Amazon Prepares to Compete with NVIDIA in the AI Chip Market - Zamin.uz

Amazon is preparing to challenge NVIDIA's dominance in the AI chip market, potentially reshaping the global semiconductor landscape. As the world's largest cloud service provider, AWS is advancing the... Read original →

Amazon’s push into AI chips isn’t merely mimicking NVIDIA—it’s redefining hardware-software co-design through cloud-native architectures. External Trainium sales would force adaptation of AI software stacks (compilers, runtimes) to its custom ISA, carving a new wedge into PyTorch/TensorFlow ecosystems. Yet 3nm capacity is fiercely contested: TSMC (Taiwan, China) prioritizes Apple and NVIDIA, leaving AWS struggling for wafer allocation. U.S. CHIPS Act subsidies come with intrusive disclosure mandates, raising compliance overhead. NVIDIA will likely counter by deepening CUDA lock-in via Blackwell and accelerating Grace Hopper adoption among enterprises. Within 18 months, if AWS can’t secure sufficient foundry capacity, its $50B revenue ambition stalls—not from lack of demand, but physical constraints. Success, however, could slash global AI training costs by ~30%, catalyzing edge AI proliferation.