Amazon just rolled out a suite of tools designed to make building custom AI models as simple as point-and-click. The announcement at AWS re:Invent comes as the cloud giant battles to catch up with OpenAI and Anthropic in the enterprise AI race, offering developers serverless model customization that could finally give AWS its competitive edge.
Amazon Web Services is making its biggest play yet for the custom AI model market. Just one day after CEO Matt Garman unveiled Nova Forge - a $100,000 annual service for fully custom models - the company dropped another round of announcements that could reshape how enterprises think about AI development.
At the AWS re:Invent conference on Wednesday, the cloud provider unveiled serverless model customization capabilities across both Amazon Bedrock and SageMaker AI. The timing isn't coincidental - AWS is fighting an uphill battle against established players like OpenAI and Anthropic, who dominate enterprise preferences according to recent Menlo Ventures research.
"A lot of our customers are asking, 'If my competitor has access to the same model, how do I differentiate myself?'" Ankur Mehrotra, general manager of AI platforms at AWS, told TechCrunch. "The key to solving that problem is being able to create customized models."
The new serverless approach in SageMaker eliminates the traditional barriers that kept smaller teams from experimenting with custom models. Developers can now choose between a self-guided point-and-click interface or an agent-led experience where they simply describe what they want in natural language. The agent feature, launching in preview, represents AWS's answer to the democratization trend sweeping AI development.
"If you're a healthcare customer and you wanted a model to be able to understand certain medical terminology better, you can simply point SageMaker AI, if you have labeled data, then select the technique and then off SageMaker goes, and [it] fine tunes the model," Mehrotra explained.
The capabilities extend beyond Amazon's own Nova models to include popular open-source options like DeepSeek and Meta's Llama series. This flexibility could prove crucial as enterprises increasingly demand control over their AI infrastructure while avoiding vendor lock-in.
Meanwhile, Bedrock is getting its own upgrade with Reinforcement Fine-Tuning, which automates the entire model customization pipeline. Developers can either define a reward function or select from pre-built workflows, then let Bedrock handle the heavy lifting from start to finish.
