🧠 Food for thought

Implications, context, and why it matters.

Leaked figures suggest huge cash needs

• A reported US$50 billion spending figure suggests OpenAI is losing money while costs climb fast. Leaked documents say computing bills may grow faster than revenue 1.
• In the first nine months of 2025, OpenAI reportedly spent US$8.65 billion on inference, running AI models for users. Revenue was estimated at at least US$4.33 billion, based on payments to Microsoft 1.
• One analysis projected cumulative negative cash flow, meaning more cash leaving than coming in, of US$143 billion from 2024 through 2029 2.
• Most of that cash burn comes from inference. Training costs were described as mostly non-cash, paid with credits Microsoft gave OpenAI as part of its investment 1.

Compute strain is changing the Microsoft-OpenAI tie-up

• High computing costs plus graphics processing unit (GPU) shortages helped push OpenAI beyond only Microsoft Azure, Microsoft’s cloud computing platform. Revised terms support a multi-cloud setup 3.
• The new agreement could let OpenAI deploy models directly on Amazon Web Services, Amazon’s cloud division, and Google Cloud. That gives it access to more chips and relieves GPU bottlenecks that had slowed growth on Azure alone 3.
• Microsoft accepted terms that end Azure exclusivity. It remains OpenAI’s primary strategic partner. Separate commentary has put Microsoft’s stake at about 27% 3.
• OpenAI also agreed last October to buy US$250 billion in Microsoft cloud services. Microsoft secured non-exclusive rights to license OpenAI models through 2032 3.