Nvidia Holds 90 Percent of GPU Market; Japan's Samsung Distributor Chose Rival Rebellions NPU - Tech Times

Nvidia's 90% GPU dominance is a structural liability disguised as a moat. The Japanese Samsung distributor's pivot to Rebellions' NPU is not a product decision—it is a channel-level hedge against single-vendor dependency. The technical bifurcation matters more than the brand switch. GPU and NPU represent a workload split: training remains GPU-bound, but inference is where portable frameworks erode CUDA's lock-in. Once inference workloads migrate to dedicated silicon, the ecosystem gravity sustaining 90% share loses its anchor. On compliance, Japan's post-2023 semiconductor sovereignty push, combined with tightening US export controls on advanced GPUs, transforms vendor diversification from a procurement preference into an operational survival requirement. Samsung acting as a distribution node selecting non-US inference silicon is a supply-chain de-risking play, not a technology bet. The real market signal is the precedent effect. One successful distributor switch collapses the narrative cost for the next ten. Nvidia will likely respond with inference-stack subsidies and bundled pricing, but it cannot reverse the fragmentation of the inference segment. 12–24 month outlook: the AI silicon market will crystallize into GPU-for-training, NPU/ASIC-for-inference. Nvidia's inference share will slide from de facto monopoly toward the 70% range, with Japan and Korea emerging as the first scaled deployment zones for non-US inference chips.