Token prices for large‑language models have collapsed from roughly $20 per million tokens in late 2022 to about $0.40 today – a 98% decline. Paradoxically, enterprise AI spending has exploded, with industry analyses estimating a 320% increase in bills since the price drop. The surge stems from the rapid adoption of agentic AI tools that multiply token consumption per task.
Anthropic’s Claude Opus 4.5, OpenAI’s GPT‑5.1 and Google’s Gemini 3 Pro, released after November 2025, have turned single‑step workflows into orchestrated agentic systems. Where a linear interaction cost $0.04 in 2023, the same process now runs roughly $1.20 – a thirty‑fold jump. Per‑developer token use has risen about 18.6 times in nine months, according to Nicholas Arcolano, head of research at Jellyfish. Heavy token users report productivity gains but spend ten times the tokens of lighter users.
Overspend is already a reality. Uber exhausted its entire 2026 AI coding budget by April. Microsoft revoked Claude Code licences for developers six months after granting them. One unnamed firm racked up a $500 million Claude bill in a single month after forgetting to set usage limits. Priceline’s senior IT finance director, Chris Reed, said a routine contract renewal with the Cursor tool returned four to five times the previous cost.
“Six months ago, I would have a conversation with a customer about ‘What can it do?’ Today it’s ‘We’re spending so much. What visibility do we have?’" said Alexander Embiricos, OpenAI’s head of enterprise. J.R. Storment, executive director of the FinOps Foundation, echoed the shift: “The conversation moved from token‑maxxing and ‘go fast’ to ‘we need guardrails, how do we control this?’”
To address the mounting financial strain, the Linux Foundation unveiled plans for the Tokenomics Foundation. The new standards body aims to define “tokenomics,” publish open standards for AI token usage and billing, and introduce metrics such as cost‑per‑intelligence and tokens‑per‑watt. A formal launch is slated for July. Nishant Gupta, chief availability officer at Salesforce, warned that “token economics is fundamentally more abstract and opaque than anything we’ve managed at this scale before.”
Industry players are already racing to fill the gap. Startups like Pay‑i and Paid offer AI‑spending optimization and value‑based billing. Companies such as Jellyfish, Waydev and Faros AI provide monitoring to prove ROI on developer tools. Larger observability platforms – Datadog and New Relic – have added token‑level metrics. Model routing is emerging as a primary cost lever; Factory, an enterprise AI coding startup, introduced a router that automatically selects the cheapest adequate model for each task.
Goldman Sachs projects global token usage will multiply 24 times by 2030. For firms already over budget, solutions are needed now, even as the Tokenomics Foundation’s first deliverables remain months away. As Vitaly Gordon, CEO of Faros AI, put it, “Maybe we created a steam engine, but we still haven’t figured out the assembly line.”
Questo articolo è stato scritto con l'assistenza dell'IA.
News Factory SEO ti aiuta ad automatizzare i contenuti delle notizie per il tuo sito.