• by Laurie Sullivan , July 28, 2026

Amazon has begun to phase out some of its in-house AI models, as well as reorganize teams and focus engineers on new strategies.

An Amazon spokesperson acknowledged an early Tuesday morning report by explaining that its AI portfolios evolve based on customer needs. This is one way that Amazon provides companies, agencies and brands with guidance and migration paths as models advance, and others become obsolete. 

"AI models remain one of the most important things we're working on, and that hasn't changed," an Amazon spokesperson told MediaPost on Tuesday. "We are continuing to invest in and support the Nova models our customers rely on today, including through Nova Forge, which lets businesses build their own models on Nova as a foundation."

Models in production are typically supported for extended periods, while continuing to invest in the next generation of frontier model research.

Amazon’s commitment to building advanced AI models remains as strong as it has been in the past, the spokesperson said. 

The report from Business Insider published today attributed the change to a restructuring of Amazon's AI strategy. "Rather than investing across a number of text, image, and video models, the company is concentrating engineering talent and scarce computing resources on its highest priorities," Business Insider wrote. 

Amazon last week quietly closed its AGI Lab, the team it set up about 18 months ago to support AI frontier models. These are the most advanced models trained on massive datasets to deliver state-of-the-art performance across many tasks and typically power advanced reasoning, image and text generation, and agentic workflows.

The AGI Lab was founded in December 2024 and built around several dozen employees from the startup Adept, including its co-founder and CEO David Luan, who departed in February. 

More than a dozen Adept employees have left Amazon during the past several months. 

Executives who ran Amazon's AGI strategy and the lab's head had already left, suggesting Amazon has stepped back from a frontier-model race.

A growing number of onlookers argue that large language models have a long way to go to reach artificial general intelligence (AGI).