“The NVIDIA RTX Spark chip is capable of delivering 1 petaflop of AI compute, compared to an H100 which outputs around 4 petaflops.”