Google's TPUs are an example of efficient chips designed for AI workloads, and were in development before the LLM boom. It's just hard to replicate the power a

Google's TPUs are an example of efficient chips designed for AI workloads, and were in development before the LLM boom. It's just hard to replicate the power and efficiency of CUDA.