Everyone is talking about the explosion of artificial intelligence and what it might mean for the future of humanity. For all the talk about algorithms and large language models, however, it might be hardware — not software — that makes or breaks this transformative technology. One piece of hardware, in particular, suggests Lucas Melo, a cloud solutions architect at Intel: the central processing unit, or CPU.
While tech companies have been wringing their hands over highly sought-after GPU chips, it’s CPU hardware that has recently proven its outsized influence, according to Melo, who has closely watched the world of machine learning over the past few years and foresees a more CPU-intensive future for this remarkable innovation.
“It’s moving so fast,” Melo told The Forecast in an interview at .NEXT 2026.
“Even the folks working in this space, building generative AI, are surprised by the speed of change. Sometimes it can be hype, but many times it’s not. You need to take the time to get hands-on with the tech. There’s so much going on.”
Debunking the Myth of GPUs
CPUs are the brains behind the computer, executing many essential functions. Their popularity was eclipsed in recent years by GPUs, driven by the rise of LLMs like ChatGPT and Claude. As a result of the rapid rise and enormous investment in data centers, including private and public cloud and neocloud services, there is big demand for specialized GPUs. This created supply constraints, as Yahoo! Finance and others reported this spring.
GPUs have always been a big deal to digital gamers and media producers, but over time their ability to process many instructions in parallel made them well-suited for training and running AI models. But the evolution of AI is revealing the growing need for CPUs, said Melo.
“The myth that Gen AI equals GPUs isn’t true,” he said.
“Depending on the SLAs (service level agreements) and the business use case that you’re trying to solve for, let’s align your infrastructure and the technology to support that. As customers look to modernize, we’re trying to showcase and teach them that they don’t necessarily need silos of AI, silos of GPU infrastructure and then traditional enterprise.”
By starting with the business problem and working backward to find the best technological solution, enterprises often find that existing CPU-based infrastructure best fits the bill, Melo said.
The Surprising Rise of CPUs
“We’re starting to see where the ratio of CPU to GPU needs to change,” reported Melo, who cited agentic AI as a major reason why.
“Agentic AI is driving additional CPU workload due to the need for orchestration and execution of Agentic AI workloads.”
With GPU demand soaring, orders are piling up faster than companies can fulfill them. GPU prices have skyrocketed accordingly, and many AI projects are stalled waiting on constrained suppliers to deliver hardware. Therefore, ensuring maximum GPU utilization is more important than ever, Melo insists. Whether its data movement and pre-processing, scheduling or networking, enterprises can employ CPUs to ensure that GPU utilization is as high and as cost-effective as possible, he said.
“You paid a lot for those GPUs,” he said. “You want to make sure they’re busy.”
Melo said this is where CPUs are growing ever more essential to what’s happening next with AI, more inference in more places inside data centers and at the edge.
Agentic AI is the Future
“We’re having conversations around agentic AI, which in my opinion, is where we’re going to actually see a lot of change in the enterprise,” Melo said.
“With agentic, we’re starting to see how enterprise workloads are going to get automated.”
While traditional AI systems like LLMs respond to prompts, agentic AI takes things a step further by autonomously executing actions or a series of tasks to reach specific goals.
“API calls, database calls, code compilation, code testing, file system operations, those are things that agents will start to do,” Melo said. “Those workloads have always run best on CPUs.”
Because these applications rely heavily on CPUs, agentic AI is paving the way for a CPU renaissance, Melo suggested.
“We already see during our tests that 90% of the agentic workflow is standard enterprise workloads,” he said.
In June 2026 alone, Google ordered 3 million AI chips from Intel, including CPUs, while Meta announced an agreement with Qualcomm on its data center CPUs.
Clearly, Melo observed, there’s a rapid transition occurring within the AI sector from a GPU-dominated industry to practical parity with CPUs.
“The turning point for enterprises is that agentic AI workflow,” he said.
Specifically, the promise is AI agents that can autonomously monitor healthcare patients, detect fraud at financial institutions, intercept cybersecurity threats, optimize supply chains and more.
Nutanix’s 2026 Enterprise Cloud Index found that 61% of IT executives expect AI agents to enhance customer experiences. Another 58% say they’ll improve productivity, and 57% expect them to create new products or services.
If agentic AI is the defining technology of the moment, which Melo believes it is, its deployment has once again reemphasized the importance of CPUs in the world of machine learning. He said GPUs alone can no longer be the primary driver of an AI IT strategy.
“Looking ahead, it’s going to be exciting times,” he said.
Related:
The Rapid Rise and Future of Neoclouds
Enterprise AI Demand Brings Hardware Scarcity Into 2027
Chase Guttman is a technology writer, an award-winning travel photographer, Emmy-winning drone cinematographer, author, lecturer and instructor. His book, The Handbook of Drone Photography, was one of the first written on the topic and received critical acclaim. Find him at chaseguttman.com or @chaseguttman.
Ken Kaplan contributed to this story. He is Editor in Chief for The Forecast by Nutanix. Find him on X @kenekaplan and LinkedIn.
© 2026 Nutanix, Inc. All rights reserved. For additional information and important legal disclaimers, please go here.
Related Articles
