en.Wedoany.com Reported - Companies are growing anxious over the soaring costs of AI usage. Uber had already exhausted its entire 2026 AI programming budget by April; Microsoft revoked licenses for Claude Code just months after granting them to developers; and a Priceline employee told TechCrunch that the company's regular Cursor contract renewal fees had increased four to five times.

Although the price per token has decreased, the widespread adoption of AI applications and increasingly autonomous agents is driving a sharp rise in total consumption. Companies that heavily subscribed to unlimited plans in early 2025 are now scrambling to trace where their money went, cut expenses, and recover return on investment from budget pressures. Meanwhile, a market catering to cost-tracking needs is emerging, with startups, existing vendors, and a new standards body all competing to provide enterprises with relevant tools and terminology.

"Six months ago, clients would only ask, 'What can it do? Is it good enough?'" Alexander Embiricos, head of enterprise at OpenAI, told TechCrunch at an event in New York City. "Now the conversation is completely different, becoming 'We're spending too much money. What visibility can you provide? What audit capabilities do you have? What token controls? How efficient are your models?'"

Against this backdrop, the Linux Foundation this week announced the Tokenomics Foundation initiative, a new standards body aimed at bringing the discipline of FinOps—the cost discipline established for cloud spending—to AI token expenditure.

"In April and May, I started hearing companies say: 'Oh my god, we've already overspent three times our entire 2026 token budget, and it's only April,'" J.R. Storment, executive director of the FinOps Foundation under the Linux Foundation, told TechCrunch. "We started hearing existential crises. The entire conversation shifted from 'token maximization' and 'move fast' to 'we need guardrails, how do we control this?'"

These calls come after CEOs had previously fervently urged teams to use the best models regardless of cost to accelerate projects. New models released in November, such as Anthropic's Claude Opus 4.5, OpenAI's GPT-5.1, and Google's Gemini 3 Pro, brought significant improvements to agent tools, leading to exponential increases in consumption. One company reportedly faced a $500 million Claude bill after forgetting to set usage limits for employees.

"It's like a crack cocaine epidemic," said Chris Reed, Senior Director of IT Financial Management at Priceline. He noted that the company has started setting token limits for certain groups: "They give you a taste for free, get you addicted, and then you're locked in."

Vitaly Gordon, CEO of engineering operations platform Faros AI, said he recently spoke with a CTO who mentioned: "One of my engineers spent $40,000 on tokens last month. I honestly don't know whether to stop him or tell everyone else to be like him."

A March survey by Faros found that among 20,000 developers, output was increasing, but so were bugs and rewrites. Engineering management platform Jellyfish similarly found that the engineers using the most tokens were roughly twice as productive as those using AI less, but they spent ten times the number of tokens to achieve that.

Nicholas Arcolano, head of research at Jellyfish, told TechCrunch via email that the explosion in AI spending is largely due to agentic features, with consumption per developer increasing about 18.6 times over nine months. These figures blur the correlation between productivity gains and spending increases. "Whether extreme spending is worthwhile ultimately depends on the final business value (e.g., revenue) of the code shipped, and most companies still can't measure that," Arcolano said.

Part of the measurement problem lies in the sheer scale of current AI usage. "Tracking cloud costs is a data problem involving hundreds of millions of rows per month," Storment said. "Tracking token costs is a data problem involving trillions of rows per month. You can't just dump this data into any spreadsheet or even basic tool. You have to fundamentally rethink your tools, norms, and accounting systems."

At Priceline, Reed has already seen discrepancies, noting issues between vendor-reported usage and Priceline's internal data. "My career started in telecom expense management, and I see all the parallels from telecom to cloud to AI," he said. "Whenever something new is introduced, there are billing errors and opportunities for auditing and optimization."

A market is forming around this issue. Pure-play companies like Pay-i can track, measure, and optimize the cost and performance of GenAI investments; Paid allows developers to track costs, measure usage, and charge users based on actual value rather than subscription fees. Others like Jellyfish, Waydev, and Faros AI offer AI agent monitoring to prove the ROI of developer tools. Storment said most of the 180 vendors under the FinOps Foundation are leaning into this space.

Companies with existing distribution channels are also adding new features. Ramp recently entered the AI spending management space; Datadog and New Relic have added services like cloud cost management, token-level observability, and GPU monitoring. At next week's FinOps X conference, AWS is expected to unveil new financial management features for enterprise AI spending.

Tiffany Luck, a partner at NEA, believes token efficiency and observability might be added "at the harness layer or application layer." She mentioned startup Factory, which this week launched a model router that automatically selects the appropriate model for each task. Gordon expects frontier labs and other model providers to adopt OpenRouter-style optimization, routing queries to the cheapest model—a trend already visible in enterprise Claude bills. "Even if you call the Opus model, part of the cost gets attributed to Sonnet or Haiku," Gordon said, "because the latter is smart enough to do the job. I think this will become increasingly common."

However, all these tools are being built without a common language or shared definitions. This is where the Tokenomics Foundation hopes to make a difference. The foundation is building canonical definitions and frameworks for "tokenomics"; developing open standards, norms, and metrics for AI token usage and billing; and creating new metrics like cost per intelligence or tokens per watt. It also plans to define metrics for token factory efficiency and consumption efficiency. The organization plans to officially launch in July and will announce more members at next week's FinOps X conference.

"The token economy is fundamentally more abstract and opaque than anything we've managed at this scale before," Nishant Gupta, Chief Availability Officer at Salesforce, said in a statement. "It requires a different operational muscle than what the industry built for the cloud."

Although Goldman Sachs predicts global token usage will grow 24-fold by 2030, companies already over budget need solutions now, and the foundation's first deliverables are still months away. "Maybe we invented the steam engine, but we haven't figured out the assembly line yet," Gordon said. According to Arcolano, a prudent approach is to adopt broadly and moderately. "The best ROI comes from moving a broad middle tier from low usage to medium usage, rather than pushing heavy users even higher," he said.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com