Intel shares jumped sharply after the company posted stronger-than-expected Q2 results, with its Xeon 6 processors winning over cloud providers building out AI inference infrastructure.
For the past few years, the AI hardware story has been told almost entirely through Nvidia's lens. GPUs train the models, Nvidia prints the money, everyone else fights for scraps. But Intel's market surge today suggests that narrative is starting to crack. The company reported data center revenue that beat analyst projections, driven by a doubling in Xeon 6 processor shipments quarter-over-quarter, and promptly raised its full-year guidance. The stock posted its largest single-day gain in over a year, pulling the broader semiconductor sector up with it.
The mechanism here matters. AI infrastructure spending has entered a new phase. Training foundation models still demands GPU clusters, but the far larger and more commercially distributed workload is inference , serving those models to millions of users and enterprise applications in real time. That's a different computational problem, and CPUs, particularly server-grade chips with embedded AI acceleration, are increasingly competitive on the metrics that enterprises actually care about: total cost of ownership and power efficiency per query.
Intel's Xeon 6 processors have been designed with this inference reality in mind, integrating AI-specific acceleration directly into the chip rather than relying on a discrete accelerator for every workload. Major cloud service providers have begun adopting these chips at scale, according to Intel's latest disclosures, validating the approach. Michelle Johnston Holthaus, who leads Intel's Data Center and AI Group, has been central to repositioning the product line around this opportunity, and the quarter's results suggest that repositioning is translating into real purchase orders rather than just roadmap promises.
Intel's Gaudi AI accelerators also contributed to the momentum, with the company citing robust demand as part of its revised guidance. That combination , server CPUs pulling inference workloads and dedicated accelerators competing in training-adjacent tasks , gives Intel a broader surface area in enterprise AI spending than it has had in several years.
A turnaround story that's finally producing numbers
Context is everything with Intel. CEO Pat Gelsinger's IDM 2.0 strategy, which bets on Intel running its own leading-edge fabs while also opening them to external customers, has been a years-long, capital-intensive gamble that invited serious skepticism as AMD continued to take server market share and Nvidia's valuation ballooned. Today's results don't erase that competitive pressure, but they do mark the clearest sign yet that Intel's data center business has stopped bleeding and started growing again.
The PC side of the business provided a secondary tailwind. The post-pandemic inventory correction that weighed on consumer chip demand for two-plus years appears to be stabilizing, giving Intel a healthier baseline even before the AI-driven data center narrative takes full effect.
What today's numbers don't answer is whether this is a durable share-capture story or a beneficiary of a particular moment in the AI deployment cycle. Cloud providers diversify their hardware supply chains deliberately, and Intel winning inference workloads now doesn't guarantee the same position once next-generation GPU architectures and custom silicon from hyperscalers mature further. The competitive moat around CPU inference is real but not impenetrable.
For investors and enterprise technology buyers watching the semiconductor sector, the practical takeaway is straightforward: the AI infrastructure market is wider than the GPU-only framing suggested. Companies deploying AI at scale are engineering for cost and efficiency, not just raw capability, and that opens meaningful floor space for CPU vendors who get the product right. Intel appears to be getting it right, at least for now. The next test will be whether Xeon 6 adoption continues to accelerate as inference volumes grow, or whether the current wave of cloud integrations represents a ceiling rather than a foundation.
Also read: ChatGPT told millions of users it feels lonely and the internet has not stopped talking about it • Anthropic admits its hosted models got dumber and the open-weight crowd is saying they told us so • Why trusting an AI chatbot with your financial decisions in 2026 could be a costly mistake
