AMD Chips Bring AI Processing Close to Data
As AI adoption grows, organizations are taking a closer look at where AI workloads should run. While large cloud models remain essential for many use cases, businesses are increasingly interested in running some AI tasks closer to users and data.
That shift is creating new opportunities for AI-ready PCs powered by advanced processors and neural processing units. AMD’s latest AI-capable silicon is designed to support local AI inferencing, enabling workloads that previously required cloud resources to run directly on the endpoint.
McGilvrey compares the trend to the evolution of enterprise storage. Years ago, organizations rushed to move data into the cloud before realizing some workloads were better suited to hybrid environments.
“The same thing is happening with AI,” he says. “Everyone was racing to take their data to the AI models in the cloud. Now, you’re seeing this platform shift of where, ‘OK, yes, maybe I do need very powerful models in the cloud, but I’m now rethinking where I’m going to run the AI and where I’m going to run the inferencing.’”
Running AI locally can offer several advantages. Sensitive data can remain on the device, improving privacy and security. AI responses can be delivered with little or no latency. And organizations can reduce their dependence on cloud-based AI services for appropriate workloads.
“Customers really need to develop a distributed inferencing strategy,” McGilvrey says. “I want to inference locally on my device. I’m going to do AI inferencing in the cloud. And then, if I have investment on-premises, I want to do AI inferencing on-prem.”
DISCOVER: AMD hardware can support your enterprise technology.
Windows 11, Copilot and AMD Create a Foundation for AI
The partnership between AMD and Microsoft is helping make that distributed AI model possible.
Microsoft is building AI capabilities directly into Windows 11 and its Copilot ecosystem, while AMD provides the hardware foundation needed to execute many of those workloads efficiently on the endpoint.
McGilvrey points to intelligent Windows search capabilities that understand natural language requests, along with AI-powered tools that help users capture, organize and retrieve information more effectively.
Those benefits extend beyond user productivity. Organizations are also grappling with the rising cost of AI consumption. Although token prices continue to decline, overall use is growing rapidly as AI becomes embedded in more business processes.
“You are seeing organizations that are burning through their AI budgets in four months due to token consumption,” McGilvrey says.
In many cases, organizations may be able to use smaller models running locally on AI-ready PCs rather than relying exclusively on large, cloud-based models. That approach can reduce costs while still delivering desired business outcomes.
READ MORE: How to achieve full-stack observability in your business.
AI-Ready Endpoints Will Support the Agentic Future
As enterprises begin deploying AI agents, endpoint strategy is likely to become even more important.
Microsoft recently introduced new capabilities aimed at helping organizations manage and govern agents running within their environments. Windows 11 is increasingly positioned as a platform for securely running and managing those workloads, while AMD-powered AI PCs provide the local processing resources needed to support them.
“Windows 11 is the platform for AI and, really, agents in general,” McGilvrey says.
For IT leaders, that means endpoint refresh decisions can no longer be made independently of AI planning. Devices purchased today will influence an organization’s ability to adopt Copilot, deploy agents and manage AI costs for years to come.
“It’s imperative that our customers are now including AI decision-makers in the endpoint decision-making process,” McGilvrey says.
