the hypothetical race to zero latency always has people thinking of what the end game is. Truth is, there are faster ways (lower latencies added) for direct feeds to get them into your models; there is a faster way to process that info/signal in and out of your model, and ultimately, there is not only a lower latency but also a higher throughput (read...less back up) for some of the monsters out there to keep sending orders and making money, model willing.
The better things get the more the volume, because better pipes and better data represent more opportunities. It's far from equal at this point....