OpenAI shifts toward paid, enterprise workflows
OpenAI introduced a new $100 developer tier aimed at teams hitting free-tier limits, signaling more commercial pricing for heavy development usage. (thenewstack.io) The company also says enterprise AI already makes up roughly 40% of its revenue as it pushes agentic, workflow‑embedded assistants rather than simple copilots. (decrypt.co) Separately, reporting suggests OpenAI is developing a cybersecurity-focused model whose release is being constrained by safety considerations, highlighting governance friction as these tools grow more powerful. (axios.com)
OpenAI just put a price tag on a problem that had been piling up for months: developers were running into coding limits on the cheap plan, and the new answer is a $100-a-month ChatGPT Pro tier with five times the Codex usage of ChatGPT Plus. OpenAI told users the $20 plan stays aimed at “steady, day-to-day” coding, while the new tier is for heavier daily use. (venturebeat.com) (engadget.com) That price slot matters because OpenAI already had a much more expensive Pro offering, so $100 is not the top shelf. It is a middle rung for teams that have outgrown casual use but are not ready to buy a full enterprise contract. (venturebeat.com) (engadget.com) At the same time, OpenAI is saying the center of gravity in its business is moving away from consumer chat. In a company post published this week, OpenAI said enterprise now makes up more than 40% of its revenue and could reach parity with consumer revenue by the end of 2026. (openai.com) (decrypt.co) The company also gave a scale number that helps explain the push: Codex has reached 3 million weekly active users, and OpenAI’s application programming interfaces now process more than 15 billion tokens per minute. Those are the kinds of volumes that turn a consumer product into an infrastructure business. (openai.com) OpenAI is not pitching those business customers on a smarter autocomplete box. Its sales message is “agentic workflows,” which means software that can take a goal like resolving a support ticket or updating a spreadsheet and carry out multiple steps inside the tools a company already uses. (openai.com) (decrypt.co) That is why the company keeps talking about names like Goldman Sachs, State Farm, DoorDash, Thermo Fisher, and LY Corporation. Those are not app-store style users buying a novelty feature; they are large organizations plugging OpenAI models into finance, insurance, logistics, and lab work. (openai.com) The new $100 tier fits that strategy because it catches a team before it becomes a procurement project. A startup or internal engineering group can spend more than $20 a month the moment coding becomes a daily workflow, without waiting for legal review, security review, and a custom enterprise deal. (venturebeat.com) (engadget.com) Then there is the other half of the story: the more useful these systems get inside real workflows, the touchier their release decisions become. Axios reported on April 9 that OpenAI is finalizing a cybersecurity-focused product with advanced hacking capabilities and plans to release it only to a small set of partners. (axios.com) That limited rollout is not a side note. Axios reported that model makers now worry their own systems can cross a threshold where they are good enough at offensive cyber tasks that wide release creates obvious abuse risks, so capability and distribution are being separated. (axios.com) (red.anthropic.com) So OpenAI’s week adds up to a simple shift in shape. The company is charging more where usage is heavy, saying business customers already supply more than 40% of revenue, and treating its most powerful tools less like mass-market apps and more like controlled industrial equipment. (openai.com) (venturebeat.com) (axios.com)
