OpenAI Plans $225 Billion in AI Operating Costs and Integrates Sora into ChatGPT

OpenAI Plans $225 Billion in Inference Costs by 2030

The Information reports that OpenAI has internally projected that they will spend over $225 billion on inference — computing capacity to run AI models — between 2026 and 2030. Simultaneously, the company plans to integrate the video generator Sora directly into ChatGPT.

Astronomical Operating Costs

$225 billion over four years is approximately $56 billion per year. That's more than Norway's entire state budget. This is not investment in development — this is just the cost of running existing models for users.

The figures reveal a brutal truth about the AI industry: even for the world's leading AI company, margins are thin and capital requirements enormous.

Sora into ChatGPT — Video for the Masses

OpenAI is preparing to integrate its AI video generator Sora directly into the ChatGPT interface. Today, Sora is available as a separate app. The integration will mean that ChatGPT users can create video directly from the chat — without switching tools.

This is a direct response to Runway, Kling, and the recently paused ByteDance Seedance 2.0.

What Does This Mean for Enterprise Pricing?

With $225 billion in planned inference costs, API prices cannot go down forever. OpenAI needs either:

• Drastically more paying users
• Higher prices
• Technological breakthroughs that reduce costs

For enterprise customers planning large AI deployments: don't lock yourself to one vendor based on today's prices. The price of AI tokens will likely rise.

My take:

$225 billion in inference is not just a number — it's a signal that OpenAI and similar companies are in a monetization race against time. Sora into ChatGPT is not just a feature, it's a desperate maneuver to keep user numbers up and justify the next investment round. For those of us building on these platforms: have a cost model ready and avoid naive dependence on "cheap AI". Cheap is temporary.