AI labs around the world are feverishly competing to build the best foundational models, but the CEO of one of the biggest companies in the game believes that there might not be much to differentiate between the models being built by the top labs.
This intriguing perspective comes from none other than Satya Nadella, CEO of Microsoft, a company deeply invested in AI research and development through its close partnership with OpenAI. On a recent podcast, Nadella shared some insightful thoughts on the evolving landscape of foundational AI models. His comments offer a glimpse into how one of the leading figures in tech views the rapidly changing dynamics of the AI industry. What makes his statement particularly interesting is his assertion that OpenAI, known for its cutting-edge models, isn’t primarily a model company, but a product company.
“We have the (OpenAI’s) IP rights, and so therefore we will want to build (models),” Nadella said. “We built Phi. Mustafa (Suleyman) and team have capability that they had in Pi before we have it in Microsoft,” he added. Nadella was talking about the chatbot Pi Mustafa Suleyman had launched while he was at inflection AI, and Microsoft’s Phi series of small AI models, hinting that Microsoft had the will — and the talent — to build models of their own.
But he hinted that foundational models might not end up being the most valuable part of the AI chain. “I do believe the models are getting commoditized in cloud,” he said. “In fact, OpenAI is not a model company; it’s a product company that happens to have fantastic models at this point, which is great for them and great for us, as both partners of theirs.” Nadella seemed to be suggesting that while having advanced models is crucial, the true competitive advantage arises from building successful products around them.
Finally, he concluded, “So I want to go now where I think the industry structure that’s emerging—models by themselves are not sufficient, but having a full system stack and great successful products, those are the two places.”
Nadella’s observations carry significant weight. His argument that foundational models are becoming commoditized suggests that simply having the most advanced model may not be a sustainable long-term advantage. The rapid pace of innovation in the field means that any lead in model performance is likely to be short-lived. This puts the focus on the next layer of the value chain: building compelling applications and services powered by these models.
The implication is that the future of AI belongs to companies that can effectively integrate these increasingly powerful, yet increasingly similar, models into user-friendly and valuable products. This shift in focus from model development to product development and system stack integration might lead to a reshuffling of the leaderboard in the AI race. Companies with strong product development capabilities and a robust ecosystem which can help them distribute their products, like Microsoft and Google, may be best positioned to capitalize on this emerging trend.