Comment by kartoolOz

Vespa.ai is pretty crazy too, a bit unkown we run a huge vespa cluster serving 1k+ queries with <100ms latency ...