Amazon is deprecating several internally developed Nova AI models, including Nova Premier, Nova Omni, the Reel video-generation model, and the Canvas image-generation model, according to a Business Insider report citing people familiar with the matter. The wind-down signals a shift in resources toward a new frontier-model project as the company works to close the competitive gap with rivals like OpenAI, Anthropic, and Google. Amazon had not responded to Reuters' request for comment when the report was published, meaning the changes have not been directly confirmed by the company.
What's being cut and what survives
The deprecation list includes flagship models that Amazon had positioned as multi-modal and generative AI tools. Nova Premier and Nova Omni were designed for advanced reasoning and multi-modal tasks, while Reel and Canvas targeted video and image generation. The cuts do not extend to the full Nova portfolio. Nova 2 Lite, Nova 2 Sonic, and Nova Forge - a service that lets customers build customized models using Amazon's technology - remain active.
The decision follows job reductions in Amazon's artificial-general-intelligence group. The company said at the time it was concentrating on initiatives it considered most important to customers. The AGI division was consolidated under longtime cloud executive Peter DeSantis in December, after several senior executives left during the previous year.
A new frontier-model bet
Amazon is redirecting resources to a frontier-model project led by AI researcher Pieter Abbeel, Business Insider reported. The new flagship model is expected to debut at Amazon's annual re:Invent conference later in 2026. The Nova brand may carry over to the new system, though no final decision has been disclosed. The reorganization suggests a narrower focus on fewer models with the potential to compete at the top tier of AI performance.
For executives tracking strategic shifts in enterprise AI, the consolidation mirrors a broader industry pattern: trimming scattered R&D efforts to concentrate spending on models that can win benchmark comparisons and developer mindshare. These decisions ripple through procurement, vendor selection, and build-versus-buy calculations - topics covered in the AI for Executives & Strategy resource hub. Leaders evaluating how to allocate internal AI budgets can also reference the AI Learning Path for CEOs for frameworks on assessing model strategies.
The competitive pressure behind the pivot
Amazon has struggled to generate the same attention around Nova that OpenAI, Anthropic, and Alphabet's Google have attracted for their models. That gap persists even as Amazon continues expanding cloud and AI infrastructure for enterprise customers through AWS. The company's strength in infrastructure has not translated into leadership in frontier model development, and the Nova retrenchment suggests an acknowledgment of that reality.
Investors may view the reorganization as an attempt to concentrate development spending on fewer models capable of competing more effectively at the frontier of AI performance. The bet on a 2026 debut under Abbeel's leadership gives Amazon roughly 18 months to deliver a model that can shift the conversation.
Why this matters for executives and strategy
Amazon's Nova pullback is a reminder that even deep-pocketed incumbents can struggle to gain traction in crowded AI markets. For executives building AI roadmaps, the lesson is not that in-house model development is a mistake - but that diffusion of resources across too many projects can leave a company without a clear winner. The reorganization under DeSantis points to a hard-nosed triage: protect what works, cut what does not, and place a concentrated bet on a single flagship effort. Whether that bet pays off will not be clear until late 2026, but the strategic logic behind it is worth watching for any organization navigating its own AI portfolio decisions.
Your membership also unlocks:
