OpenAI Considers Shifting Compute Needs to SoftBank - Report

OpenAI's Future Computing Infrastructure

Recent reports from The Information indicate that OpenAI anticipates a significant transformation in its sourcing of computing resources over the coming five years.

By the year 2030, the company projects that approximately 75% of its data center capacity will be supplied by Stargate, a venture poised to receive substantial funding from SoftBank, a recent investor in OpenAI.

A Shift Away From Microsoft

This represents a considerable departure from the current reliance on Microsoft, OpenAI’s largest shareholder, which presently provides the majority of the startup’s computational power.

However, this transition will be gradual. OpenAI intends to continue increasing its investment in data centers owned by Microsoft in the near future.

Projected Cost Increases

Concurrently, OpenAI’s overall expenditures are expected to rise substantially. The Information suggests that OpenAI forecasts a cash burn of $20 billion in 2027.

This figure is significantly higher than the estimated $5 billion consumed in 2024.

Inference Costs to Surpass Training

Furthermore, by 2030, OpenAI anticipates that the expenses associated with running AI models – a process known as inference – will exceed the costs related to the initial training of those models.

This shift highlights the growing importance of deploying and utilizing AI, rather than solely focusing on its development.