“We Just Need More Power”: OpenAI Will Gradually Overcome Its Dependence on Microsoft

OpenAI currently relies heavily on the computing power of its major shareholder Microsoft to develop new AI models and provide AI services to companies and users. But that is set to change significantly over the next five years, The Information reported on Friday.

Image source: Growtika/Unsplash.com

By 2030, OpenAI expects to get three-quarters of its data center computing power from Stargate, which will be largely funded by SoftBank, one of OpenAI’s new investors, according to The Information. That marks a significant shift in OpenAI’s relationship with Microsoft, which has invested nearly $14 billion in the company since 2019.

The change won’t happen overnight, The Information writes. Over the next few years, OpenAI will continue to increase its spending on Microsoft-owned data center computing power.

According to The Information, OpenAI plans to spend $20 billion on its developments in 2027. For comparison, the company’s expenses in 2024 amounted to $5 billion. According to OpenAI’s forecasts, by 2030, its costs for the operation of already trained AI models (inference) will exceed the costs of their training.

When Stargate was announced, one X user noted that the friendship between OpenAI and Microsoft was over. In response, OpenAI CEO Sam Altman said, “Not at all! This is a very important and large partnership for the long term.” He added, “We just need more computing power.”

Microsoft, for its part, says the relationship between the two companies remains strong. It will continue to host OpenAI services on its Azure cloud computing platform. Microsoft has also invested in its French competitor Mistral AI and offers Meta✴ Llama models on the Azure platform. Last year, Microsoft added support for OpenAI competitor Anthropic to its GitHub Copilot AI assistant, which helps programmers write code.