Investing into making their stuff work with the abstracted libraries. Pytorch etc. Collaboration with Microsoft to make their hardware easier to get time on and test with. Starting to support tooling on their lower tier consumer GPUs.

There have been several announcements:

"Azure announces new AI optimized VM series featuring AMD's flagship MI300X GPU" https://news.ycombinator.com/item?id=38280974

Doubling down on ROCm: https://www.theverge.com/23894647/amd-ceo-lisa-su-ai-chips-n...

Putting resources into pytorch: https://pytorch.org/blog/experience-power-pytorch-2.0/

Build a developer flywheel by making ROCm more accessible.

https://www.tomshardware.com/pc-components/gpus/amd-arms-thr...

That said, I'm approaching this from the other end, by make their high end GPUs available to developers.