Alternative(s) to run CUDA on non-Nvidia hardware

Spectral Compute has developed SCALE, a compiler tool that allows CUDA code—the dominant programming language for AI and HPC applications—to run on non-Nvidia hardware including AMD GPUs and other accelerators. Founded by former HPC engineers frustrated with Nvidia's hardware lock-in and expensive GPUs, the company aims to break Nvidia's dominance in the market while also optimizing performance on Nvidia's own chips. CUDA currently controls approximately 80% of parallel computing development tools, making Spectral's effort to democratize its use across different hardware platforms potentially significant for the industry.