Intel's blowout Q1 2026 earnings, powered by CEO Lip-Bu Tan's argument that agentic AI is swinging compute demand back toward CPUs, sent SanDisk stock jumping 6.5% as investors connected the dots between inference-era hardware and memory demand.
For most of the past three years, the semiconductor story wrote itself: Nvidia printed money on GPU demand as every major AI lab raced to train the next frontier model. Intel sat on the sidelines, its Gaudi AI chip failing to gain traction, its Falcon Shores GPU quietly abandoned. That narrative shifted sharply this week. On Intel's Q1 2026 earnings call, CEO Lip-Bu Tan delivered his sixth consecutive quarterly beat and made the case that the AI market's center of gravity is moving. The training boom that made Nvidia untouchable is giving way to inference and agentic workloads, and those workloads, he argued, are where CPUs come back to life. Intel shares surged 27% on the news. SanDisk, which barely registers in most AI hardware discussions, jumped 6.5% the same morning, because investors understood immediately what the shift means for memory.
The mechanism is straightforward. When AI models are being trained, the compute ratio inside a data center runs roughly seven or eight GPUs for every CPU. Intel CFO Dave Zinsner told analysts that for inference workloads, that ratio drops to three or four GPUs per CPU. As agentic AI scales further, Zinsner said the ratio could reach parity or flip entirely in the CPU's favor. More CPUs per rack means more adjacent demand for the fast NAND flash memory that sits alongside them. SanDisk, which spun out of Western Digital in 2024 and has spent the past 18 months positioning itself as the AI era's memory supplier of choice, is the clearest direct beneficiary of that shift.
The memory specialist's trajectory over the past year has been extraordinary by any standard. SanDisk stock rose 559% in 2025, making it the top S&P 500 performer. It has gained another 142% in 2026 year-to-date. Its fiscal second quarter results, reported in February, showed data center revenue growing 64% sequentially to $440 million, and for the first time in its history the data center segment is on track to become SanDisk's largest market in 2026, surpassing traditional consumer markets in PCs and smartphones. As Reuters reported in January, the company forecast revenue of $4 billion at the midpoint for the following quarter, against Wall Street expectations of $2.77 billion. Those are not incremental beats. They reflect a fundamental repricing of what NAND flash memory is worth in an AI infrastructure context.
The agentic AI dynamic amplifies this further. Agentic systems, where AI models execute multi-step tasks autonomously rather than responding to a single prompt, generate far more intermediate storage requirements than traditional inference. Every reasoning step, tool call, and context retrieval in an agentic workflow touches storage in ways that a simple query-response model does not. Nvidia has already flagged what it calls
Also read: China's crackdown on US investment in AI firms signals a new front in the tech cold war • Starlink's billions are quietly bankrolling the AI race from orbit • A 21-year-old cyberweapon called Fast16 just rewrote the history of state-sponsored sabotage
