The battle for enterprise AI just got more interesting. Amazon has launched AI Factories, a new solution that lets large companies and governments run Amazon’s AI systems inside their own data centers. Instead of sending sensitive data to the cloud, organizations can keep everything on-premises—while still tapping into AWS’ powerful AI ecosystem.
Why does this matter?
Because data sovereignty has become a top priority. Many enterprises want full control over their data to avoid security risks, regulatory issues, or exposure to competitors. With AI Factories, AWS installs and manages the AI infrastructure directly at the customer’s location. Customers provide the facility and power — AWS brings the AI stack.
What’s inside these AI Factories?
A mix of AWS technology and Nvidia’s high-performance hardware. Companies can choose between Nvidia’s advanced Blackwell GPUs or Amazon’s new Trainium3 chips. Plus, it integrates seamlessly with services like Amazon Bedrock and SageMaker for model building, training, and deployment.
Interestingly, AWS isn’t alone. Microsoft is also investing heavily in Nvidia-powered AI Factories for its global data centers, though it hasn’t fully opened them for on-prem private cloud use. Its push for “Azure Local” shows the same trend: AI is pushing the cloud giants back toward hybrid and on-prem infrastructure.
The takeaway?
We’re seeing a full-circle moment. As AI demands more control, power, and security, the biggest cloud companies are racing to build high-performance systems that sit closer to clients — just like the old data center days, but supercharged for the AI era.
A new wave of enterprise AI is here, and on-prem is suddenly cool again.
