Amazon CEO Andy Jassy dropped an AI bombshell on employees today, announcing that Rohit Prasad—who has led Amazon’s so-called AGI (artificial general intelligence) team since 2023, overseeing the development of the company’s Nova models—will depart at the end of the year.
Prasad previously served as the head scientist behind Amazon’s Alexa voice assistant, a role he held from the product’s earliest days. When he was appointed to lead the new ambitious AGI effort after ChatGPT launched in November 2022, as part of a scramble to develop a competitive LLM that could help reinvigorate the Alexa voice assistant. it was led almost entirely by ex-Alexa executives.
In a blog post, Jassy announced that longtime Amazon Web Services (AWS) executive Peter DeSantis will lead a new organization that drives the development of its AI models, custom computer chips (which include its Graviton, Trainium and Nitro chips), and quantum computing efforts. DeSantis had overseen the many teams designing AWS’ global infrastructure.
“With our Nova 2 models just launched at re:Invent, our custom silicon growing rapidly, and the advantages of optimizing across models, chips, and cloud software and infrastructure, we wanted to free Peter up to focus his energy, invention cycles, and leadership on these new areas,” Jassy wrote, adding that DeSantis would report directly to him.
Jassy also said that as part of the organizational change, Pieter Abbeel, an Amazon Distinguished Scientist in robotics who is also an AI and robotics professor at UC Berkeley, will lead the company’s frontier model research team. Abbeel came to Amazon in 2024 along with other cofounders of his robotics startup Covariant, in a deal that also saw Amazon licensing Covariants software, which included AI models that gave robots the ability to quickly adapt to new environments and tasks.
“Pieter is one of the world’s leading AI researchers, and co-founder of Covariant, which pioneered the first commercial foundation model for robotics,” Jassy wrote. “His deep expertise in generative AI and reinforcement learning makes him well-suited to advance Amazon’s AI research as we push the boundaries of what’s possible for customers.”
The news of Prasad’s departure comes as somewhat of a surprise, given that he was recently at Amazon’s Re:Invent conference discussing the latest Nova models. However, over the past two years there has been significant media coverage suggesting that Amazon’s Alexa AI and AGI-related efforts have struggled and fallen behind competitors.
A year ago, for example, Fortune’s Jason Del Rey reported exclusively that leaked Amazon documents identified critical flaws in the delayed AI reboot of Alexa. And in June 2024, Fortune reported that Amazon’s had blown Alexa’s shot to dominate AI, according to more than a dozen employees who worked on it—partly due to a lack of adequate data, even though Prasad, pushed the AGI team to work harder and harder, with a message to “get some magic” out of the LLM.
In addition, last week’s Amazon layoffs fueled concerns about whether Amazon’s was still lagging behind in AI, and whether the cuts reflected slowing growth. That came on the heels of comments in October by analyst Mark Shmulik of Bernstein, who said Amazon’s AWS was in “last place” in the AI cloud race.
However, The Information as well as Bloomberg reported this week that Amazon was in talks to invest $10 billion in OpenAI. OpenAI, in turn, had agreed to use Amazon’s Tranium AI chips, perhaps helping to counter the narrative that the company is behind in AI. OpenAI had previously agreed to spend $38 billion using AWS for computing.
Amazon also has a deal with AI company Anthropic, in which Amazon has invested $8 billion. Anthropic has agreed to use AWS’s Trainium chips for training and Anthropic’s Claude model is being used to answer some queries in the new Alexa Plus.
