CFOtech UK - Technology news for CFOs & financial decision-makers
Flux result a1f1c93f c5b0 42b1 ad83 0078be2fe4c0

Anthropic shifts enterprise billing to token-based pricing

Fri, 17th Apr 2026 (Yesterday)

Anthropic has changed Claude's enterprise billing model from fixed per-seat subscriptions to per-token pricing with mandatory monthly spending commitments. The change will affect customers on existing enterprise plans when they renew their contracts.

The new structure replaces Anthropic's Premium and Standard tiers with two role-based products: Claude Code at USD $20 per user per month for technical staff, and Claude.ai at USD $10 per user per month for business users. Those seat charges now cover platform access only. Usage across Claude, Claude Code and Cowork is billed separately at standard API rates based on actual consumption.

Customers must also accept a monthly spending commitment based on Anthropic's estimate of their token use. They pay that amount whether usage reaches the estimate or not.

The changes also eliminate API volume discounts of 10 to 15 per cent that previously applied to larger enterprise users. Together, lower seat fees, separate usage billing and mandatory consumption commitments are expected to raise overall costs for many businesses.

NPI Financial, an IT procurement advisory firm, said the revised model would increase total cost of ownership for most organisations. It advised enterprise buyers to seek more detail on how Anthropic calculates expected consumption, challenge commitment levels and negotiate concessions elsewhere in contracts to offset the loss of discounts.

Pricing shift

The revised terms mark a significant change in how Anthropic charges large customers. Under the previous system, enterprises could buy seats on a fixed monthly basis at USD $200 per user for Premium and USD $40 per user for Standard, giving procurement teams a more predictable software bill.

Now, a lower headline seat price sits alongside metered charges that can rise or fall with employee activity. For finance and procurement teams, that creates a variable cost model more commonly associated with cloud infrastructure than workplace software subscriptions.

The change comes as demand for generative AI services continues to surge. Anthropic's annualised revenue run rate has risen from about USD $9 billion at the end of 2025 to USD $30 billion, and more than 1,000 customers now pay over USD $1 million a year.

That growth has increased pressure on the computing resources needed to run large AI models. Market reporting has linked Anthropic's billing overhaul to a shortage of compute capacity that has strained infrastructure across the sector.

Compute pressure

API reliability has become one visible sign of that strain. Anthropic's API uptime over the 90 days ending April 8 was 98.95 per cent, according to figures cited by The Wall Street Journal, below the 99.99 per cent benchmark commonly associated with established cloud providers.

At the same time, the cost of hardware used to train and run AI models has increased. Blackwell GPU rental prices have climbed 48 per cent in two months, while Bank of America has forecast that compute demand will exceed supply through 2029.

Against that backdrop, the new enterprise billing model shifts more of the financial risk of uncertain usage onto customers. Instead of offering broad flat-rate access and absorbing the risk of higher-than-expected consumption, Anthropic now locks in a customer commitment in advance and charges separately for the resources used.

For Anthropic, the approach could make it easier to sign larger contracts without first securing enough GPU capacity to support unlimited, or broadly inclusive, seat-based access at a set monthly price. For buyers, it reduces cost certainty at a time when internal demand for AI tools can be hard to forecast.

Market trend

The shift also reflects a broader move across the AI software market away from simple subscription plans. OpenAI moved Codex from flat-message pricing to token metering in early April, and GitHub tightened usage limits for Copilot on April 10.

That pattern suggests enterprise buyers may need to adapt procurement processes built around fixed software licence models. AI products are increasingly sold with pricing tied to underlying compute consumption, making spend management, usage monitoring and contract negotiation more central to purchasing decisions.

Anthropic's existing customers on older seat-based contracts will have to move to the new framework at renewal to keep using the company's enterprise products, or lose their grandfathered terms.

NPI Financial recommends that buyers press for transparency into Anthropic's consumption estimates, test whether monthly commitments reflect realistic demand and seek to recover the value of discontinued API discounts elsewhere in negotiations.