Amazon is stepping up its AI offerings with the launch of “AI Factories,” a new on-premises solution designed for large corporations and government agencies. Announced Tuesday, the service allows customers to run Amazon’s AI systems directly in their own data centers. AWS manages the AI system while customers provide the power and infrastructure, and the solution can be integrated with other AWS cloud services.
The main appeal of AI Factories lies in data sovereignty. Organizations concerned about keeping sensitive information away from competitors or foreign adversaries can now deploy AI systems without sending data to external models or sharing hardware. Essentially, the AI comes to the data, rather than the other way around.
This new offering is built in collaboration with Nvidia, borrowing a page from the GPU maker’s own “AI Factory” concept. Companies can choose between Nvidia’s cutting-edge Blackwell GPUs or Amazon’s proprietary Trainium3 chip. The system also leverages AWS networking, storage, databases, security, and AI services such as Bedrock for model selection and SageMaker for model training.
Amazon isn’t alone in this approach. Microsoft unveiled its AI Factories in October, rolling out Nvidia-powered systems in its global data centers to run OpenAI workloads. Microsoft also emphasized building local data centers and hybrid cloud solutions to meet data sovereignty requirements, including Azure Local, which allows hardware to be installed directly on customer sites.
The push toward on-premises AI marks a shift in cloud computing, as providers invest heavily in private corporate data centers reminiscent of the early days of hybrid cloud adoption. As AI continues to evolve, these “factories” highlight a growing trend: major tech companies are increasingly helping clients run AI where the data lives, giving businesses more control, privacy, and flexibility than ever before.
source: Techcrunch
