

Not only are Nvidia’s AI chips faster and more powerful than most other chips out there, Naji said the firm has also “developed the broadest ecosystem” of developers and software.
“And so it’s just so much easier to … build an application, build an AI model on top of those chips,” he said.
Where are the other semiconductor companies? The Intels and AMDs of the world, surely they can build an AI chip and software somebody wants?
The article first mentions the reason Intel and AMD has a hard time competing (the development/software for AI are all focused on running on Nvidia and CUDA) and then asks where they are?
The chips are here already… well almost
https://www.amd.com/en/products/graphics/workstations/radeon-ai-pro/ai-9000-series/amd-radeon-ai-pro-r9700.html
https://www.intel.com/content/www/us/en/products/sku/243916/intel-arc-pro-b60-graphics/specifications.html
They will most definitely be cheaper than the 5090 though, and if 640GB/s is good then the AMD card could be interesting as it matches the 32GB. But then it falls down to software support and Nvidia gets a really good lead in usability again.
And Intels card is even further behind: