Microsoft has demonstrated a new, highly efficient “1-bit” AI model that can run on standard CPUs using only 0.4 GB of memory—significantly less than traditional models. This advance addresses a key bottleneck in deploying capable AI features on devices like laptops and smartphones, which are limited by memory and computational power. The model matches the performance of similarly sized AI models while requiring fewer resources and can operate at speeds comparable to human reading. This breakthrough could shift some AI workloads away from data centers, potentially benefiting chip makers like Intel and AMD, while reducing dependence on expensive accelerators popularized by companies like Nvidia.































