A year ago, Intel (INTC 0.72%) looked like a company investors had given up on. Its stock briefly traded below $18, its foundry business was bleeding cash, and rivals had lapped it in the race to power artificial intelligence (AI). So far in 2026, though, the chipmaker has staged one of the more improbable comebacks in big tech, with shares up about 225% as of this writing.
Plenty of headlines have fueled the move. There's a reported preliminary agreement to make chips for Apple, a tie-up with Elon Musk's sprawling "Terafab" project, and a number of significant supply deals with cloud customers. But strip those away and one argument keeps surfacing as the reason the rally could have legs: a shift in where AI workloads run that may finally play to Intel's oldest strength.
Image source: Getty Images.
A turn investors didn't see coming
For years, the chip conversation was about graphics processing units (GPUs) -- the accelerators from Nvidia and others that train large AI models. Intel, whose bread and butter is the central processing unit (CPU) that anchors servers and PCs, was largely left out of the story. But the CPU is suddenly becoming a key component in the AI era.
When Intel reported its first-quarter results last month, the standout wasn't the headline revenue figure of $13.6 billion, up about 7% year over year. It was the data center and AI division. Revenue there climbed 22% year over year to $5.1 billion, and the segment's operating margin more than doubled to roughly 31% from about 14% a year earlier. Notably, this was an acceleration from 9% growth in the prior quarter.
Intel CEO Lip-Bu Tan tied the improvement directly to how AI is evolving.
"For the last few years, the story around high-performance computing was almost exclusively about GPU and other accelerators," he said during the company's first-quarter earnings call. "In recent months, we have seen clear signs that the CPU is reinserting itself as the indispensable foundation of the AI era."
The agentic AI bet
Sure, training an AI model leans heavily on GPUs. But once a model is built and put to work answering questions -- a phase called inference -- and especially once AI "agents" start orchestrating tasks on their own, CPUs do more of the coordinating. Intel's chief financial officer, David Zinsner, framed it in terms of a ratio: training setups might run eight GPUs for every CPU, inference closer to four to one, and agentic work could push that toward parity or even tip the other way.
If that plays out, the addressable market for Intel's core products could expand meaningfully just as the company's manufacturing is starting to find its footing.
Tan called the demand strong enough that it's outstripping what Intel can currently supply. The company's AI-related businesses already make up about 60% of revenue and grew about 40% year over year last quarter.
NASDAQ: INTC
Key Data Points
The downside here, however, is the stock's valuation.
After tripling, Intel trades at well over 100 times its expected earnings for the next year -- a multiple that prices in years of the agentic shift unfolding in Intel's favor.
Meanwhile, Intel's business is still working through some challenges. On a standard accounting basis, Intel still posted a loss of about $3.7 billion last quarter, weighed down by restructuring and a write-down tied to its Mobileye unit. Its foundry unit lost roughly $2.4 billion.
Yes, the CPU comeback is a massive catalyst for the business. And the Apple report, which the company hasn't confirmed, would be a notable vote of confidence in Intel's factories if it holds up.
But a stock priced for near-flawless execution leaves little cushion if the company's foundry business fails to ramp up fast enough or if the CPU-to-GPU ratio Zinsner describes shifts more slowly than hoped. For investors tempted to chase one of this year's biggest turnarounds, the more measured move may be to wait for a price that bakes in some of those risks rather than assuming they all break Intel's way -- something I think the current valuation largely does.
With all of this said, it's always possible that Tan is even underestimating the importance of the CPU in this agentic era of AI. For this reason, investors can't rule out the possibility that the stock continues to rise from here as the business outperforms even today's bullish assumptions. So, for the aggressive investor who believes we are still in the early innings of this AI boom, it may make sense to take a small position in the stock despite the risks, building it out to a full position if the stock takes a significant hit in the future.
