Irony
The last paragraph's irony is spot on, since VMware's current infrastructure software already adds a compute tax of 10-25%. VMware has little interest in controlling that since host sprawl sells more licenses.
SmartNICs have tremendous potential but will suffer from what's possible versus what's practical. E.g., today's NICs and ToR switches have all sorts of capabilities and optimizations that mostly go unused because software solutions built around them are are seen as too complex, brittle and choice-limiting. SmartNIC success depends on them being indispensible to a single "killer app". VMware's NSX aims to be that app but it hasn't been the runaway success VMware has hoped for.
I'm also curious how enterprise vendors see SmartINC, since it will reduce their ability to differentiate. For example, as Palo Alto and Fortinet and VMware NSX implement their firewall stacks on DPU, their dataplane performance, capabilities, and limitations will look increasingly similar. Current competitive advantages -- like PAN's ability to inspect overlay tunnels -- either won't translate to a DPU at all, or it will translate and suddenly be available to all vendors.
Biting the hand that feeds IT
