ETR asked 36 technology leaders why their OpenAI spend is moving. Twenty-two plan to raise it by 6% or more over the next six months, and 14 plan to cut it by the same margin. The split matters less than the pattern underneath it. Growth is coming from new users and new budget, while the losses land almost entirely with one rival, and coding is the most contested workload.
Expansion is additive. Only three of 22 increasers said their added OpenAI spend came at the direct expense of another product or legacy project. The rest describe new budget, more users, and more use cases.
Anthropic is absorbing the cuts. Twelve of 14 decreasers named Anthropic or Claude as a destination for spend leaving OpenAI.
Coding is the contested workload. Three decreasers tie their cuts to code generation or developer workflow fit, while one increaser credits Codex for OpenAI's improving position.
Not all churn is permanent. One decreaser calls its switch a short-term call that could reverse on price or product, while bundled suites pose a stickier threat.
The larger accounts describe OpenAI as embedded in long-range plans. A senior practitioner at a Fortune 500 manufacturer called it "heavily woven into most future strategies." Financial services and insurance firms made up five of the 22 increasers, citing efficiency targets and, in one case, a common agentic layer built while decommissioning technical debt and legacy robotic process automation.
Expansion is not exclusivity, though. An architect at a Fortune 100 IT/TelCo firm is growing OpenAI spend while keeping a rival as the primary vendor: "Their models are improving, especially with Codex. Anthropic is still the leader." A small IT/TelCo executive put the multi-vendor logic plainly: "The different providers and models are better at different aspects at different times."
Even committed buyers are hedging. "Things are changing all the time in AI, who knows where OpenAI will be in 6 months," said a director at a large services firm.
The decreaser group tells a more concentrated story. Twelve of 14 named Anthropic or Claude as a beneficiary, and three tied the move directly to code generation or developer workflows. An advisor at a midsize healthcare company described a near one-for-one swap: "Claude is demonstrably better at that task," with spend shifting "just about dollar for dollar."
Others cited depth of analysis for security operations work, models they found more grounded, and in one case ethical concerns. Leadership changes mattered too. "New CFO is more aligned with Claude than OpenAI," said a VP at a midsize financial services firm, which now expects to treat OpenAI as a second-tier vendor.
Not every loss is about model quality. Two decreasers are moving to Microsoft and Google tools that fit stacks or contracts they already have. That is a procurement decision, not a benchmark result, and it is harder for any standalone model vendor to counter. "Unless OpenAI offers substantial benefits that cannot be realized through other products that we already license, we will continue to limit its use to help contain costs," said a C-level executive at a midsize education organization. Cost pressure surfaced in five decreaser accounts overall.
Company size also separates the groups. Six of 14 decreasers are midsize organizations, compared with two of 22 increasers. Seven increasers sit in the Fortune or Global 1000, versus two decreasers.
Three signals stand out for anyone sizing the model layer.
First, watch the coding workload. It is the one workload decreasers named specifically, and also where at least one large buyer sees improvement. Whether Codex changes that is the thread to follow.
Second, separate reversible churn from structural churn. One education leader called the switch to Claude "a short term decision" that "could easily change as prices change and products improve." Losses driven by bundled contracts or new executive alignment look harder to win back.
Third, budgets are still additive, but that may not last. Another small IT/TelCo leader expects that "over time we may consolidate some overlapping AI capabilities if OpenAI provides better functionality, integration, or value." If consolidation takes hold, today's multi-vendor hedging turns into a contest with clear winners and losers.
ETR designs its Edge Reports to capture both spend groups, so the 22-to-14 split is not a read on the broader market. The value is in the reasoning behind each decision. Request access to the Edge Report on OpenAI to read every account's rationale, including exactly which platforms are capturing displaced spend, by contacting our sales team.