Lazovsky, EVP and GM of the Data Center Networking Business Group at Marvell Technology
Arvind Krishna, chairman and CEO of IBM
Rob Thomas, SVP of software and chief commercial officer at IBM
John F. Akers, former CEO of IBM
Prith Banerjee, SVP of innovation and member of the executive leadership team at Synopsys
Matt Murphy, partner at Menlo Ventures
Charlie Kawwas, president of Broadcom
Donald Trump, 47th president of the United States of America

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

You can also listen to theCUBE Pod on Apple Podcasts or on Spotify. For video episodes, visit out YouTube playlist. Tune in now, and be part of the ongoing conversation.

Photo: SiliconANGLE