Older computer tech is getting new attention as AI use shifts

In short: As AI spreads through businesses, companies that sell older building blocks of computing, like CPUs, servers, and some software, are seeing renewed interest.

What's going on

For much of the recent AI boom, the spotlight has been on specialized chips called GPUs, which are widely used to train large AI models. Training is the expensive part where an AI system learns from lots of data.

The Financial Times says attention is starting to shift toward what it calls “Old IT,” meaning the more traditional parts of computing. That includes servers, storage, and CPUs (central processing units, the general-purpose “brains” in most computers). One reason is the growing importance of “inference,” which is the day-to-day use of an AI model after it has been trained (like asking it questions and getting answers).

ARM and AMD recently reported stronger demand for CPUs, and their share prices rose. AMD said it now expects CPU sales to grow about 35 percent per year over the next few years, roughly double its forecast from six months ago.

Another driver is interest in AI agents, which are AI tools designed to take actions on a user’s behalf, not just answer questions. Intel’s finance chief told investors that training AI might need one CPU for every eight GPUs, but agent workloads could need as many CPUs as GPUs. Intel shares have jumped about 400 percent since the US government took a stake last summer, and Seagate shares rose 60 percent over the past month.

What to watch

A big open question is who captures the profits if AI agents become common. The article points to software companies like Salesforce, which is testing “headless” software (software built to run without a human-facing screen), a step meant to make it easier for AI agents to work inside business systems. If AI reduces the number of human users logged into software, pricing models based on per-person licenses may need to change.

Source: Financial Times