Amazon introduced a brand new product Tuesday known as βAI Factoriesβ that enables large companies and governments to run its AI techniques in their very own knowledge facilities. Or as AWS places it: Clients provide the facility and the information middle, and AWS plunks within the AI system, manages it, and might tie it into different AWS cloud providers.
The concept is to cater to firms and governments involved with knowledge sovereignty, or absolute management over their knowledge so it could possiblyβt wind up in a competitorβs or international adversaryβs arms. An on-prem AI Manufacturing facility means not sending their knowledge to a mannequin maker and never even sharing the {hardware}.
If that product identify sounds acquainted, it ought to. Thatβs what Nvidia calls its {hardware} techniques which might be chock-full of instruments wanted to run AI, from its GPU chips to its networking tech. This AWS AI Manufacturing facility is, in truth, a collaboration with Nvidia, each firms say.
On this case, the AWS Manufacturing facility will use a mix of AWS and Nvidia expertise. Firms that deploy these techniques can go for Nvidiaβs newest Blackwell GPUs or Amazonβs new Trainium3 chip. It makes use of AWSβ homegrown networking, storage, databases, and safety and might faucet into Amazon Bedrock β the AI mannequin choice and administration service, and AWS SageMaker AI, the mannequin constructing and coaching instrument.
Curiously, AWS is way from the one big cloud supplier putting in Nvidia AI Factories. In October, Microsoft confirmed off its first of many-to-come AI Factories rolling out into its world knowledge facilities to run OpenAI workloads. Microsoft didnβt announce on the time that these excessive machines could be accessible for personal clouds. As an alternative, Microsoft highlighted the way it was leaning on a bunch of Nvidia AI Manufacturing facility knowledge middle tech to construct and join its new βAI Superfactories,β aka new state-of-the-art knowledge facilities being inbuilt Wisconsin and Georgia.
Final month, Microsoft additionally outlined the information facilities and cloud providers that might be inbuilt native international locations to handle the information sovereignty challenge. To be truthful, its choices additionally embody βAzure Native,β Microsoftβs personal managed {hardware} that may very well be put in on buyer websites.
Nonetheless, it’s a bit ironic that AI is inflicting the most important cloud suppliers to speculate so closely in company non-public knowledge facilities and hybrid clouds prefer itβs 2009 over again.
Techcrunch occasion
San Francisco
|
October 13-15, 2026





