AI Kernel gen is easy with the rapid advancement of coding agents - if you have a compiler and dsl like triton/gluon.

The real metric is end to end enablement of the entire stack, plus unlocking/integrating a developer community that can share code from NVIDIA ecosystem👇

Modular (@Modular)
Support for new AI hardware usually takes a big team, many repos, and a year.

Two and a half engineers got a frontier open model serving on @Qualcomm Cloud AI 100 in under six months using the Modular stack, followed by GPT-2 on Qualcomm Dragonfly™ AI 200 in a week.

Inside the bringup at ModCon 2026: yewtu.be/watch?v=orJP0cxf…
Link
Inside the Qualcomm NPU Bring Up: A Technical Deep Dive (yewtu.be/watch) Bringing up new AI hardware usually means a huge team, 15 repos, an... youtube.com