What Local AI Chips Mean for the Next Wave of PCs
Local AI chips are processors designed to run advanced artificial intelligence models directly on personal computers, enabling on-device AI processing for tasks such as language generation, image creation, and autonomous software agents without relying on remote cloud servers. This shift changes what a PC can be: instead of sending data to distant data centers, AI PCs can respond instantly and keep sensitive information on the user’s machine. For consumers, that means lower latency, fewer worries about constant internet connections, and stronger privacy by default. For the industry, it signals a move away from AI being an enterprise-only feature toward AI-powered PCs that ship as standard. Nvidia’s RTX Spark processor and Euclyd’s planned Craftwerk architecture sit at the center of this AI PC competition, each offering a different path to bringing powerful local AI to everyday laptops and desktops.
Inside Nvidia’s RTX Spark: AI PCs Built for On-Device Processing
Nvidia’s RTX Spark processor is described as a “superchip” aimed at turning conventional laptops and desktop computers into full-fledged AI PCs with strong on-device AI processing. According to TechDigest, Nvidia’s chief executive Jensen Huang said the RTX Spark will ship this year in systems from Dell, Lenovo, Asus, and HP running Windows, with the goal of making AI “replace the mouse and keyboard” as the main way people use computers. The chip is designed to run large language models, AI agents, and creative AI tools locally, avoiding slow or unreliable cloud links and keeping data on the device for greater privacy and speed. Nvidia is also preparing its Vera Rubin architecture, built on a 3-nanometer platform with 288 gigabytes of HBM4 memory, to power future Spark systems and even large-scale “AI factories,” with mass production planned toward the end of the decade.
Euclyd’s Craftwerk: An Energy-Efficient Challenger to Vera Rubin
While Nvidia builds out the RTX Spark ecosystem, Euclyd is positioning itself as a challenger with a different vision for local AI chips. Founded by Bernardo Kastrup, the company is developing its Craftwerk architecture, which it claims could be up to one hundred times more energy-efficient than Nvidia’s Vera Rubin-based systems. Craftwerk is planned as a massively parallel design using 16,384 tightly connected processors that together target a computing power of 32 petaflops. Euclyd is seeking 100 million euros in growth capital and aims to begin commercial production in 2028, targeting AI inference workloads that currently rely heavily on Nvidia hardware. If its claims hold up in real-world deployments, Craftwerk could give PC makers and data centers a highly efficient alternative for on-device AI processing, potentially reshaping how AI PCs and local AI infrastructure are built and powered.
Privacy, Latency, and the Shift Away from the Cloud
Both RTX Spark and Euclyd’s Craftwerk architecture focus on the same strategic goal: moving AI workloads from centralized cloud servers to local devices. Running models on a PC removes the latency of sending prompts and data across the network, which is critical for real-time assistants, creative tools, and interactive agents. It also reduces dependence on large data centers, which have drawn complaints over noise, energy use, and local impact, as seen in reports of residents frustrated by constant data center hum near suburban neighborhoods. For consumers, on-device AI processing promises that sensitive documents, voice data, and images can stay on their own machines instead of in remote systems. This local AI shift also gives PC makers more control over user experience, as they are less bound to subscription-based cloud services and can differentiate through hardware and software integration.
AI PCs Go Mainstream: Market Implications of RTX Spark vs. Euclyd
The emerging AI PC competition between Nvidia and Euclyd signals that advanced AI is moving from specialized enterprise setups into mainstream consumer PCs. With RTX Spark set to appear in laptops and desktops from major brands, AI-capable hardware will soon be part of standard configurations rather than niche add-ons. At the same time, Euclyd’s long-term plan suggests a market where energy efficiency and cost of inference become key differentiators, especially as AI workloads scale. Both companies are positioning their chips as alternatives to cloud-based AI services, giving developers and users more choice in how they deploy models. This could encourage software that is designed from the ground up for local AI, similar to how past shifts in graphics and CPUs created new application categories. The result is likely a new baseline expectation: future PCs will be AI-ready out of the box.
