Like a lot of the commenters here, I have a problem with this headline. They don't "seek to become a platform company", as in making their own cloud platform where they rent GPU time, meaning they stop selling GPUs to other cloud platforms. That easy misinterpretation makes good clickbait, but no, that's not what the article says - the article has Huang bragging that CUDA already is a parallel computing platform, for a decade or more, and Blackwell Architecture is so integrated and customizable with CUDA (with all its user-extendable kernels and community) that it's thought of as a platform rather than just chip architecture.
"Platform" can have several different meanings and some people in this thread are picking the most evil meaning as an excuse to shit on Nvidia. It's true that Nvidia is providing the GPU, the CPU, the server, the rack, the network, the drivers, and the orchestration software to run AI training/inference. (If you want that stuff. You could just buy GPUs if you want.) It's fair to call that a platform.
Nvidia is not becoming a cloud provider (beyond a small eval environment perhaps).
