As artificial intelligence transitions from experimentation to production, businesses are facing a fundamental shift in how they pay for the technology, according to an analysis published by InformationWeek by Cheri Williams, senior vice president at Hewlett Packard Enterprise. The piece argues that consumption-based pricing—paying per request or token—becomes economically inefficient once AI workloads reach steady, predictable scale. At that point, organizations need to consider whether owning dedicated capacity makes more financial sense than continuing to purchase AI one interaction at a time.

The analysis cites Deloitte's 2026 State of AI in the Enterprise, which shows worker access to AI increased 5% in 2025. More significantly, the proportion of companies with at least 40% of their AI projects running in production is projected to double within six months. This represents a major acceleration in how businesses deploy AI—moving from isolated pilots to production portfolios that include assistants, retrieval-and-knowledge systems, and agentic applications capable of executing multi-step workflows across enterprise systems. Customer-service, IT, research, and business-process agents now generate recurring demand across models, data, and tools.

According to Williams, when AI becomes a portfolio of always-on workloads rather than a collection of experiments, the financial calculus changes entirely. Consumption pricing offers teams flexibility and avoids upfront commitment, but when usage grows steady, predictable, and large enough to keep capacity productive, leaders must ask whether it still makes economic sense to buy AI one request at a time. The analysis emphasizes that ownership isn't automatically cheaper—it only becomes cost-effective when an enterprise can maintain high utilization. Every organization has a crossover point where sustained use makes owning capacity more economical than purchasing it per request, but there's no universal threshold because it depends on the models used, the ratio of input to output tokens, performance needs, system architecture, energy expenses, and the operating model needed to support it.

The shift matters because it reframes AI infrastructure as a strategic asset rather than a variable expense. Williams explains that a retrieval-heavy knowledge system carries a very different cost profile from a simple assistant because it may process far more context per interaction. Agentic workflows present yet another pattern: a single business task can involve repeated reasoning, retrieval, model calls, and tool use. That's why generic cost benchmarks fall short—enterprises need to model their actual workloads, understand expected demand, and size capacity accordingly. At the right utilization level, the benefit isn't only lower effective cost but also greater predictability: the ability to manage AI capacity as strategic infrastructure investment rather than watch monthly spending fluctuate with model use and workload demand.

Williams recommends leaders ask three questions before committing capital: Is demand becoming steady, predictable, and large enough to justify dedicated capacity? At what level of usage does ownership make economic sense? Can the organization keep that capacity productive through adoption, governance, and continued use-case expansion? The analysis concludes that capital investment is only half the equation—even when the economics support ownership, capacity creates value only when the business gets workloads into production quickly and keeps them running. That requires an operating model connecting technology to adoption and business outcomes: bringing users and workloads on board, governing how AI is used, reviewing utilization, and continually identifying the next high-value use case. The goal is to create value early, then build on it by measuring use, identifying underutilized capacity, and bringing additional high-value workloads onto the platform over time—turning AI capacity into a productive asset the business can optimize, expand, and use to create measurable value. Organizations that look beyond token prices and the latest model will know when recurring demand calls for a different economic model—and have the operating discipline to make that capacity productive. The calculus isn't whether to own infrastructure outright, but whether the business has reached the scale where treating AI as fixed capital rather than variable cost unlocks better economics and control.