Hybrid infrastructure decisions put AI costs and control in focus
SiliconANGLE Jonathan Anthony ● Covered by 3 sources
Enterprises are shifting AI work to hybrid setups as inferencing costs jump. Dell says the real fight now is control, not just model quality.
Based on reporting by SiliconANGLE, Jonathan Anthony — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Enterprise AI is hitting a more cautious phase, and the bill for inference is doing a lot of the talking. Companies are no longer asking only what a model can do. They are asking where it should run, which model they should pick and how much data they’re willing to hand off along the way.
That pressure is showing up in buying decisions across industries, including finance, where firms are trying to turn scattered data into something useful in real time. Dell says more than 6,500 enterprises have adopted its AI Factory with Nvidia updates, and the company is hearing a clear change in what customers want.
Kyle Dufresne, Dell’s senior vice president of sales, said the economics around tokens have flipped quickly. In his telling, frontier models were generating 80% of tokens just three months ago. Now, he said, open models are generating 80% of tokens. The reason, he argued, is simple: cost.
That cost squeeze is pushing AI work toward hybrid infrastructure and, increasingly, toward the edge. Dell is seeing inference happen wherever data is created — in the cloud, in the data center and on PCs — because business owners are looking for ways to cut inferencing costs without slowing down innovation. And that brings a harder question with it: whether companies want to outsource not just workloads, but data and differentiation too.
The supply side is getting uglier as well. Dufresne said memory prices are rising, while next year is expected to bring about 20% more memory to market against roughly 200% more demand. Dell says it is responding by helping customers plan ahead, steer toward more available components and get advance visibility into coming cost increases. In some cases, buyers are moving early because their cost of capital is lower than the inflation they expect from the hardware itself.
My take — AI-written commentary, not fact-checked reporting
This is the part of AI that sounds boring until the invoice arrives. The shiny model race is sliding into a procurement fight, and that usually means open models and controlled infrastructure get the last laugh. Turns out “AI strategy” is often just a very expensive spreadsheet with better branding.
Read more about this at: SiliconANGLE