December 5, 2025

Amazon Web Services unveiled four new “Nova” AI models, Nova Lite, Nova Pro, Nova Sonic and Nova Omni, alongside Nova Forge, a tool that lets customers pretrain their own models by injecting proprietary data at different stages. This democratizes frontier‑model development and signals a shift toward customizable AI infrastructure. Entrepreneurs can now tailor models to niche use cases without building everything from scratch.

Full Article

Amazon’s move into frontier AI continues. According to Wired, the company introduced a family of Nova models in early December that span a range of sizes and capabilities. Nova Lite targets casual use with modest compute requirements; Nova Pro adds more parameters and reasoning ability; Nova Sonic emphasizes speed; and Nova Omni is a full‑scale model competing with GPT‑5. What really sets Amazon apart is Nova Forge, a service that allows customers to inject their own data into the training process at different stages.

Traditionally, customizing a frontier model meant fine‑tuning an existing one, with limited influence over early layers. Nova Forge upends that by letting companies add their data during pretraining. This means a manufacturing firm could train the model with decades of maintenance logs to create a specialized predictive‑maintenance assistant, or a media company could infuse its archive to develop a proprietary summarization tool. By using Amazon’s compute infrastructure, customers avoid the enormous costs of building from scratch.

For entrepreneurs, Nova represents a strategic opportunity:

•             Own your data: You can differentiate by blending your unique datasets into the model rather than relying solely on public training corpora.

•             Stay compliant: Controlling training data helps meet regulatory requirements around data provenance and privacy.

•             Iterate faster: With Nova’s managed environment, you can experiment with different pretraining splits and quickly evaluate performance on your tasks.

However, there are trade‑offs. Locking into one cloud provider could reduce flexibility, and custom pretraining still requires significant technical expertise. Entrepreneurs should weigh whether the productivity gains justify potential vendor lock‑in.

Subscribe to our newsletter to stay up to date with the latest breakthroughs in AI, Business, Technology, and Mindset.