Why consumption pricing is breaking budgets - and how to combat it

When news broke back in May this year that Uber developers had chewed through millions of tokens experimenting with AI coding tools, and burned through its entire annual AI budget in just four months, the reaction across the tech industry was split. To outsiders, it sounded like an operational error, a classic case of runaway experimentation. To those of us monitoring the underlying economics of software, it was something far more systemic: a high-profile symptom of a massive, structural shift in how software is bought, sold, and monetised.

 

For years, the software industry ran on a relatively predictable mechanism: the per-seat subscription. Today, that model is dying. 

 

In the past 24 months alone, 23% of software vendors have moved away from flat seat-based pricing towards usage, hybrid, and token-based models. The shift is being driven by the aggressive rollout of AI capabilities and multi-model options, as vendors look to protect baseline seat revenue while monetising extra AI features.

 

The vendor pitch sounds universally attractive: “Only pay for what you use.” In reality, consumption-based pricing is turning out to be one of the most effective revenue extraction mechanisms the SaaS industry has ever devised.

 

It is not offering financial control; it is delivering absolute chaos.

 

The mirage of pay-as-you-go

On paper, usage-based pricing feels inherently fair. It promises to align software costs directly with actual business value while eliminating the waste of shelfware - the decades-old problem of per-user pricing models, where expensive licences sit unused or even unassigned. 

 

In practice, the underlying economics favour the software vendor entirely. Data reveals that when the top SaaS vendors switched to consumption or hybrid pricing models, their Average Contract Values increased over 50%.

 

To make matters worse, traditional procurement levers are failing. Average contractual discounts when paying by consumption are 30% lower than those secured when paying by seat.  Vendors can easily push back with a simple rebuttal: “If you want to control spend, just control usage.”

 

It is a perfect trap, as no forward-thinking organisation wants to stifle internal adoption just to hit an arbitrary financial target (in truth, few could control usage even if they wanted to). 

 

But the main reason is that traditional financial forecasting simply wasn’t built to handle software contracts that act like volatile commodities. And few finance leaders know how to adapt.

 

Taking charge of runaway AI and SaaS costs

Managing usage-based software with a static, point-in-time procurement process makes runaway costs inevitable, so organisations that want to regain control must overhaul how they negotiate contracts and manage costs.

 

1. Understand exactly what you’re paying for before you sign: Buyers need to look past vague credit bundles and explicitly demand a granular breakdown of what constitutes a task cost (e.g. whether billable units are metered per API call, compute hour, data volume processed, or completed agent outcome). 

 

Organisations should match the precise price of a single task to how often this tool would be required to perform it over the course of a week. Understanding these baseline micro-costs means having access to internal tool activity data, a good understanding of how departments operate functionally and what their longer term goals are. 

 

2. Set flexible usage limits and use market data to your advantage: Since big-volume discounts are harder to win, set flexible spending limits based on how much you use. 

 

Contracts should include pre-negotiated, capped overage rates, strict vendor notification alerts when usage approaches budget thresholds, and automated pricing adjustments that decrease unit costs as consumption thresholds are crossed. Teams should automate their contract analysis to flag non-compliant terms and leverage real-time, cross-market pricing benchmarks to evaluate whether consumption offers are truly competitive.

 

3. Set up your request process to flag problems early: Controlling spend also requires using information at intake to create an “early warning system” within your procurement workflow.

             

Pre-approval processes should identify that a consumption-based pricing model is being offered. These checks should include what sort of tasks the tool will be used for, by whom and how often, what would trigger an overage and who owns usage governance post-signature.

 

4. Track spending and progress as it happens: Finally, enforcing continuous usage monitoring turns vendor management into an ongoing operational discipline. Usage spikes occur in hours rather than months. Real-time visibility into cross-platform execution and spend is the only way to catch operational anomalies, runaway loops, and orphan integrations before they escalate into multi-million-pound budget overruns.

 

Taking back control

The transition toward consumption and hybrid pricing in SaaS is becoming the baseline standard as AI capabilities deepen across every software category. Vendors will continue to push these models because they align with their underlying infrastructure costs and dramatically raise revenue ceilings.

 

For buyers, the solution is not to resist AI or ban usage-based tools entirely. It is to recognise that a dynamic financial model cannot be managed with static administrative tools. To successfully govern these environments, organisations must deploy dedicated AI cost governance platforms - such as Vertice’s AI Cost Optimisation tool - to gain a single, unified view of all model usage, spend, and commitments. 

 

By combining granular cross-platform tracking, automated 90-day spend forecasts, and intelligent recommendations to right-size commitments or swap to cheaper models, finance leaders can proactively curb runaway spend. Meanwhile, most of their peers will continue to be just one R&D project away from being the next Uber…

 

---

 

Eldar Tuvey is CEO and Co-Founder at Vertice

 

Main image courtesy of iStockPhoto.com and Shinsei Motions