Business

AWS introduces on-premises AI factories to offer private, hyperscale model deployment

Thursday, 4 December 2025 5:37AM UTC

Amazon Web Services has launched AI Factories, a managed service delivering full‑stack AI infrastructure within clients' data centres, enabling large‑scale models to be run locally while maintaining control over sensitive data.

Amazon Web Services has launched AI Factories, a managed service that installs full‑stack AI infrastructure inside customers’ own data centres so organisations can run large‑scale models without moving sensitive data off‑site. According to the original report, the offering combines AWS’s Trainium chips with NVIDIA GPUs and integrates networking, storage, databases and AI tools such as Amazon Bedrock and SageMaker to deliver a private, low‑latency environment. ^[1]^[2]^[3]

AWS says customers supply space, power and connectivity while AWS manages procurement, installation, networking and software integration, shortening what can be a months‑or‑years build‑out into a managed deployment. Industry material from AWS frames the AI Factory as a private environment similar to a dedicated AWS Region, with services and support for large‑scale workloads. ^[2]^[3]

The technical stack layers AWS Trainium processors with NVIDIA accelerators , including Grace Blackwell and Vera Rubin GPUs in the UC Today reporting , and uses high‑speed interconnects such as Elastic Fabric Adapter and Nitro virtualisation to optimise throughput for modern models. Speaking to UC Today, Ian Buck, Vice President and GM of Hyperscale and HPC at NVIDIA, said: "Large‑scale AI requires a full‑stack approach – from advanced GPUs and networking to software and services that optimise every layer of the data centre." ^[1]

Data sovereignty and compliance are central to the pitch. AWS positions AI Factories for enterprises and government agencies that must keep controlled workloads on‑site; the company says the infrastructure can handle classification levels from Unclassified up to Top Secret. The service is being marketed as a way to retain control over sensitive information while accessing hyperscale compute. ^[1]^[2]^[3]

AWS has already announced a major regional deployment with HUMAIN in Saudi Arabia to create an AI "zone" in Riyadh, targeting up to 150,000 AI accelerators. Tareq Amin, CEO of HUMAIN, said the project “represents the beginning of a multi‑gigawatt journey for HUMAIN and AWS.” Businesswire and AWS material also specify inclusion of NVIDIA’s latest GB300s alongside Trainium chips in that deployment. ^[1]^[4]

The move sits within a broader industry shift toward hybrid models that pair cloud services with on‑premises control. Microsoft and other cloud vendors have launched comparable local and managed on‑premises offerings to address sovereignty and latency demands. Reuters coverage and AWS commentary also underline ongoing chip and server advances , including AWS’s Trainium3 servers and plans to integrate NVIDIA’s NVLink Fusion concepts into future Trainium designs , that together aim to boost performance and energy efficiency. ^[1]^[5]^[6]

For enterprise IT, AI Factories promise faster access to high‑performance infrastructure but bring new operational responsibilities. Organisations must budget for power and space, plan integration with existing systems, and secure staff with expertise in model deployment, monitoring and security; AWS manages the hardware layer but not every aspect of run‑time model engineering. Analysts and AWS messaging note that upskilling or hiring specialists will be crucial. ^[1]^[2]^[7]

AWS’s announcement underscores a strategic recalibration: AI is reshaping infrastructure choices and encouraging hybrid architectures that balance cloud scale with on‑site control. According to AWS and reporting, the AI Factory model may accelerate regional deployments and give regulated sectors a path to adopt advanced AI while retaining sovereignty and compliance oversight. ^[1]^[2]^[4]

##Reference Map:

^[1] (UC Today) - Paragraph 1, Paragraph 3, Paragraph 4, Paragraph 5, Paragraph 7, Paragraph 8
^[2] (AWS, About AWS) - Paragraph 1, Paragraph 2, Paragraph 4, Paragraph 7, Paragraph 8
^[3] (About Amazon) - Paragraph 1, Paragraph 2, Paragraph 4
^[4] (Businesswire) - Paragraph 5, Paragraph 8
^[5] (Reuters) - Paragraph 6
^[6] (Axios) - Paragraph 6
^[7] (Reuters, Trainium credits) - Paragraph 7

Source: Noah Wire Services

More on this

https://www.uctoday.com/unified-communications/aws-launches-on-premises-ai-factories-powered-by-nvidia/ - Please view link - unable to able to access data
https://aws.amazon.com/about-aws/global-infrastructure/ai-factories/ - AWS AI Factories offer fully managed AI infrastructure that can be deployed within customers' existing data centres. This service combines AWS's Trainium chips with NVIDIA GPUs, networking, storage, databases, and AI tools like Amazon Bedrock and SageMaker AI. It enables organisations to run large-scale AI workloads on-site, ensuring data sovereignty and compliance with regulatory requirements. AWS handles the deployment and management of the infrastructure, allowing customers to focus on AI innovation without the complexities of hardware setup and maintenance.
https://www.aboutamazon.com/news/aws/aws-data-centers-ai-factories/ - AWS AI Factories provide dedicated AI infrastructure within customers' data centres, integrating NVIDIA GPUs, AWS's Trainium chips, and AI services such as Amazon Bedrock and SageMaker AI. This approach allows organisations to maintain control over sensitive data while leveraging AWS's advanced AI capabilities. The service is designed to meet the needs of enterprises and government agencies requiring strict data sovereignty and compliance, offering a secure and efficient solution for deploying AI applications at scale.
https://www.businesswire.com/news/home/20251119637708/en/AWS-and-HUMAIN-Expand-Partnership-with-NVIDIA-AI-Infrastructure-and-AWS-AI-Chip-Deal-to-Drive-Global-AI-Innovation - AWS and HUMAIN have expanded their partnership to deploy up to 150,000 AI accelerators, including NVIDIA's latest GB300s and AWS's Trainium chips, in a dedicated AI Zone in Riyadh. This collaboration aims to provide customers with access to advanced AI infrastructure and services, including Amazon Bedrock and SageMaker AI, facilitating the development and deployment of AI applications. The initiative underscores AWS's commitment to driving global AI innovation and supporting organisations in meeting data sovereignty and compliance requirements.
https://www.reuters.com/business/retail-consumer/amazon-use-nvidia-tech-ai-chips-roll-out-new-servers-2025-12-02/ - AWS has announced the integration of NVIDIA's NVLink Fusion technology into its future AI chip, Trainium4, to enhance AI infrastructure performance. This partnership aims to enable faster inter-chip communication for large AI models. Additionally, AWS unveiled new servers using its Trainium3 chip, delivering over four times the performance of existing AI infrastructure while using 40% less power. These developments highlight AWS's commitment to advancing AI capabilities and providing efficient solutions for AI workloads.
https://www.axios.com/2024/03/12/aws-ceo-ai-bedrock-amazon-anthropic - AWS CEO Adam Selipsky revealed that over 10,000 organisations currently use AWS's Bedrock service to build AI applications, representing about 1% of the potential market. AWS aims to dominate the AI space, aspiring to control up to 99% of the market. The company’s AI strategy spans from data centres and proprietary chips to hosting and refining external AI models. A key development is AWS’s deepening partnership with AI startup Anthropic, which launched its Claude 3 models in early March 2024.
https://www.reuters.com/technology/artificial-intelligence/amazon-offers-free-computing-power-ai-researchers-aiming-challenge-nvidia-2024-11-12/ - AWS announced it will provide $110 million worth of free cloud computing credits to AI researchers to utilise its custom Trainium chips, aiming to compete with Nvidia, AMD, and Google's cloud offerings. The initiative is designed to attract researchers to AWS’s AI infrastructure by offering access to 40,000 of its first-generation Trainium processors. Participants already include researchers from Carnegie Mellon University and UC Berkeley. This move underscores AWS's commitment to advancing AI research and challenging competitors in the AI chip market.

Noah Fact Check Pro

The draft above was created using the information available at the time the story first emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed below. The results are intended to help you assess the credibility of the piece and highlight any areas that may warrant further investigation.

Freshness check

Score: 10

Notes: The narrative is current, with the earliest known publication date being December 2, 2025. No earlier versions with differing figures, dates, or quotes were found. The content is original and not recycled from other sources. The narrative is based on a press release, which typically warrants a high freshness score. No discrepancies or outdated material were identified.

Quotes check

Score: 10

Notes: The direct quote from Ian Buck, Vice President and GM of Hyperscale and HPC at NVIDIA, appears to be original, with no earlier matches found online. This suggests the content is potentially original or exclusive.

Source reliability

Score: 8

Notes: The narrative originates from UC Today, a reputable source in the unified communications sector. However, it is not as widely recognised as major outlets like the Financial Times or BBC, which introduces a slight uncertainty.

Plausability check

Score: 9

Notes: The claims about AWS's AI Factories align with recent announcements from AWS and NVIDIA, including the integration of NVIDIA's NVLink Fusion technology into AWS's future AI chip, Trainium4, and the unveiling of new servers using AWS's Trainium3 chip. The narrative lacks specific factual anchors such as names, institutions, and dates, which slightly reduces its credibility. The language and tone are consistent with typical corporate communications, and there is no excessive or off-topic detail.

Overall assessment

Verdict (FAIL, OPEN, PASS): PASS

Confidence (LOW, MEDIUM, HIGH): HIGH

Summary: The narrative is current, original, and aligns with recent developments from AWS and NVIDIA. While sourced from a less widely recognised outlet, the content is plausible and consistent with corporate communications. The lack of specific factual anchors slightly reduces its credibility, but overall, the assessment is positive.

AWS
AI infrastructure
hybrid cloud