OpenAI Says 1 Million Businesses Now Pay for Its AI

OpenAI announced today that more than 1 million business customers worldwide now pay the company directly for AI — a figure it calls the fastest-growing business platform in history. The count includes every organization paying OpenAI for business use, whether through ChatGPT for Work subscriptions or direct model consumption via its developer platform.

The announcement, published on OpenAI's official site, arrives as competition for enterprise AI spending intensifies among Microsoft, Google, and Anthropic. OpenAI is using the milestone to argue that its technology has crossed from experimentation into production across financial services, healthcare, retail, and other industries — and to preview where it plans to take the platform next.

Consumer scale feeds enterprise adoption

OpenAI attributes part of its enterprise momentum to a consumer base of more than 800 million weekly ChatGPT users. Because employees already know the product, the company says, business pilots run shorter and rollouts face less friction.

The numbers behind that claim are concrete. ChatGPT for Work now has more than 7 million total seats, up 40% in just two months. ChatGPT Enterprise seats specifically have grown 9x year-over-year. Those growth rates matter because seat expansion, not just logo count, is what converts consumer familiarity into durable enterprise revenue.

A new wave of tools for full-scale deployment

To support that acceleration, OpenAI has shipped a set of products aimed at moving teams from early experimentation to companywide deployment.

The first is company knowledge, a capability that lets ChatGPT reason across workplace tools including Slack, SharePoint, Google Drive, and GitHub. It answers questions, performs analysis, and takes action using a version of GPT-5 optimized for tool use and citations. In practical terms, this positions ChatGPT as a layer over a company's existing software stack rather than a standalone chat window.

Codex, OpenAI's model for code generation, refactoring, and workflow automation, has seen usage grow 10x since August. OpenAI cites Cisco as a flagship deployment: the networking giant rolled Codex into its engineering workflows and cut code review times by 50%, shrinking project timelines from weeks to days.

AgentKit, launched to make enterprise agents practical to build and deploy, lets teams go from idea to production in days instead of months, according to OpenAI. The private equity firm Carlyle reports the AgentKit evaluation platform cut development time on its multi-agent due diligence framework by more than 50% and improved agent accuracy by 30%.

OpenAI also points to its multimodal portfolio — the Image Generation API, Sora 2 for video, and gpt-realtime with the Realtime API for production voice agents — as evidence that teams can now work across text, images, video, and audio in one system.

The ROI question

Whether enterprise AI generates returns remains contested among researchers, with several academic studies finding limited measurable productivity gains. OpenAI counters with a recent Wharton study reporting that 75% of enterprises see a positive ROI from AI, while fewer than 5% report a negative return. The company acknowledges other studies reach different conclusions but says the Wharton findings match what it observes with its own customers: when AI is deployed with the right use case and infrastructure, teams see real results.

The announcement names four customers with specific outcomes:

• Indeed uses OpenAI APIs in its Invite to Apply feature, driving a 20% increase in applications and a 13% uplift in downstream hiring success.
• Lowe's equips associates in more than 1,700 stores with Mylow Companion, an in-store app built on OpenAI models that provides expert project guidance.
• Intercom built its customer service agent Fin on OpenAI models and accelerated development cycles from quarters to days.
• Databricks, starting today, is making OpenAI frontier models natively available on its platform — bringing the models to where enterprise data already lives and, in OpenAI's words, making it easier to build and run high-quality agents.

The Databricks integration is notable for a different reason: it signals OpenAI pushing its models into third-party data platforms rather than insisting businesses move data to OpenAI's cloud, a shift in posture that matters for enterprises with strict data-residency and governance requirements.

From using AI to building on it

Beyond internal adoption, OpenAI identifies a second trend: businesses increasingly want to build on its platform rather than simply use its products.

Canva, Figma, Zillow, and Spotify have plugged their apps directly into ChatGPT to meet users where they already are. On the commerce side, Shopify, Etsy, Walmart, PayPal, and Salesforce are building shopping experiences through the Agentic Commerce Protocol (ACP) in ChatGPT, aiming to bring conversational commerce into everyday purchasing decisions.

That dual strategy — OpenAI as both a workplace tool and a distribution channel for third-party apps — puts the company in direct competition with Apple and Google's app-store economics and with Amazon's commerce funnel. If ChatGPT's 800 million weekly users increasingly transact and work inside the product, the platform becomes an operating layer for both consumer and business activity.

OpenAI frames the ambition exactly that way. "There's a big opportunity to rethink the operating system for work, and we're excited to build the platform to power it," the company wrote, adding that it is looking ahead to what comes next in 2026 and beyond.

Why the milestone matters

The 1 million customer figure gives OpenAI a defensible answer to the two questions that have shadowed enterprise AI: are businesses actually paying, and are they seeing returns? Seat growth of 40% in two months and 9x year-over-year growth in ChatGPT Enterprise suggest the paying base is expanding quickly, not just churning through trials.

The stakes extend beyond OpenAI's own revenue. Microsoft, which has built its Copilot strategy on OpenAI's models, Google with Gemini, and Anthropic with Claude are all chasing the same budget line. The customer proof points OpenAI cites — Cisco, Carlyle, Lowe's, Intercom, Databricks — are names every competitor will now have to answer in sales conversations.

The next test is whether OpenAI can convert deployment breadth into depth: more seats per customer, more mission-critical workloads, and more third-party builders choosing its platform over rivals. The company's stated roadmap — company knowledge, agents, and agentic commerce — signals it intends to be the substrate for how work gets done, not just a chatbot vendor. Whether the Wharton-era ROI numbers hold as deployments move from pilots to core operations will define the next phase of the enterprise AI market.