Amazon is overhauling its AI portfolio, phasing out flagship Nova models and focusing on frontier‑model research.

Amazon is dramatically reshaping its artificial‑intelligence roadmap, announcing the retirement of most of its Nova‑series models while pledging to double down on frontier‑model research and cloud‑based AI services.

Why Amazon Is Pulling Back on Nova

The Nova line, introduced in 2023 as a suite of in‑house large language models, failed to meet internal performance benchmarks and lagged behind competitors in cost efficiency. Executives cited “resource constraints” and “strategic misalignment” as key reasons for the shift.

According to internal memos, the company plans to decommission the majority of Nova instances by the end of the fiscal year, redirecting engineering talent toward the development of next‑generation models that can be offered through Amazon Web Services (AWS).

Focus on Frontier‑Model Research

Amazon’s new AI agenda centers on building “frontier” models that rival the scale of OpenAI’s GPT‑4 and Google’s Gemini. The effort will be anchored in the AWS AI research labs, with an emphasis on multimodal capabilities, real‑time inference, and tighter integration with the company’s cloud infrastructure.

The company also announced a partnership with leading academic institutions to accelerate breakthroughs in model alignment and safety, aiming to address growing regulatory scrutiny.

Implications for Customers and Developers

Existing customers who rely on Nova for custom applications will be given a migration window to transition to AWS’s managed AI services, such as Bedrock and SageMaker. Amazon has promised “no disruption” for critical workloads, but developers may need to refactor code to accommodate different model APIs.

  • Enhanced access to large‑scale models via AWS Bedrock
  • Expanded support for multimodal inputs (text, image, audio)
  • New pricing tiers aimed at enterprise and startup users

Industry analysts view the move as a pragmatic response to the escalating costs of training and maintaining proprietary models, and as an effort to consolidate Amazon’s AI offerings under the more profitable cloud services umbrella.

We are reallocating resources to where we see the greatest long‑term value for our customers and shareholders,

— Amazon senior vice president of AI, in a statement released to the press.

For a detailed look at the restructuring plan, see Business Insider coverage of Amazon’s AI strategy overhaul.