OpenAI’s decision to restructure ChatGPT Enterprise from a fixed-price model to a flexible, credit-based system took effect this week, giving corporate buyers granular control over how—and how much—they pay for advanced artificial intelligence features. The pivot lands just as the San Francisco firm claims a $10 billion annual recurring revenue run rate and finds itself sparring with longtime backer Microsoft over discounts that have swung big contracts OpenAI’s way.
From Flat Fees to Flexible Credits
For two years, OpenAI previously sold ChatGPT Enterprise at a fixed price. Executives now say that the model no longer fits a customer base whose usage varies widely by team and workload.
Under the new scheme, enterprise customers purchase a shared credit pool at the contract level. Every seat still enjoys unlimited access to core ChatGPT models—including GPT-4 and GPT-4.1-mini — credits kick in only when a user taps premium capabilities, such as large-context analysis or multimodal generation. Credits only apply to advanced models and features, not core ChatGPT options such as GPT-4o, GPT-4.1-mini, search, file upload, and canvas.
Customers buy those credits through their OpenAI Account Teams. Because pricing varies by volume, the company has not published a public rate card; however, procurement managers say that larger bundles unlock meaningful discounts.
How the Credit Pool Works
Finance administrators track consumption in real-time. Enterprise Workspace Owners can view their remaining credits and download usage reports through Settings > Billing and set red-line alerts to ensure bills never surprise the CFO. Credits remain usable for the duration of the agreement; they do not expire or rollover.
When a pool dries up, workspace owners can pay in arrears or pause premium-model access until fresh credits arrive. The platform sends emails to owners when thresholds are reached and allows them to configure usage alerts and hard caps to control runaway costs. For larger organizations, Role-Based Access Controls can limit which job roles may invoke specific models—a safeguard that keeps interns from burning through tokens intended for research scientists.
Discounts and Spend Controls
As part of the rollout, OpenAI offers 10% to 20% discounts on ChatGPT Enterprise subscriptions to customers who bundle the service with other OpenAI products or sign multi-year commitments. The company frames the move as a way to put generative AI “in every worker’s hands.” Yet, finance chiefs view it as a potent guardrail against over-adoption: the credits make budget planning straightforward, while deep discounts sweeten the decision to standardize.
Microsoft Tension Boils Over
That discounting strategy is where harmony with Redmond ends. Sources at both companies confirm that OpenAI’s aggressive concessions have created tension with Microsoft, which typically does not offer discounts on enterprise services. Azure sales reps, lacking a comparable subsidy, have reportedly lost deals because they cannot match OpenAI’s offers.
Emails reviewed by the Wall Street Journal show OpenAI leaders debating whether to accuse Microsoft of anticompetitive behavior as a “nuclear option” should the dispute escalate. Another sticking point is whether Microsoft would gain access to intellectual property from OpenAI’s acquisitions, including the $3 billion purchase of Windsurf, a coding startup. Both matters are part of a broader renegotiation of Microsoft’s equity stake and board influence following OpenAI’s late-2024 governance overhaul.
New Tools for Power Users
While pricing grabs headlines, the product itself continues to expand. ChatGPT Enterprise now includes Connectors that bring internal tools and content into conversations with real-time data fetches and inline citations. Supported connectors include Google Drive, SharePoint, Dropbox, Box, Outlook, Gmail, Google Calendar, Linear, GitHub, HubSpot, and Microsoft Teams.
The release also introduces Record Mode, which captures meetings, brainstorms, and voice notes with transcription and summarization. For now, Record Mode is available only on the macOS desktop app, supports up to 120-minute sessions, and deletes raw audio once transcription finishes. A “Deep research” workflow generates long-form, citation-rich answers by pulling from internal knowledge bases through those same connectors, and admins can build custom connectors using the Model Context Protocol.
Explosive Revenue Growth
OpenAI reached $10 billion in annual recurring revenue as of June 2025, almost doubling from roughly $5.5 billion just six months earlier. The figure excludes licensing income from Microsoft and other one-time deals, indicating that subscription sales alone can support the company’s research expenses. The user base is vast—more than 500 million people drop into ChatGPT weekly, and 3 million businesses now pay for the tool.
Executives peg the opportunity even higher, telling investors that the firm projects approximately $15 billion in revenue from business customers once the credit model reaches a steady state. Analysts note that the forecast assumes both an increase in per-seat spending and a new wave of AI adoption outside the tech industry’s traditional early adopter set.
What the Shift Means for the Market
OpenAI’s credit system follows a broader move among AI vendors—Anthropic and Cohere included—to link fees directly to compute-heavy actions instead of user count. The rationale is that heavy users pay more, light users get in at a lower cost, and finance teams see a clear return on investment (ROI). Yet, few rivals enjoy OpenAI’s scale, financial cushion, or brand recognition. Microsoft itself may need to revise its pricing on Azure AI Studio if it wants to stop further defections.
For buyers, the arithmetic looks simple. Finance teams can pin an upper limit on what the research department or marketing squad can burn in a quarter. Procurement can dangle discounts for volume purchases. Security leaders can sleep easier knowing RBAC fences limit the number of people who can push confidential data through large language models.
Looking Ahead
Negotiations with Microsoft will test CEO Sam Altman’s diplomacy. OpenAI relies on Azure for its infrastructure and on Windows’ dominance to integrate its desktop tools within Fortune 500 firewalls. However, it also needs to keep prices low enough to outbox nimble open-source challengers. The credit scheme is a balancing act: a lever to boost revenue while fending off rivals and placating cash-strapped customers. Whether the strategy succeeds will come down to how many credits those customers burn—and whether Microsoft decides to match terms rather than fight them.
Key Takeaways:
- OpenAI has shifted ChatGPT Enterprise to a credit-based model that charges only for premium AI tasks while core usage stays unlimited.
- Credits live at the workspace level, never expire during the contract, and come with granular controls, email alerts, and hard caps.
- Volume buyers get 10–20 % discounts, a move that has strained the partnership with Microsoft, which rarely cuts prices.
- New enterprise features, such as Connectors and Record Mode, aim to integrate companies more deeply into the OpenAI ecosystem.
- OpenAI says it now rings up $10 billion in recurring revenue and targets $15 billion from business users alone.
- The shake-up signals a wider industry pivot toward usage-based billing that could redraw the cloud-AI pricing map.