Nvidia banks on AI PCs and expands sovereign AI infrastructure with NAVER

Reuters reports Nvidia is wagering on demand for AI PCs that has yet to materialize beyond niche power users. The pitch: Nvidia-powered Windows laptops and desktops could rival Apple Macs on memory bandwidth, the bottleneck that most constrains local AI inference, narrowing the long-standing gap to Apple's unified-memory silicon. This dovetails with Nvidia's RTX Spark Superchip push to run agentic workloads locally without cloud dependency.

On the infrastructure side, Nvidia announced that NAVER is expanding its sovereign AI buildout on the Nvidia DSX platform, constructing production-scale 'AI factories' to serve Korean industries and global customers. Sovereign-AI deals — where nations or national champions build domestically controlled compute — have become a key Nvidia growth vector amid export controls and data-residency demands.

The AI-PC bet is the riskier one. Skeptics note that consumer demand for on-device AI remains speculative, and that most users still reach for cloud models. Nvidia's advantage is its software moat (CUDA) and the ability to position laptops as developer and prosumer machines for local agents. Community enthusiasm exists at the high end — Nvidia AI amplified a builder's local-AI rig boasting '16x DGX Spark, 3x RTX Spark' — but that's hardly the mass market.

The NAVER deal, by contrast, is concrete revenue and reinforces Nvidia's dominance in nation-scale buildouts. Watch whether AI-PC attach rates justify the marketing spend, and how memory-bandwidth claims hold up against Apple's M-series and the new Apple Intelligence/Gemini stack.