Microsoft and OpenAI have announced a pivotal update to their partnership agreement. This change, revealed on Tuesday, allows OpenAI to tap into computing resources from competitors, a move that could reshape the dynamics of AI development and deployment in the industry.

The new deal

The core of this new arrangement involves altering the exclusivity terms that once bound OpenAI to Microsoft’s Azure cloud infrastructure for all its computing needs. Now, the deal includes a “right of first refusal” (ROFR) clause for Microsoft, meaning while Microsoft gets the first dibs on providing new computing capacity, OpenAI can look elsewhere if Microsoft declines or cannot meet the demand. “To further support OpenAI, Microsoft has approved OpenAI’s ability to build additional capacity, primarily for research and training of models,” Microsoft stated, signaling a more flexible approach to their collaboration.

Despite these changes, the foundational aspects of the partnership remain intact. Microsoft retains exclusive rights to integrate OpenAI’s technology into its products like Copilot, and OpenAI’s API will continue to be exclusive to Azure. The revenue-sharing model, where Microsoft reportedly receives 20% of OpenAI’s revenue, also continues unchanged, ensuring both parties benefit financially from their joint efforts.

This adjustment comes on the heels of a major tech alliance aimed at building an ambitious network of data centers called Stargate, involving giants like Arm, Microsoft, NVIDIA, Oracle, and OpenAI. This initiative underscores the immense computing power required to push the boundaries of AI model development, particularly as OpenAI aims to cater to an ever-expanding user base.

The necessity for this change has roots in previous tensions where some OpenAI shareholders expressed concerns over Microsoft’s pace in providing computing resources. This led to a temporary solution in June when OpenAI, with Microsoft’s approval, partnered with Oracle to fulfill its compute needs, highlighting the urgency for more flexible arrangements.

The relationship between Microsoft and OpenAI hasn’t been without its public scuffles. The dramatic, albeit brief, ousting of OpenAI’s CEO Sam Altman last year cast a spotlight on the complexities of their partnership. The New York Times reported on the growing strains, attributing them to financial pressures within OpenAI, concerns about its stability, and interpersonal friction among the tech companies’ employees.

A notable point of contention was Microsoft’s acquisition of Inflection AI’s CEO Mustafa Suleyman, along with much of his team, for a hefty $650 million last March. This move was perceived by some at OpenAI, including Altman, as Microsoft potentially grooming a rival technology or diversifying its AI strategy away from OpenAI’s technologies.

The AGI clause

Perhaps one of the most intriguing aspects of the Microsoft-OpenAI partnership is the AGI clause. This clause stipulates that if OpenAI achieves artificial general intelligence (AGI) — defined here as an AI system capable of generating over $100 billion in profits — Microsoft could lose access to some of the most advanced models developed thereafter. Initially, this was seen as a safeguard to prevent the commercialization of highly advanced AI. However, current rumors suggest OpenAI might be willing to drop this clause to secure additional investment or support from Microsoft, indicating a strategic pivot towards more immediate commercial gains over long-term ethical considerations.

Looking ahead

This revised partnership could signal a new era of flexibility and possibly competition in AI development, where companies like OpenAI aren’t tethered to a single provider for their computational needs. For Microsoft, this might be about ensuring they remain central to AI innovation while also hedging against the rapid evolution and potential unpredictability of AI technology development. For OpenAI, it’s an opportunity to accelerate its research without being bottlenecked by a single company’s capabilities or strategies.