OpenAI ditches fixed pricing—now devs pay per API call

OpenAI has quietly flipped the switch on Codex pricing for its ChatGPT business plans, replacing fixed licenses with a usage-based model. The move, buried in an update on The Decoder, positions OpenAI directly against GitHub Copilot and Cursor, which have long relied on seat-based pricing. For enterprise customers, the promise is simple: pay only for what you use. The reality, as always, is messier.

Usage-based pricing isn’t new—it’s been the default for most cloud APIs for over a decade. What’s notable here is OpenAI’s timing: just as Copilot’s enterprise adoption hits critical mass, the company is betting that developers will prefer metered billing over fixed costs. The question is whether this reduces spend or merely redistributes it. Early signals suggest that for high-volume users, the math could tilt in OpenAI’s favor—especially if usage scales unpredictably during peak coding sprints.

The bigger story isn’t the pricing change itself, but what it reveals about OpenAI’s strategy. By targeting Copilot’s core market, OpenAI is signaling that it sees developer tools as a battleground worth fighting for. The move also reflects a broader trend in AI: vendors are realizing that fixed pricing struggles to capture the value of models that improve over time. Yet, for teams already invested in Copilot’s workflows, the switch may feel less like an opportunity and more like a forced migration.

Where this gets interesting is in the competitive dynamics. GitHub Copilot, which has been aggressively expanding its enterprise features, suddenly faces a pricing model that could undercut its seat-based revenue. Cursor, the scrappy upstart, may see its growth trajectory disrupted as OpenAI muscles in with a more flexible (and potentially cheaper) alternative. But flexibility comes with trade-offs: usage-based pricing can introduce unpredictability, particularly for teams operating under strict budgets.

For developers, the shift is a mixed bag. On one hand, smaller teams or those with sporadic usage could see genuine cost savings. On the other, enterprises with consistent workloads might end up paying more—especially if OpenAI tweaks its rate card down the line. The lack of grandfathering for existing customers suggests this isn’t just a pricing experiment but a deliberate pivot.

The real tell will be how GitHub and Cursor respond. Copilot has already been testing usage-based tiers, but its parent company, Microsoft, has deeper pockets and more experience in enterprise lock-in. Cursor, meanwhile, lacks the scale to compete on price but could double down on niche features or open-source alternatives. One thing is clear: the days of fixed-price AI coding tools are numbered, and the winners will be the ones who can turn metered billing into a selling point rather than a gotcha.