Microsoft considers suing over Amazon-OpenAI $50B cloud deal

Microsoft is weighing legal action against Amazon and OpenAI over a $50bn deal that could breach its exclusive cloud partnership with the ChatGPT maker, setting up a clash between the Big Tech rivals.

The dispute centres on whether Amazon Web Services can offer OpenAI’s new commercial product, known as Frontier, without violating a longstanding agreement that requires all access to the start-up’s models to be routed through Microsoft’s Azure cloud platform, News.Az reports, citing The Finanial Times.

The arrangement is highly lucrative for Microsoft, with OpenAI’s products helping drive Azure revenues to record highs.

Amazon and OpenAI say they are building a system that works around the contract. Microsoft executives dispute this, saying the approach is not feasible and would violate the spirit, if not the letter, of their agreement, according to people familiar with the discussions.

Ahead of Frontier’s launch, the companies were still in talks to resolve the dispute without litigation, they added.

“We know our contract,” said a person familiar with Microsoft’s position. “We will sue them if they breach it. If Amazon and OpenAI want to take a bet on the creativity of their contractual lawyers, I would back us, not them.”

The legal threat underscores a broader rift between Microsoft and OpenAI as the start-up pushes to loosen the constraints of its early contracts and diversify its cloud partnerships while its biggest backer increasingly views it as a competitor in enterprise AI services.

OpenAI believes its plans with Amazon are compatible with its deal with Microsoft, according to a person familiar with its positions.

The person added that Microsoft was unlikely to pursue legal action and invite further scrutiny while it is facing regulatory probes in the US, UK and EU into its alleged anti-competitive licensing practices with Azure.

OpenAI’s plans for a public listing as early as this year could be derailed if the dispute ends up in court. Even after closing a $110bn funding round last month, it needs to raise more cash to pay for the vast computing resources needed to train and run its large language models.

The IPO is already complicated by a lawsuit filed by Elon Musk against chief executive Sam Altman, with whom he co-founded the start-up in 2015. The world’s richest man is accusing Altman of abandoning its non-profit mission to enrich himself and other executives, with a trial scheduled to start in Oakland next month.

“The last thing OpenAI needs is another court case right now,” said the person familiar with Microsoft’s position.

OpenAI’s Frontier deploys fleets of AI agents — bots that can operate independently under human instructions — within businesses. The platform is the centrepiece of the OpenAI-Amazon partnership announced last month, alongside a pledge to buy $138bn in cloud services from AWS.

Microsoft had been OpenAI’s exclusive cloud provider since investing $1bn in the start-up in 2019 but gave up that right when it signed off on its restructuring in October.

However, it retained a clause covering application programming interfaces (APIs), the connections developers and businesses use to access OpenAI’s models. The clause requires all API calls to be routed through Azure.

Behind the scenes, the companies’ lawyers clashed for weeks over the scope of Amazon’s agreement and how it could be described, according to two people familiar with the matter.

When the three groups released parallel statements describing the Frontier product, Microsoft asserted that nothing had changed from the October agreement and it remained the exclusive cloud provider for OpenAI APIs.

The dispute hinges on the definition of “stateless” and “stateful” access to AI models. Large language models are “stateless” by default, retaining no information between user interactions. “Stateful” layers are added via applications to give them memory and context, which makes them more useful for businesses.

Amazon and OpenAI are developing a system known as a “Stateful Runtime Environment” that runs in Amazon’s Bedrock AI platform. The system would access company data stored on AWS, allowing OpenAI agents to remember prior work, operate across software tools and data sources, and access computing power.

Those claims concerned Microsoft because its experts do not believe the technology exists to avoid running Frontier on Azure under the terms of its contract, the people said.

To avoid provoking Microsoft, Amazon has given staff strict guidance on how to describe the SRE, according to an internal memo seen by the FT and first reported by Business Insider.

AWS employees may tell customers that SRE is “powered by”, “enabled by” or “integrates with” OpenAI, but are banned from saying that SRE “enables access” or “calls on” ChatGPT. Staff should also not suggest that OpenAI’s most advanced frontier models are available on AWS.

OpenAI maintains that its Amazon deal does not allow backdoor access to its stateless models, according to the person familiar with its position. The start-up has the right to make new products with third parties so long as they do not cross the “red line” of being primarily offered as an API.

Amazon and OpenAI declined to comment. Microsoft said: “We are confident that OpenAI understands and respects the importance of living up to [its] legal obligation.”

News.Az 

By Ulviyya Salmanli