Amazon takes on competitors with on-premises Nvidia “AI Factory”

Amazon on Tuesday announced a new product called “AI Factories” that will allow large companies and governments to run AI systems in their data centers. Or, in AWS’s words, the customer supplies the power and data center, and AWS integrates and manages the AI ​​system and connects it to other AWS cloud services.

The idea is to respond to businesses and governments concerned about data sovereignty, or having complete control over data so it doesn’t fall into the hands of competitors or foreign adversaries. An on-premises AI factory means no data is sent to the model maker and no hardware is shared.

If the product name sounds familiar, it should. That’s what Nvidia calls its hardware system, which is chock-full of the tools needed to run AI, from GPU chips to networking technology. The companies say this AWS AI Factory is actually a collaboration with Nvidia.

In this case, AWS Factory uses a combination of AWS and Nvidia technologies. Companies deploying these systems can choose between Nvidia’s latest Blackwell GPUs or Amazon’s new Trainium3 chips. It uses AWS proprietary networking, storage, database, and security, and is powered by Amazon Bedrock, an AI model selection and management service, and AWS SageMaker AI, a model building and training tool.

Interestingly, AWS isn’t the only large cloud provider deploying Nvidia AI Factory. In October, Microsoft unveiled the first of many upcoming AI Factories that will be deployed in its global data centers to run OpenAI workloads. Microsoft did not announce at the time that these Extreme Machines would be available in a private cloud. Instead, Microsoft highlighted how it’s relying on a host of Nvidia AI Factory data center technologies to build and connect its new “AI Superfactory,” a new state-of-the-art data center it’s building in Wisconsin and Georgia.

Last month, Microsoft also outlined data centers and cloud services that will be built in local countries to address data sovereignty issues. To be fair, that option also includes “Azure Local,” Microsoft’s own managed hardware that can be installed at customer sites.

Still, it’s a little ironic that AI is causing the biggest cloud providers to invest heavily in enterprise private data centers and hybrid clouds again, as they did in 2009.