Intel surges 20%, CPUs return to the spotlight in the Agent era

Last night, Nvidia’s stock price briefly touched $70 during intraday trading, then surged 20% after the market closed, because its latest earnings report beat everyone’s expectations.

Intel disclosed its results for the first quarter of fiscal year 2026 on Thursday: revenue of $13.6 billion, up 7% year over year, and 11% higher than Wall Street’s consensus estimate. Non-GAAP earnings per share were $0.29, compared with analysts’ expectations of $0.01—29 times the forecast—which is an extremely rare magnitude of surprise among large-cap stocks. After the news was released, Intel’s stock price rose by as much as 20% in after-hours trading.

The Q2 guidance also pointed in a more aggressive direction. Revenue guidance was between $13.8 billion and $14.8 billion, above the consensus estimate’s midpoint. The newly appointed CEO, Lip-Bu Tan, summarized the company’s performance during the earnings call with a single line: in essence, CPUs are moving back into an indispensable foundational position in the AI era.

This is one of the most hotly discussed propositions in the market about Intel over the past two years. The company had once been widely seen as having completely missed the first wave of AI.

On the one hand, it didn’t deliver GPUs that could match Nvidia. On the other hand, its advanced manufacturing nodes couldn’t keep up with TSMC. But over the past 12 months, as more AI deployments shifted from model training to inference and the orchestration of autonomous “agents,” the CPU—once viewed as a basic “computer brain”—has ironically become needed again. Intel’s rebound this quarter is the first financial validation of this technical narrative.

Data Center business turns into a U-shaped reversal

Breaking down the $13.6 billion in Q1, the most critical change comes from the Data Center and AI (DCAI) segment. According to Intel’s earnings report, DCAI generated quarterly revenue of $5.1 billion, up 22% year over year, setting a record high.

This isn’t a one-off burst. If we roll the timeline back to 2025, DCAI reached $4.1 billion in Q1, fell to $3.9 billion in Q2, and then returned to $4.1 billion in Q3. The sideways stretch around mid-2025 led the market to question whether the so-called “CPU revival” was merely a narrative. Then in Q4, according to the Intel disclosure figures compiled by Tom’s Hardware, DCAI jumped from $4.1 billion in Q3 to $4.7 billion—up 15% quarter over quarter. This was Intel’s fastest quarter-over-quarter growth in a decade.

Entering Q1 2026, that $5.1 billion figure drew a clear U-shape across the curve: the trough was in mid-2025, the turning point was in Q4 2025, and it was confirmed in Q1 2026. Management’s explanation was that the Xeon 6th-generation “Granite Rapids” processors began ramping at scale, combined with an AI infrastructure refresh cycle. The company even voluntarily sacrificed part of its client CPU production capacity—allocating wafer supply to data centers—to raise the overall profit margin of the entire DCAI segment. According to Intel’s Q3 2025 earnings report, the operating margin for this segment rose from 9.2% in Q3 2024 to 23.4%, nearly 2.5 times higher.

Same AI narrative, three different trajectories

If you put Intel’s rebound into a peer comparison, you’ll see a chart that’s more interesting than just looking at percentage gains and losses.

Using January 2023 as the baseline, by April 2026 Nvidia’s stock index has climbed to 1023, AMD has risen to 406, and Intel sits at 245. The three lines start from the same point, but the endpoints differ by nearly five times. What’s even more worth watching is the shape of Intel’s blue line: it doesn’t rise gradually. Instead, it first drops all the way down to 64 in September 2024 (equivalent to a 36% decline from the starting point), then draws a V-shaped rebound, before finally catching up to 245 in early 2026.

In fact, this chart is really describing two rounds of market repricing of “who truly makes money during the AI capital cycle.” From 2023 to 2024, money flowed to Nvidia because training requires GPUs. AMD took the second slice with its MI300 series, and its stock moved along. Intel, however, was systematically crossed off the AI trading shortlist because Gaudi accelerator sales came in below expectations and advanced process mass production was delayed. According to third-party estimates cited by Fortune in January 2025, Nvidia’s share in the AI chip market rose from 25% in 2021 to 86% in 2024, while Intel’s share fell from 68% to 6%.

The second repricing happened in the second half of 2025 into early 2026. As the market began to revisit the question of whether the demand structure for compute would change if AI moved from training into the inference and Agent stages, the answer to that question directly determines how far Intel’s blue line can go.

The closer the scenario gets to Agents, the more CPUs return to center stage

When you break AI workflows into three categories of scenarios, CPUs’ weight differs dramatically across them. According to Deloitte’s 2026 technology trends report, in the large-model training stage, CPUs account for only about 8% of the bottleneck in workflows; the remaining 92% of compute pressure is on parallel synchronization in GPU clusters—Nvidia’s home turf. Once you enter the large-scale inference stage, CPU weight rises to 25%, but GPU parallel throughput and memory bandwidth still remain the bottlenecks.

The real change happens in the Agent orchestration scenarios. According to a study jointly published by Georgia Tech and Intel in November 2025, the CPU processing used for tool calls in an Agent workflow accounts for 50% to 90% of total end-to-end latency, depending on the tool type and the complexity of orchestration. In other words, when an AI agent performs things like “calling APIs, fetching data, coordinating subtasks, and managing context memory,” the bottleneck isn’t on the GPU—it’s on the CPU.

This trend has measurable scale to reference. Deloitte estimates that the share of AI total compute attributable to inference workloads was about 1/3 in 2023, about 1/2 in 2025, and is projected to reach 2/3 by 2026. According to Futurum Group, the server CPU market size will grow from $26 billion in 2025 to $60 billion by 2030, with a growth rate exceeding the historical long-term average. A more specific signal is OpenAI’s disclosed compute roadmap: the company plans to obtain “hundreds of thousands of the most advanced Nvidia GPUs, and scalable compute power for tens of millions of CPUs to support Agent workloads.” GPUs remain the top player, but for the first time, CPU scale is publicly listed in the same line.

The rebound didn’t start in Q1 2026

If you stack Intel’s share price over the past five years together with six key events, the 20% after-hours jump in Q1 is actually the tail end of decisions made earlier.

In February 2021, Pat Gelsinger returned to serve as CEO, rolling out the “IDM 2.0” strategy to make Intel both a chip design company and an open foundry serving external customers. When Gaudi 3 was launched in April 2024, Intel set a target for 2024 AI accelerator sales of $500 million.

On August 2, 2024, Intel’s Q2 2024 earnings report sent a shock: revenue fell year over year, with revenue of $12.8 billion; GAAP earnings per share were -$0.38; the company announced 15% layoffs and paused dividends. The stock dropped 26% in a single day—the worst one-day performance since 1974. As Intel disclosed at the time, management later admitted that Gaudi 3 could not meet the $500 million annual target and took a $300 million inventory write-down.

According to Intel’s official announcement, Gelsinger resigned on December 1, 2024, and the company entered a temporary co-CEO stage. In February 2025, the new management decided to cancel the independent GPU project “Falcon Shores,” acknowledging that its in-house AI accelerator roadmap could not match Nvidia’s ecosystem lock-in. On March 18, 2025, former Cadence CEO and a veteran semiconductor executive, Lip-Bu Tan, officially became Intel’s CEO. Around that time, Intel’s stock price was near $22—only a little more than a fifth higher than the $18 low in September 2024.

From the time Lip-Bu Tan took office to this Q1 earnings report, Intel’s stock rose from $22 to nearly $65 before the report was released. Adding the 20% after-hours move means it had just reached around $78. If August to December 2024 was the company’s darkest period, then the real start of the rebound wasn’t in Q1 2026—it was the moment it canceled Falcon Shores and selected Tan as CEO. The company gave up its fantasy of matching Nvidia and returned to its true CPU stronghold.

A 29x EPS beat is a financial signal, but behind it there are actually two things happening at the same time. The market began to reprice the position of CPUs in AI architectures, and Intel happened to complete the management leadership transition while making product line trade-offs. Neither of these two things occurred in Q1.