Nvidia's once-unassailable hold on the AI chip market is under mounting assault from rivals, tech giants, and even the software it helped create.
Nvidia still sells roughly 80 percent of the world's AI accelerators, a lead built on years of foresight from founder Jensen Huang. That dominance is narrowing fast. AMD has climbed from a two percent share in 2022 to as much as ten percent this year, while Google, Amazon, and Microsoft are all designing their own silicon to cut costs and loosen their dependence on Nvidia.
The pressure is global. In China, where Nvidia is barred from selling its most powerful chips, its slice of company revenue has collapsed to eight percent, down from twenty percent in 2024, even as Huawei rolls out supercomputers engineered to train AI models on domestic hardware. Startups like Cerebras and AMD's newly acquired Taalas are pushing specialized designs that promise to beat Nvidia's core products on efficiency.
The subtler threat comes from AI itself. New models can now read and rewrite code so efficiently that they port workloads onto a rival's hardware within hours, chipping away at the software advantage long treated as Nvidia's deepest moat. With finance chief Colette Kress now steering the company after founder Huang's abrupt departure amid controversy, some analysts warn Nvidia could be heading for an Intel-style stumble of its own making.