The news
OpenAI and Anthropic are encountering rising price resistance from major customers. AI-powered software startups are increasingly turning to cheaper open models instead of paying the rates charged by the two leading closed providers.
Context
The shift comes as companies that build products on top of large language models look for ways to lower their operating costs. Previously, many of these startups relied on API access from OpenAI and Anthropic for core capabilities. The move to open alternatives reduces that dependency and directly affects revenue at the closed-model companies.
Bloomberg Technology reported the trend on October 6, 2026. The account centers on “big AI users” who now treat inference cost as a controllable line item rather than a fixed expense.
Details
The Bloomberg report states that AI-powered software startups are increasingly turning to cheaper open models. No specific pricing figures or customer names appear in the report. The pressure is described as coming from high-volume users who have found viable lower-cost options.
The article does not name the open models gaining share, nor does it quantify the volume of workloads that have already moved. It also does not indicate whether OpenAI or Anthropic have adjusted prices or introduced new tiers in response.
Reactions and counterpoints
The single source available offers no statements from OpenAI or Anthropic executives. It likewise contains no comments from the startups that have switched or from vendors of the open models now in use. Without additional reporting, it is not possible to determine whether the two closed providers view the shift as temporary or structural.
Why it matters
For companies whose primary product is access to frontier models, customer willingness to pay is the central variable. When a meaningful share of high-volume users finds acceptable substitutes at lower cost, the pricing power that supported earlier valuations weakens. The economics are straightforward: inference spend scales with usage, so even modest per-token savings compound quickly for any startup that serves real traffic.
Startups building AI products operate under tight unit economics. A sustained difference in inference cost changes which provider they choose and how much they can afford to spend on other parts of the business, from sales to model fine-tuning. If open models close the gap on the tasks that matter most to these customers, the closed providers lose their strongest argument for premium rates.
OpenAI and Anthropic must therefore demonstrate clear performance or reliability advantages that justify their rates, or accept lower revenue per customer. The Bloomberg account gives no sign that either company has yet reversed the trend. Absent new data showing that performance gaps remain decisive for the workloads now moving, the pressure on margins is likely to continue.
The development also affects how investors value the two companies. Revenue forecasts built on the assumption that closed frontier models would retain most high-volume usage now face a direct test. If the pattern holds, growth in API revenue may slow even as overall AI adoption rises, because a larger fraction of that adoption will run on lower-priced alternatives.
For the startups making the switch, the choice is primarily financial rather than ideological. They are selecting the lowest cost option that still meets their product requirements. That calculation favors open models whenever the quality delta does not translate into measurable user-facing improvement. Over time, this dynamic rewards open-model providers that can maintain competitive capability at commodity prices and penalizes closed providers that cannot.
The Bloomberg report does not forecast how far the shift will extend. It simply records that the migration has begun among the customers whose spend matters most to OpenAI and Anthropic. That observation alone is enough to alter expectations about the durability of current pricing levels.
---
Sources:
No comments yet