- Anthropic unbundles Claude, Claude Code, and Cowork from enterprise seat fees, moving to per-token billing at standard API rates on top of a flat $20/month per-user fee.
- The company’s annualized revenue tripled from $9 billion at the end of 2025 to $30 billion by April 2026, with over 1,000 enterprise customers paying more than $1 million annually.
- GPU rental prices surged 48% in two months, CoreWeave raised prices over 20%, and Bank of America projects compute demand will outstrip supply through 2029.
- OpenAI, GitHub, and Windsurf have made similar moves, suggesting every major AI provider running agentic workloads will adopt usage-based enterprise billing within six months.
Anthropic has quietly restructured its enterprise pricing model in one of the most significant commercial shifts in the AI industry this year. According to an exclusive report by The Information on April 14, the company is moving away from flat-rate seat licenses and billing enterprise customers based on actual AI consumption. The change arrives as Anthropic’s revenue rockets to $30 billion in annualized run rate and the broader industry grapples with a deepening compute crunch that is forcing every major provider to rethink how they charge for intelligence.
What Changed and Why It Matters
From Flat Fees to Metered Tokens
Under the previous model, Claude Enterprise customers paid up to $200 per month per licensed user with set token allocations bundled in. The new structure decouples usage entirely: organizations now pay a flat $20 per month per user plus separate charges for all compute consumed across Claude, Claude Code, and Cowork at standard API rates. Anthropic’s enterprise help center now states plainly that “all usage across Claude, Claude Code, and Cowork is billed separately at standard API rates, based on what your team actually consumes.”
Organizations on older seat-based plans with fixed usage allowances must migrate by their next contract renewal or lose grandfathered terms. Fredrik Filipsson, co-founder of Redress Compliance, warned that the pricing changes “will double or even triple the cost for heavy users.” The shift is particularly impactful for engineering teams relying heavily on Claude Code, where prompt-cache time-to-live was already reduced from one hour to five minutes in early March.
Revenue Growth That Forced the Hand
The pricing overhaul is inseparable from Anthropic’s explosive growth trajectory. The company’s annualized run rate climbed from $9 billion at the end of 2025 to $30 billion by April 2026, a tripling in roughly four months. More than 1,000 enterprise customers now pay over $1 million annually. That kind of demand on flat-rate plans creates an unsustainable cost structure when every inference burns real GPU cycles, especially as agentic workloads consume significantly more compute than traditional chat interactions.
The Compute Crunch Driving the Shift
Infrastructure Costs Are Spiraling
The macro context makes the shift almost inevitable. GPU rental prices for Blackwell chips increased 48% in just two months. CoreWeave, one of the largest GPU cloud providers, raised prices over 20% and now enforces three-year minimum contracts for smaller customers. Bank of America projects that compute demand will outstrip available supply through 2029. PJM, the eastern U.S. grid operator, is seeking 15 gigawatts of additional AI-related power capacity by early 2027.
Meanwhile, usage is exploding across the industry. OpenAI’s token consumption jumped from 6 billion to 15 billion tokens per minute between October and late March. Anthropic tightened five-hour session limits for Pro and Max users during weekday peak hours in late March, with roughly 7% of users hitting the new caps. These are not cosmetic adjustments; they reflect a fundamental supply-demand imbalance that flat-rate pricing cannot absorb.
An Industry-Wide Repricing Event
Anthropic is not alone. OpenAI shifted Codex from flat-message pricing to token metering in early April. GitHub tightened Copilot usage limits on April 10. Windsurf replaced its credit system with daily and weekly quotas in March. Industry analysts predict that every major AI provider running agentic workloads will move to usage-based enterprise billing within six months. The flat-fee era for enterprise AI appears to be ending.
AI Biz Insider Analysis ― Anthropic’s pricing pivot reveals the central tension in AI commercialization: the marginal cost of intelligence is not zero. As agentic workloads replace simple chat queries, a single enterprise developer session can consume 100x more tokens than a standard conversation. Flat-rate pricing effectively subsidized power users at the expense of margins. With ARR tripling to $30 billion and a $380 billion valuation to justify, Anthropic must align revenue to actual compute costs. The risk is churn to competitors, but with OpenAI and GitHub making identical moves, there is no cheaper alternative. The winners will be companies that optimize their AI consumption patterns; the losers will be those caught off guard by bills that could be “50 times higher” than expected. This is the cloud computing billing model applied to intelligence, and it changes the unit economics of every AI-powered workflow in the enterprise.
Related
- Project Glasswing: Anthropic Claude Mythos Preview and $100M Cybersecurity Alliance
- Hightouch Hits $100M ARR as AI Marketing Agents Drive 70% of Revenue
- Google Launches Native Gemini App for Mac with System-Wide AI Access
- TOP3 DIGEST: Anthropic Opus 4.7, OpenAI Trusted Access, and the Human 20%
Sources
- The Information — Anthropic Changes Pricing to Bill Firms Based on AI Use Amid Compute Crunch (Apr 14, 2026)
- PYMNTS — Anthropic Switches to Usage-Based Billing for Enterprise Customers (Apr 2026)
AI Biz Insider · AI Business EN · aibizinsider.com
댓글 남기기