NVIDIA Drops CUDA-Oxide 0.1: An Experimental Rust-to-CUDA Compiler
NVIDIA finally lets Rustaceans write GPU code without drowning in unsafe blocks — meet CUDA-Oxide 0.1, a proof-of-concept compiler.

NVIDIA has answered the prayers of developers who wanted to write CUDA code in Rust but weren't ready to deal with unsafe blocks the size of a bedsheet. Enter CUDA-Oxide 0.1 — an experimental compiler that translates a subset of Rust into CUDA C++.
It's more of a proof of concept than a production tool right now. The compiler doesn't support all Rust features — closures and traits are still missing. But it handles basic types, loops, and even some pointers. In other words, it's like any 1.0 release: "works on my machine" (and not always there).
For developers dreaming of Rust on the GPU, this is an invitation to dance, not a wedding march. NVIDIA is hinting that if the community steps up with bug reports and feature requests, the project could grow. Meanwhile, you can try compiling hello world and feel like a pioneer.
Comment from METABYTE studio: If you want your code to run fast without surprises but aren't ready to wait for Rust to conquer all architectures, we can help with C++ or CUDA optimization. And if you do take the Rust plunge, we'll debug it with you.
NEXT STEP
Liked the approach?
We apply the same principles to client projects: AI, automation, products that don't die after launch.