Intel tries to loosen Nvidia’s AI grip with chips that keep computing off the public cloud - Ynetnews

At Computex 2026 in Taipei, Intel unveiled an ambitious AI chip strategy aimed at challenging NVIDIA’s dominance by promoting local computing capabilities. The company introduced the Core Ultra 3 and ... Read original →

Intel’s Computex 2026 (Taiwan, China) push—anchored on its 18A node and Core Ultra chips—is a structural counteroffensive against NVIDIA’s GPU-centric AI hegemony. Technically, the OpenVINO + Core Ultra X7 358H stack not only halves cost and latency versus Jetson AGX Orin but redefines edge AI stacks: CPU-NPU co-processing displaces pure-GPU inference, forcing compiler and quantization toolchains to adapt. From a compliance angle, on-device processing sidesteps tightening cross-border data rules in the EU and U.S., cutting cloud dependency for enterprises. NVIDIA will likely accelerate Jetson Thor ecosystem lock-in via CUDA, yet rising AI agent adoption could erode that moat as hardware heterogeneity grows. Over the next 18 months, the battleground shifts from raw compute to 'edge intelligence density'—the winner controls the low-power, integrated Physical AI platform that becomes the gateway to ambient intelligence.