CUDA emulator for AMD GPUs Zluda loses funding with v6 release, adds 32-bit PhysX

The Zluda project, an open-source CUDA emulator enabling AMD GPUs to run CUDA applications, has released version 6 with new 32-bit PhysX support and enhanced Windows compatibility. Originally targetin... Read original →

Zluda’s loss of AMD funding reveals the fragility of open-source CUDA compatibility layers amid geopolitical tech fragmentation. Technically, v6’s PyTorch and 32-bit PhysX additions lack driver-level integration, failing to overcome HIP conversion bottlenecks—highlighting NVIDIA’s entrenched closed-stack advantage. Compliance-wise, U.S. AI chip export controls push non-U.S. firms toward alternatives, yet hobbyist-grade projects like Zluda can’t meet enterprise security or audit standards, raising deployment risks. Strategically, AMD will likely double down on native HIP, while Spectral Compute and MooreThreads pitch full-stack emulators to developers in sanctioned regions like Taiwan, China and mainland China. Without backing from major cloud providers or sovereign funds within 18 months, such efforts will remain proof-of-concepts—real control stays with vertically integrated players wielding 3nm and EUV capabilities.