Oracle has once again grabbed headlines, but this time not just for its strong financial results. The company has announced a strategic shift in its billing model for artificial intelligence services, moving towards an outcome-based scheme instead of the traditional token-based pricing. This move, although incipient, could redefine how enterprises acquire and consume AI in the cloud.

Record financial results and a multi-billion dollar bet

In the fourth quarter of fiscal year 2026, Oracle reported total revenues of $67.4 billion, up 17% year-over-year, and net income of $17.087 billion, a 36% increase. Cloud was the main driver: cloud infrastructure revenues grew 93% year-over-year, fueled by demand for AI workloads and database services. To sustain this growth, the company plans to invest over $70 billion in capital expenditure next year, according to CFO Hilary Maxson during the analyst conference.

This level of investment reflects Oracle's confidence in its customers' committed demand, which CEO Mike Sicilia says "are now focused on how to leverage AI in their own businesses. They want AI to boost productivity, improve customer service, and generate real competitive advantages. So they want to do it quickly and within their current budget."

The outcome-based billing pilot program

In response to this demand, Oracle has launched a pilot program that, according to Sicilia, "will align pricing directly with the value obtained." The program, deployed in the last quarter, involves 33 organizations and focuses on offering enterprise-ready AI agents. "Over the past year, we have deployed more than 1,000 AI agents in our application suites. These agent-based offerings can reason, decide, and execute tasks across processes," added Sicilia.

This outcome-based billing model aims to simplify AI procurement for enterprise customers, who often face the complexity of variable consumption models based on tokens. However, analysts warn that this is not a complete abandonment of token economics, but rather a commercial abstraction.

The end of tokens? Not so fast

Sanchit Vir Gogia, chief analyst at Greyhound Research, states that "Oracle's move towards outcome-based AI billing should not be interpreted as the death of token economics. Tokens are not disappearing; they are being hidden behind a more business-friendly interface. The counter is still running somewhere."

Gogia points out that token-based pricing is a "terrible language for enterprise budgeting," as agentic workloads can consume up to a thousand times more tokens than simple tasks, with variations of up to thirty times in identical executions. Oracle's advantage, according to him, lies in its proximity to the system of record, allowing it to define outcomes more credibly than other providers. However, he warns that "the danger begins precisely there: outcome-based pricing sounds cleaner until the provider becomes both supplier and arbiter."

Scott Bickley, advisor at Info-Tech Research Group, agrees that this model is where CIOs want to go, because it's easier to define business value. However, he notes that the current abstraction of tokens into units like "AI credits" or "agent actions" still requires understanding the underlying mechanics. "If you can completely abstract that pain and offer me an outcome for a price, then you are far ahead of most of the market," says Bickley.

Implications for CIOs and the multicloud market

Gogia advises CIOs to separate the construction impulse from execution certainty, as "confusing the two would be naive." The real bottleneck in AI is no longer GPUs, but energy, permissions, and policy. Additionally, he highlights that these results are not just about AI infrastructure, but also Oracle's attempt to make itself harder to avoid in a multicloud enterprise world.

This move by Oracle aligns with broader industry trends, where providers like JetBrains and Cisco are also adapting their models to integrate AI more seamlessly. Outcome-based billing could be the next step in the evolution towards more accessible and predictable AI for businesses, although tokens will remain the invisible engine under the hood.

Original source: ComputerWorld. Analysis and adaptation by ForgeNEX.