Microsoft announces a multibillion-dollar extension of its OpenAI partnership

Background

Microsoft's partnership with OpenAI began in 2019, with a $1 billion investment and a plan to build a large Azure computing platform. A further investment followed in 2021. Early in 2023 the two companies were preparing the next step, as OpenAI's models were moving into commercial products.

What happened

On 23 January 2023 Microsoft announced the third phase of its long-term partnership with OpenAI. It was a multiyear, multibillion-dollar investment following its earlier investments in 2019 and 2021.^[1] Media reports put the investment at $10 billion. Microsoft's own announcement did not state an amount.^[2]

Microsoft said it would increase investment in specialised supercomputing systems for OpenAI's independent research. It also said it would continue building out Azure AI infrastructure.^[1]

Key details

Microsoft would deploy OpenAI's models across its consumer and enterprise products. Azure would be OpenAI's exclusive cloud provider, powering all OpenAI workloads across research, products and API services.^[1]

The announcement listed products already built on the partnership: Azure OpenAI Service, GitHub Copilot, DALL·E 2 and ChatGPT.^[1] The agreement let each company independently commercialise the resulting AI technologies.^[1]

A week earlier, on 17 January 2023, Azure OpenAI Service had become generally available with GPT-3.5, Codex and DALL·E 2. ChatGPT was promised soon.^[3]

Impact

The 2023 extension tied OpenAI's research and product work to Azure as its exclusive cloud. Products such as GitHub Copilot and ChatGPT were among those already built on the partnership at the time.^[1][3]

What the partnership covers

A partnership of this kind has several parts that can be changed separately. The money is one part, and it determines how much stake or influence each side has. The cloud arrangement is another, and it determines where OpenAI's models run and who earns from that usage. The product rights are a third, and they determine which company can sell what. The 2023 announcement touched all three, which is why it is described as an extension rather than a single payment.^[1]

Why the exclusive cloud mattered

Exclusive cloud provider means that other cloud companies do not host the workloads in question. For OpenAI, this meant that its research, its products and its API services all ran on Azure, so the capacity Microsoft built in its data centres became central to the company's work. For Microsoft, it meant that the models would be available to its customers through Azure first and most directly.^[1][2]

Context for readers

The announcement came as large language models were moving from research into everyday use. Readers can compare the 2023 extension with the 2019 start of the partnership, and with the later agreements in this archive, to see how the terms changed as the products grew.^[1][3]