Reported by Song Jianan, Tech Journalist
On June 1, the chip industry witnessed a truly dramatic showdown—Nvidia made a bold entry into the PC chip territory that Intel has owned for decades, while Intel aggressively ramped up its presence in data center and AI chip markets, Nvidia’s home turf.
At the Computex Taipei 2026 GTC conference, Nvidia unveiled a brand-new super chip called RTX Spark, built for Windows PCs. Packing an NVIDIA Blackwell RTX GPU with 6,144 CUDA cores and fifth-gen Tensor Cores, this move marks Nvidia’s official leap into the core processor market for personal computers, directly challenging the long-standing duopoly of Intel and AMD.
This chip was co-designed by Nvidia and MediaTek, using TSMC’s cutting-edge 3-nanometer process. It integrates a Blackwell RTX GPU with a 20-core Grace CPU, fully compatible with Microsoft’s Arm-based Windows OS. In terms of graphics performance, the RTX Spark is on par with the RTX 5070 mobile GPU, and its CPU performance reportedly hits the top tier in the Windows ecosystem, ready to go toe-to-toe with mainstream rivals.
Starting this fall, major PC makers like Dell, Lenovo, and ASUS will roll out laptops and desktops powered by RTX Spark, according to the company’s roadmap.
During his keynote, Nvidia CEO Jensen Huang said the PC industry, after more than 40 years since the Windows 95 era, is hitting a fresh inflection point. AI agents are reshaping the PC landscape, and together with Microsoft, Nvidia is “reinventing the personal computer”—giving local PCs the ability to host standalone AI agents and truly bring artificial intelligence to edge devices.
Industry insiders believe the RTX Spark is Nvidia’s strategic move to deeply integrate CPU, GPU, and on-device AI computing. By leveraging its expertise in AI compute, Nvidia could shake Intel’s traditional dominance in the x86 PC market at the architectural level, pushing the PC industry from conventional computing toward an AI-first paradigm.
Right as Nvidia dropped its PC chip bombshell, Intel counterpunched with major product news and a future roadmap. That same day, Intel officially launched its new Xeon 6+ processor—its first data center CPU built on the Intel 18A process node. Designed for cloud-native workloads, agent-based AI, and network-intensive tasks, it delivers improvements in performance density, energy efficiency, and operational scalability.
Kevork Kechichian, Intel’s Executive Vice President and GM of the Data Center & AI Group, emphasized that scaling AI is not just about piling up hardware components—it’s about system-level orchestration. In the age of AI agents, task scheduling, concurrent processing, and data flow have become new bottlenecks, and the CPU remains the central control plane for modern AI infrastructure.
Additionally, Intel revealed a clear AI chip roadmap. The company said it plans to launch a new AI chip called “Crescent Island” by the end of 2026, focusing on accelerating AI inference tasks. This is a deliberate attempt to sidestep Nvidia’s stranglehold on the model training market and differentiate through cost-effectiveness. The big selling point: Crescent Island uses a lower-cost memory configuration and air-cooling technology, drastically cutting hardware deployment and maintenance costs compared to Nvidia and AMD’s comparable products. That could give it a solid edge in the mid-range AI inference segment by offering strong bang for the buck.
For Intel, doubling down on data center CPUs and budget-friendly AI chips is both a natural move to round out its product portfolio and shore up its data center stronghold, and a strategic counterpunch against Nvidia’s cross-border assault.
For a long time, Intel has ruled the roost thanks to its PC chip business and data center CPUs. But in recent years, Nvidia’s AI GPUs have stormed the scene, steadily eating into the high-end compute market. Now, Intel is reinforcing its grip on the data center with a 18A process chip and carving out an inference niche with low-cost AI silicon—all in an effort to bust Nvidia’s monopoly on AI compute and build a complete compute ecosystem from edge to data center, from training to inference. Meanwhile, Nvidia’s foray into PC chips directly targets Intel’s core revenue source, which will force Intel to accelerate architectural innovation and technology iteration.
The clash between these two titans is bound to trigger a deep restructuring of the global chip landscape. On one hand, the PC chip market will no longer be a two-player game between Intel and AMD. Arm-based chips, riding the AI wave, will penetrate faster, and the battle between x86 and Arm architectures will heat up. OEMs will have more chip options, which could also speed up the adoption of AI PCs.
On the other hand, with its process advantages, ecosystem legacy, and cost control skills, Intel could siphon off market share in AI compute and high-end inference. That’s good news for cloud service providers and enterprise users looking to cut compute procurement costs. At the same time, this cross-industry rivalry will drive upstream and downstream collaboration all along the chain—from process tech and chip architecture to system manufacturing and AI software development—potentially kicking off a new cycle of innovation across the industry.
However, given Nvidia’s current dominance in the AI chip space, turning the tables anytime soon won’t be easy for Intel. Nvidia itself is racing to build a comprehensive AI ecosystem across the entire value chain.
Beyond announcing the PC chip, Nvidia also dropped several other important products on the same day. According to Jensen Huang, Nvidia’s next-gen Vera Rubin platform has already entered full production. Vera Rubin uses HBM (High Bandwidth Memory) from Micron, SK Hynix, and Samsung, providing a rack-scale (POD) all-in-one AI factory foundation. Nvidia says the supply chain it has built for Rubin is “twice as large” as its predecessor, Blackwell. While it used to take two hours to assemble a massive Blackwel rack, Vera Rubin now takes just five minutes.
Additionally, Nvidia launched a new open-source AI model and tech stack called Nemotron 3 Ultra. It not only offers a mature AI model but also comes with training data and development tools—lowering the bar for developers to build AI agents and further strengthening Nvidia’s own AI ecosystem.
