Amazon is shaking up the AI landscape with a provocative new offering that lets enterprises and governments run cutting-edge AI systems inside their own data centers. Dubbed “AI Factories,” the program lets customers supply the power, space, and physical infrastructure while AWS installs, operates, and integrates the AI stack with other AWS services.
The core appeal is clear: data sovereignty. Organizations that want absolute control over their data—and to prevent it from flowing to competitors or foreign actors—can keep data onsite and avoid sharing hardware or models externally. In essence, an on-prem AI Factory keeps data out of the cloud model marketplace while still enabling access to advanced AI capabilities.
The product name echoes Nvidia’s own branding for its comprehensive AI hardware ecosystem. In fact, AWS describes the collaboration as a joint effort with Nvidia, tying together Nvidia’s GPU and networking hardware with AWS software and services.
Practically, the AWS Factory blends Nvidia’s latest hardware with AWS technology. Companies can choose Nvidia’s Blackwell GPUs or Amazon’s Trainium3 chips. The setup leverages AWS networking, storage, databases, and security, and can be connected to Bedrock for model selection and management, as well as SageMaker for model building and training.
This isn’t a one-off move by AWS. Other tech giants are pursuing a similar path. In October, Microsoft unveiled its own fleet of AI Factories—high-powered systems deployed across its global data centers to handle OpenAI workloads. Microsoft highlighted these facilities as part of its broader strategy to create private-cloud-ready, state-of-the-art AI infrastructure, branding some facilities as AI Superfactories in development sites in Wisconsin and Georgia.
Earlier, Microsoft also laid out plans for sovereign cloud solutions designed to serve European organizations with data localization in mind, including options like Azure Local, which enables managed hardware to be installed on customer premises.
The trend reveals a curious irony: as AI accelerates, the biggest cloud providers are doubling down on private data centers and hybrid cloud configurations that resemble an early-2000s pattern—bringing critical processing closer to the user rather than relying solely on centralized public clouds.
TechCrunch event
San Francisco | October 13–15, 2026.
