Banks pay HW engineers a lot. I know a dude at JPMC (nyc) who is clearing 450k doing HW (dc/priv cloud) design. I wouldn't be surprised if folks at somewhere like two sigma make around the same.
Well, if latency is the issue, most likely the software just loads data, models and supervisory systems (telemetry, setup, etc). All the processing must be done in the FPGAs straight to the firehose.
I should clarify, by private cloud, I mean a few custom designed servers running open stack or something, and DC I mean some built up "pod/rack" design you might colo. Ether way, I mean end to end, I was just trying to account for colo and on-prem (two sigma, for example, have a small private cloud in their office that's quite well architected.)
My guess is that he's talking about design custom equipment for efficiency and performance. It's a movement akin to moving from EC2 to in premises: if you're big enough, it starts making sense to custom build, often with in-house development.
Many of your recent comments are asking for clarification of acronyms. While I sympathize with the desire to understand, it would be good practice to take a stab at it yourself (my preferred search engine is a common starting point for me—they're remarkably good at narrowing in on an answer given the acronym and some context); and ask for confirmation or clarification if it's still necessary.