Dell and H2O.ai target the token-cost problem with vertical AI models

As artificial intelligence adoption accelerates inside enterprises, the economics of generative AI are forcing a fundamental rethink. Runaway token costs, data sovereignty demands and a growing gap between AI pilots and production ROI are pushing organizations to reconsider where their models run — and what kind of vertical AI models they actually need.

The answer increasingly points toward on-premises infrastructure and vertical AI models purpose-built for specific industries, rather than large general-purpose models consuming tokens at scale in the cloud, according to Satish Iyer (pictured, right), vice president and chief technology officer of technology innovation and ecosystems at Dell Technologies Inc.

“Our aspire[ation] in Dell [is] to bring AI to where the customer data is, as simple as that,” Iyer told theCUBE, SiliconANGLE Media’s livestreaming studio. “There is no AI without data, and most of enterprise data stays on-prem. It’s important for us to support an enterprise journey where we can make sure that enterprises are able to leverage AI to drive the right outcomes within that business without worrying about token cost.”

Iyer and Sri Ambati (left), founder and chief executive officer of H2O.ai Inc., spoke with theCUBE’s Gemma Allen at Dell Technologies World, during an exclusive broadcast on theCUBE. They discussed vertical AI models, token economics, sovereign AI deployment and the Dell–H2O.ai partnership for enterprise AI activation. (* Disclosure below.)

Vertical AI models power the next wave of enterprise ROI

The urgency behind the conversation is real. Dell has surpassed more than 5,000 AI factory deployments globally, with enterprises in financial services, healthcare and telecommunications leading adoption as they seek predictable costs and data control. The pressure on token spend is now a boardroom-level conversation, with top developers at H2O.ai burning $1,000 a day in tokens — but many users generating far less measurable return on that investment, according to Ambati.

“It’s how do you do different pieces, orchestrate them across different ecosystems, whether it’s Gemini or … OpenAI models and open-source models,” Ambati said. “Open-source models, which run on a Dell AI factory, can give you that predictability for your token consumption, token pricing as well. Some of our customers are in billions of tokens, tens of billions of tokens a day. We moved almost half of them onto Dell and H2O with small language models.”

H2O.ai’s approach reflects a broader shift in the market toward vertical AI models that combine predictive and generative capabilities within a single, industry-specific framework. The company’s recently released TabH2O tabular foundation model is designed to deliver predictions across structured enterprise data without requiring expensive parameter tuning, Ambati explained. The path forward lies in distilling large models into smaller, more deployable assets that can run at the edge — in mining operations, hospital networks and contact centers — where the data is generated and decisions must be made locally, according to Ambati and Iyer.

“H2O has been doing a lot of vertical-specific AI models, solving a lot of specific industry-specific problems,” Iyer said. “For them to leverage our AI Factory to build some of those vertical models and solve some of the vertical problems is a great way for us to activate enterprise.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Dell Technologies World 2026:

(* Disclosure: TheCUBE is a paid media partner for Dell Technologies World 2026. Sponsors of theCUBE’s event coverage do not have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE