Amazon.com (AMZN) is orchestrating a dramatic overhaul of its artificial intelligence strategy, effectively pulling the plug on active development of most of its flagship Nova AI models. The move, reported just days before the company’s second-quarter earnings release, signals a sharp pivot away from competing across a broad range of generative AI categories and toward a concentrated bet on a single, next-generation frontier model.

The decision to deprecate the high-end Nova Premier and Omni models, along with the Reel video-generation model and Canvas image-generation model, was confirmed by people familiar with the matter. While Amazon will continue to support existing customers using these tools, internal resources are being redirected. Some employees have reportedly described the affected models as being in “KTLO,” or “keep the lights on” mode—an engineering term indicating that while maintenance and customer support will persist, the products are no longer a primary focus for innovation.

The strategic reset comes on the heels of significant organizational upheaval. Last week, Amazon confirmed layoffs within its Artificial General Intelligence (AGI) organization and the closure of its AGI Lab, a research unit formed in 2024 after the company hired much of the team behind AI startup Adept. These actions, combined with the Nova pullback, paint a picture of a company that is ruthlessly prioritizing resources after struggling to generate the same market buzz around its proprietary models as rivals like OpenAI, Anthropic, and Google.

The New Frontier

At the center of Amazon’s reworked AI roadmap is a new initiative known internally as Frontier Model Research (FMR). Leadership of this top-priority project has been handed to Pieter Abbeel, a prominent researcher who joined Amazon following its acquisition of the AI robotics startup Covariant. Abbeel’s team is racing to develop a new flagship foundation model, which is expected to make its debut at Amazon’s annual re:Invent conference later this year.

This represents a significant consolidation of strategy under Peter DeSantis, the longtime cloud executive who took over Amazon’s AGI efforts in December following the departure of former AGI lead Rohit Prasad. Unlike the previous approach of fielding a sprawling portfolio of models for text, image, and video generation, DeSantis’s vision is far more focused: channeling engineering talent and massive computing resources into a single, cutting-edge system designed to compete at the highest tier of AI capability.

Nova Isn’t Dead Yet

The reported shift does not mean Amazon is entirely abandoning the Nova brand. The company continues to actively support Nova 2 Sonic, Nova 2 Lite, and Nova Forge—a service that allows customers to build and customize their own AI models using Nova technology. Additionally, Nova Act, Amazon’s AI agent platform, remains operational. The upcoming frontier model could even eventually be released under the Nova banner, preserving brand continuity even as the underlying technology strategy transforms.

Wall Street Remains Bullish Ahead of Earnings

The strategy overhaul lands in a critical window for Amazon. The stock has suffered five consecutive losing sessions, and the company is scheduled to report quarterly earnings on Thursday. Despite the apparent internal turbulence, Wall Street analysts are maintaining a positive outlook.

Evercore ISI reiterated an ‘Outperform’ rating on Amazon shares with a $315 price target. The firm believes Amazon’s second-quarter revenue estimates of roughly $196 billion—representing approximately 17% year-over-year growth and 9% sequential growth—appear achievable and carry more upside than downside risk. Shares of Amazon traded flat in pre-market hours on Tuesday, suggesting that investors are taking the strategic pivot in stride, potentially viewing the consolidation as a necessary step to compete more effectively in the AI arms race.

Amazon did not immediately respond to requests for comment on the report. The company’s AGI efforts have undergone several iterations since taking shape in 2023, and the latest restructuring underscores the immense pressure on tech giants to deliver not just a wide array of AI products, but a single, dominant model that can capture developer mindshare and enterprise budgets.