Intel’s Two-Front War: Nvidia Invades the PC as Xeon 6+ Fights for the Data Center

Intel shares drop 4.67% at Computex as Nvidia unveils Arm-based RTX Spark PC chip; Intel fights back with Xeon 6+ and new AI inference strategy.

Intel’s annual Computex showcase was supposed to be a victory lap for its next-generation data-center silicon. Instead, the stock took a 4.67% hit, closing at $109.33 on June 2, after Nvidia used the same Taipei stage to announce a direct assault on Intel’s PC heartland. The contrasting narratives — a promising CPU renaissance versus an encroaching Arm-based rival — left investors parsing whether the long-term AI story can offset near-term competitive pressure.

Nvidia’s RTX Spark: A Superchip Aimed at Intel’s Core Business

Nvidia, together with MediaTek, unveiled the RTX Spark, an Arm-based system-on-chip that packs a 20-core CPU alongside a Blackwell GPU with 6,144 CUDA cores. Jensen Huang called it a “reinvention of the PC,” targeting a market worth roughly $200 billion that Intel and AMD have dominated for decades. The chip delivers up to 1 petaflop of AI performance and is set to debut in laptops from Microsoft, Dell, HP, ASUS, Lenovo and MSI by autumn 2026.

The market’s reaction was immediate: Intel shares swung between $107 and $114 during the session, and the pre-market slipped further to $107.65. In Frankfurt, the stock traded at €93.01, down 1.04% on the day. The sell-off punished traditional processor stocks as Nvidia signaled it is no longer just a GPU supplier but a full-platform player that intends to rewrite PC architecture.

Intel Counters with Xeon 6+ and a New AI Inference Strategy

Intel’s Computex response centered on the Xeon 6+ processor, codenamed Clearwater Forest, built on the 18A process node — the same technology underpinning the company’s foundry ambitions. With up to 288 efficient cores, Intel claims a 2.5× performance lift over the prior generation and 45% better performance per thread per watt versus competitors. A single rack of Xeon 6+ CPUs can execute up to 150,000 agents simultaneously, underscoring Intel’s bet that the shift from model training to agentic AI will reshape data-center economics.

During the training era, the typical ratio was one CPU for every four GPUs. Intel argues that agentic inference flips that to roughly one-to-one or even fewer GPUs per CPU. To demonstrate, Intel teamed up with SambaNova, Vista Equity Partners and Cambium Capital for a live inference system: Xeon 6 handles orchestration and execution, SambaNova SN40 RDUs manage decoding, and Nvidia Blackwell GPUs handle prefill. The setup runs in a Vector Core Compute data center in Los Angeles, with Together.ai as the first commercial customer, clocking the fastest enterprise inference on the MiniMax-2.5 model.

Complementing the CPU push, Intel introduced Crescent Island, a data-center GPU built for token-heavy agentic workloads. It features LPDDR5x memory with up to 480GB capacity and runs at 350 watts in an air-cooled PCIe design. The company also expanded its 800-series Ethernet portfolio with the E835 controller and adapters, claiming 200 Gbit/s throughput and 1.9× better performance per watt than Nvidia equivalents.

Market Share Erosion and Analyst Divergence

The competitive threat is not just rhetorical. Intel’s share of the server CPU market slid to 54.9% in the first quarter of 2026, down from 64.4% a year earlier. Yet several investment banks raised their price targets, betting that the broader AI infrastructure buildout will create demand that Intel can capture — especially if rivals hit capacity constraints.

• Mizuho lifted its target from $124 to $128, maintaining “Neutral,” citing demand for agentic AI and potential supply tightness at competitors through 2027.
• Wells Fargo boosted its price target from $85 to $110, pointing to positive signals from AI data-center demand.
• Barclays increased its target from $65 to $100 but kept an “Equal Weight” rating, acknowledging the valuation risk.

Intel’s first-quarter revenue rose 7% year over year to $13.6 billion, with the data-center and AI segment jumping 22% to $5.1 billion and the foundry unit climbing 16% to $5.4 billion. Second-quarter guidance calls for revenue between $13.8 billion and $14.8 billion, with non-GAAP EPS of $0.20. The stock’s 52-week range — from a low of $18.97 to a high of $132.75 — highlights how far the turnaround has come and how much execution risk remains.

The 18A Yield Hurdle and the Autumn Countdown

The wild card remains Intel’s manufacturing turnaround. Morgan Stanley estimates the yield on the critical 18A process at roughly 50%, with Apple identified as the most important external customer. Intel’s roadmap — five nodes in four years — depends on steady improvements there. Simultaneously, potential bottlenecks in memory and CPU supply could constrain growth in the second half of 2026.

With a forward price-to-earnings ratio near 189 as of June 1, the market is pricing in a successful foundry pivot and strong AI adoption. By the time Nvidia’s PC platform launches in autumn, Intel must demonstrate tangible progress on 18A yields and tangible demand for Xeon 6+. If production stays on track, the AI narrative can support the premium valuation. If the manufacturing engine stalls, the new pressure from Nvidia’s Arm-based PCs will weigh far more heavily.