In a twist that few would have predicted just a few years ago, Microsoft is no longer the sole provider of data centre infrastructure powering OpenAI’s ambitious AI models. It is a significant shift for an industry where Microsoft’s cloud, Azure, once appeared to be OpenAI’s permanent home base.
Under the newly unveiled “Stargate” deal, OpenAI’s sweeping AI infrastructure agreement with tech giants like SoftBank and Oracle, Microsoft has signed an updated contract that grants it “right of first refusal” on any fresh OpenAI cloud requirements. In simpler terms, Microsoft gets the first crack at hosting OpenAI’s expanding AI workloads. If the software giant cannot keep up, OpenAI now has the freedom to take its business elsewhere.
According to a Microsoft blog post, OpenAI has recently made a “large Azure commitment” intended to bolster its products and training activities. Moreover, Microsoft has greenlit OpenAI to establish additional capacity, with a focus on research and training for its growing suite of models.
For its part, OpenAI has frequently cited limited compute resources as a stumbling block, slowing the release of new AI powered offerings. This crunch has reportedly caused friction between OpenAI and Microsoft, ironic, given that Microsoft is both a key partner and one of OpenAI’s biggest backers.
Still, Microsoft is eager to reassure observers that the bedrock of its long running partnership with OpenAI remains intact until 2030. That includes Microsoft’s access to OpenAI’s intellectual property, various revenue sharing deals, and exclusivity over OpenAI’s application programming interfaces, APIs.
However, there is one massive asterisk in these arrangements: if OpenAI attains artificial general intelligence, AGI, as the two companies define it, essentially AI able to rake in at least AUD$100 billion in profit, Microsoft will no longer have access to OpenAI’s technology, according to reports. In other words, the moment the technology crosses a jaw dropping profitability threshold, the current deal unravels.
“The OpenAI API is exclusive to Azure, runs on Azure and is also available through the Azure OpenAI Service,” Microsoft wrote. “This agreement means customers benefit from having access to leading models on Microsoft platforms and directly from OpenAI.”
All told, this latest arrangement hints at the intense jockeying under way among tech heavyweights in the scramble to supply the computational backbone for tomorrow’s most advanced AI systems. As OpenAI’s needs grow ever larger, and with Microsoft eager to hold onto its golden goose, only time will tell if these new terms truly satisfy both parties.
HARNIS specialises in the safe, efficient and ethical implementation and usage of AI by business.
If you’d like to reach out to discuss the implementation of AI in your business, please reach out to us at hello@harnis.ai
