Artificial intelligence is no longer just a software story — it is the foundation for AI factories, where chips, data pipelines and networks work together as production systems for digital outcomes. This emerging model is redefining how performance is measured and how value is created across industries.

These AI factories combine custom semiconductors, low-latency fabrics and massive data instrumentation into real-time feedback loops that generate competitive advantage. They are reshaping business models as chipmakers secure long-term custom contracts and enterprises harness telemetry to transform traditional experiences into living, digital ecosystems. The economics of compute now rival the economics of manufacturing.

This framing sets the stage for the latest episode of theCUBE Pod with theCUBE Research’s John Furrier (pictured, left) and Dave Vellante (right), who explored how Broadcom Inc., Nvidia Corp. and IBM Corp. are navigating this shift. Their conversation ranged from real-time analytics at the U.S. Open to the rise of custom silicon deals and the broader race to scale AI-powered infrastructure.

“This is going to be a continuing boom for the Broadcoms and the Nvidias because they’re the arms dealers and the supply chain,” Furrier said. “Nvidia is going after the enterprise. Broadcom is taking more of an open approach.”

How AI factories are built and monetized

The concept of AI factories came up repeatedly as a lens for interpreting today’s market. The phrase underscores that compute resources are not abstract tools, but production lines for data-driven outcomes. Broadcom and Nvidia illustrate two ends of this spectrum. Nvidia pursues a vertically integrated model where hardware, software and systems are tightly coupled. Broadcom takes a modular approach, pursuing multi-year custom silicon deals while pairing semiconductors with software to balance cycles, according to Vellante.

“They’ve turned the US open into an AI data factory,” he said about theCUBE’s recent attendance at the sporting event. “They don’t use that marketing. That’s Dell and Nvidia, but it really is a data factory. They’ve got over a hundred years of matched data. They got all the history on where the balls lands. The note I have here is four and a half million data points per match and it’s unbelievable.”

The US Open became a showcase of AI factory economics, with IBM’s watsonx processing millions of data points in near real time across 17 courts. Broadcom’s $16 billion quarter underscored the same theme: Long-term chip deals and capital-intensive ecosystems are redefining vendor strategies, customer experiences and enterprise value, Furrier explained.

“The OpenAI deal represents what I call the white box model for the large-scale clusters, which is simply there will be a market that will rise up to feed the demand for essentially custom chips that will feed into what could be a clone of Nvidia, meaning purpose-built systems,” he said. “OpenAI had a huge incentive in cash to vertically integrate their own. We’ve been having this debate, heterogeneous versus homogenous.”

Ecosystems, enterprise shifts and competitive dynamics

Beyond the headlines on earnings and chip design, the broader conversation centered on how ecosystems evolve. AI factories are not just hyperscaler phenomena. Their influence touches enterprises through partnerships, data integration strategies and the application of agentic workflows. That’s why upcoming conversations in the AI factories series are significant: They extend the debate from chips and cloud clusters into enterprise readiness, interoperability and governance, Vellante emphasized.

“Broadcom … almost $16 billion in revenue, up 23% in the quarter. That’s just the quarter, operating income at six billion, almost 37% of revenue,” he said in reference to their place in the industry. “Their free cash flow is $7 billion, 45% of revenue — that’s higher than Microsoft; that’s higher than Nvidia. The thing about Broadcom … they have these long duration, high value, multi-year, multi-billion-dollar custom chip deals.”

Nvidia’s integrated approach, framed by “Jensen’s Law,” highlights doubling performance annually, though its enterprise impact remains in question. Both Nvidia and Broadcom are benefiting from AI demand, while Salesforce Inc., IBM and others push agentic, data-driven growth. Yet Wall Street often overlooks these strategies, underscoring why financial and technical roadmaps remain inseparable.

“I would say that Nvidia’s on the learning curve and Nvidia’s ahead of everybody to the extent that you need the latest greatest,” Vellante explained. “They are furthest along on the learning curve, which means they have the most innovation and the lowest cost, not the lowest price of course, but that’s their moat. I see it continuing for quite some time unless and until those laws become non-immutable.”

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

Mark Zuckerberg, CEO of Meta Platforms
Sam Altman, co-founder and CEO of OpenAI
David Sacks, White House AI and crypto czar
Tim Cook, CEO of Apple
Jensen Huang, founder and CEO of Nvidia
Elon Musk, CEO of Tesla
Jim Cramer, American TV personality and author
Arvind Krishna, chairman and CEO of IBM
Bruno Aziza, group VP, data, BI and AI at IBM
Aaron Rakers, managing director and technology analyst at Wells Fargo
Pat Gelsinger, former CEO of Intel
Paul Maritz, chairman at Pivotal and former CEO of VMware
Jeff Bezos, chairman of Amazon
Michael Dell, chairman and CEO of Dell Technologies
Marc Benioff, chair and CEO of Salesforce
George Gilbert, principal analyst at theCUBE Research
Jamie Dimon, chairman and CEO of JPMorgan Chase
Sarbjeet Johal, founder and CEO of Stackpane
Jonathan Cantor, SVP and deputy chief privacy officer of Truist
Lina Khan, former chair of the FTC
Bill Gates, co-founder of Microsoft
Janet Reno, former U.S. attorney general
Jack Ma, Chinese business magnate and investor
Donald Trump, 45th and 47th president of the United States of America

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