OpenAI Quietly Tests Pay-Per-Outcome Pricing for Enterprise Customers
OpenAI has quietly begun offering select enterprise customers a new pricing model that charges only when AI completes tasks like customer service, according to The Information, citing a person with direct knowledge. The move signals that pay-per-outcome pricing is expanding beyond AI startups, with Sierra, Salesforce (via its $3.6B Fin acquisition), and Cognition already adopting or promising similar models. An OpenAI spokesperson declined to comment.
Coverage timeline
机器之心机器之心
如果有一天，OpenAI 告诉企业：AI 没把事情办成，这笔钱你可以不付，会发生什么？ 这件事已经开始了。据 The Information 援引一名直接知情人士称，过去几个月， OpenAI 已经悄悄向部分大客户提供了一种新的付费方式：只有当 AI 真正完成客户服务等任务时，企业才需要付钱。 不过，OpenAI 发言人拒绝对此置评。 看起来只是换了一种计费方式，背后其实牵动着整个软件行业过去二十多年形成的商业模式。 过去买软件很好算账：一个员工一个账号，一个月多少钱，企业根据人数购买订阅就行。到了生成式 AI 时代，这套规则越来越不好用了。一个 Agent 可能一分钟回答一个问题，也可能连续工作几个小时，替几个人完成一整套流程；更麻烦的是，模型调用量越来越高，并不代表给企业创造的价值也越来越高。 所以，越来越多公司开始问一个更直接的问题：既然 AI 号称能够干活，为什么不等它真的把活干完以后再收费？ AI 卖得越来越多，软件公司却不知道该怎么收费了 软件行业过去两年其实一直在摸索这个问题。 最容易想到的办法是按使用量收费，比如 Token、调用次数或者完成多少次任务。这种方式至少比传统的按账号订阅更贴近 AI 的成本结构，但很快又出现另一个问题：使用得多，不一定意味着价值高。 比如，一个 Agent 打了 1000 通电话，和它最终促成了多少销售，显然不是一回事。一家公司的工程师每天调用几亿 Token，也不能说明这些 Token 最终给公司赚了多少钱。 所以，用使用量计费，企业很容易觉得自己的钱花得不明不白。他们真正关心的，还是投资回报。 这也是为什么「按结果收费」正在迅速进入 AI 公司的定价表。客户管理公司 Sierra 就采用类似模式：只有 AI 在没有人工介入的情况下真正完成任务，客户才付费。Salesforce 正在以 36 亿美元收购的 Fin，同样按照 AI 成功解决的问题收费。 编程智能体公司 Cognition 走得更远。它向企业客户承诺，如果最终交付的工程成果价值达不到客户支付的费用，公司最高可以提供 1000 万美元的 credits。Adobe、HubSpot 和 Zendesk 等老牌软件公司，也已经开始朝着「AI 真干完了活再收钱」的方向调整。 OpenAI 如今加入其中，意味着这种收费方式已经不只是 AI 创业公司的尝试。 类似的探
Related stories
OpenAI's ChatGPT Ads hits $1B annualized revenue; outcome-based pricing tested
OpenAI announced that its ChatGPT Ads business has reached a $1 billion annualized revenue run rate and is expanding globally, touting a diversified business model ahead of a potential IPO. Separately, The Information reports that OpenAI has begun offering some major customers the option to pay only when its AI completes tasks, joining Salesforce and other providers in testing outcome-based pricing.
3 reports · 2 sources
OpenAI on track to exceed $40B annualized revenue; enterprise overtakes consumer
OpenAI is on track to generate annualized revenue of more than $40 billion, roughly doubling its run rate from the end of 2025, according to sources. CFO Sarah Friar told investors that the enterprise business now generates more revenue than the ChatGPT-led consumer segment, with enterprise customers growing 32% in July.
OpenAI previews Private Safety Processing for zero-data-retention misuse monitoring
OpenAI announced it is testing Private Safety Processing, an automated system that monitors for potential misuse of its frontier models while retaining none of the customer's data, preserving zero data retention protections for eligible API customers. The preview, offered to select customers, contrasts with Anthropic's recently announced data-retention policy, which keeps user data for 30 days for covered models.
4 reports · 3 sources
OpenAI Q2 revenue up 18% QoQ to $6.7B; Anthropic more than doubles to $11.6B
According to Wall Street Journal sources, OpenAI's Q2 revenue grew 18% quarter-over-quarter to $6.7 billion, but its operating margin continued to decline. In the same period, Anthropic more than doubled its revenue to $11.6 billion. New data also indicates OpenAI is gaining ground on Anthropic among business users, highlighting volatility in enterprise AI spending.
News updates every hour. Subscribe in your RSS reader to get new stories automatically.
Subscribe