OpenAI pivots on enterprise
An internal memo outlines OpenAI's effort to reduce dependence on Microsoft, lean into an alliance with Amazon, and ship a new model codenamed “Spud” that the company says will boost its products — and it has published formal rate cards for business and enterprise plans. Those moves sit alongside vendors packaging governance tooling for regulated customers, with new product announcements aimed at moving AI from pilots into auditable production ( ).
OpenAI is reshaping its business around large corporate customers, pairing a new Amazon alliance with formal ChatGPT pricing for business buyers. (cnbc.com, help.openai.com) In a staff memo viewed by CNBC and The Verge, chief revenue officer Denise Dresser said Microsoft had been “foundational” but had also limited OpenAI’s ability to meet customers on Amazon Web Services’ Bedrock platform. She said customer demand since the Amazon deal was announced in late February had been “frankly staggering.” (cnbc.com, theverge.com) Amazon said in February it planned to invest up to $50 billion in OpenAI, while Microsoft has invested more than $13 billion in the company since 2019. Dresser told CNBC earlier this month that enterprise now contributes 40% of OpenAI revenue and is on track to reach parity with its consumer business by the end of 2026. (cnbc.com) OpenAI’s own help pages now read more like software procurement documents than consumer marketing. The company published a rate card for ChatGPT Business, Enterprise, and Edu plans and, on April 2, added two seat types for Business and Enterprise: a standard ChatGPT seat and a new Codex-only seat. (help.openai.com, help.openai.com) That rate card puts explicit usage prices on higher-cost features inside ChatGPT workplaces: GPT-5.4 Thinking at 10 credits per message, GPT-5.4 Pro at 50 credits, Agent at 30 credits per message, Deep Research at 50 credits per task, Images at 5 credits per generation, and Voice at 5 credits per minute. Business customers get per-seat limits first, while Enterprise and Edu customers draw from a shared contract-level credit pool. (help.openai.com, help.openai.com) The memo also points to a coming model codenamed “Spud.” The Decoder, citing the leaked memo first reported by The Verge, said Dresser described it as an “important step” for the next generation of work, with early customer feedback pointing to stronger reasoning, better handling of intentions and dependencies, and more reliable production output. (theverge.com, the-decoder.com) The sales pitch has shifted with the market. Dresser wrote that enterprise artificial intelligence is entering a “more mature phase,” where customers care less about benchmark wins and more about whether models fit into workflows, controls, and day-to-day operations. (the-decoder.com, theverge.com) That is where governance vendors are trying to attach themselves to the spending wave. Charli AI said on April 14 that its Version 6.5 adds a “Governance Control Plane” for financial institutions, and the company says its platform sits above existing models to make decisions traceable, reproducible, and policy-aligned. (markets.businessinsider.com, charliai.com) Charli is describing the same bottleneck from the buyer side that OpenAI is describing from the seller side: regulated companies cannot move a chatbot into production if every answer changes and no audit trail shows how it was produced. Its website says standard generative systems remain stuck in experimental workflows in finance because they are not deterministic enough for sign-off. (charliai.com) OpenAI is still selling intelligence, but the enterprise push now turns on cloud access, seat design, usage accounting, and audit controls. The next test is whether that package can turn “staggering” demand into long contracts and systems companies will actually run in production. (cnbc.com, help.openai.com, charliai.com)
