Insiders Claim OpenAI Buried Under Microsoft Compute Fees

The latest leaked documents reveal that Microsoft’s Azure arm collected roughly $865.8 million from OpenAI in the first three quarters of 2025, a steep rise from $493.8 million the previous year. This figure reflects the 20 percent revenue‑share clause tied to Microsoft’s $13 billion total investment, including the original $10 billion infusion that helped launch ChatGPT. Analysts estimate OpenAI’s total revenue to be at least $4.3 billion for the same period, suggesting the AI firm is scaling rapidly while still under intense financial scrutiny ahead of a potential IPO.

Compute costs dominate OpenAI’s expense profile. Inference spending alone jumped from about $3.8 billion in 2023 to $8.65 billion in the first nine months of 2025, underscoring the cash‑intensive nature of serving billions of model calls. Historically, Azure supplied the bulk of this compute, but recent agreements with CoreWeave, Oracle, AWS, and Google Cloud indicate a diversification strategy to manage capacity and pricing pressures. While much of the training budget is covered by Microsoft‑issued credits, inference remains a largely cash‑based outlay, tightening the margin between revenue and operating costs.

These financial dynamics fuel the broader debate over an AI bubble. If OpenAI’s models are still operating at a loss despite soaring revenue projections—such as Sam Altman’s claim of a $20 billion annualized run rate and a $100 billion target by 2026—investors must reassess valuation assumptions across the sector. The opaque reporting of Microsoft’s own Azure‑OpenAI earnings adds further uncertainty. Nonetheless, the partnership’s scale and the relentless demand for generative AI services suggest that both companies will continue to shape market expectations, even as they navigate profitability challenges and regulatory scrutiny.