<p>➀ Modular's team of 120 has been working on a CUDA alternative for three years, aiming to replace the entire AI software stack from scratch.</p><p>➁ The existing AI software stack has issues due to rapid evolution and the addition of layers to keep up with new use cases and models.</p><p>➂ Modular's AI inference engine, Max, launched in 2023, now supports Nvidia GPUs, offering a full-stack replacement for CUDA.</p>