Why would you even need dedicated hardware for just 40 Gb/s? That is within single-core decryption performance which should be the bottleneck for any halfway decent transport protocol. Are we talking 40 Gb/s at minimum packet size so you need to handle ~120 M packets/s?

Because the entire stack is auditable here. There's no Cisco backdoor, no Intel ME, no hidden malware from a zombie NPM package. It's all your hardware.