HACKER Q&A
📣 thetrimtab

Why don't quant hedge funds have hardware labs?


If I'm a hedge fund, I want to have data that (a) is useful for figuring out what to buy and sell and (b) other hedge funds don't have. The latter lets me outcompete hedge funds. Hedge funds even have systems for people (mostly brokers/investment banks) submit ideas to them on what to trade, and get rewarded if it's profitable.[1] Numerai has an explicit platform for this.[2]

Why, then, don't hedge funds run their own hardware / R&D labs that design and deploy sensors to get data about the world other hedge funds don't have? Why don't hedge funds build and launch a bunch of cheap satellites with interesting sensors that only the hedge fund has access to? Or footfall sensors in major cities? Or agricultural sensors in target geographies?

Taken to the extreme, why don't hedge funds build e.g. Twitter to figure out what people are talking about and why, but then not let anyone use the API? Or why didn't they build Google Streetview to figure out what was in the world and where?

In other words: why don't we see more hedge funds trying to go get information about the world themselves instead of buying it from data vendors?

I guess the best answer is the return doesn't justify the cost. Better to just buy thousands of different datasets from different data vendors instead.

But is there maybe a different answer where it's just "culturally, that's not what people at hedge funds do or think about"?

[1] https://en.wikipedia.org/wiki/Alpha_capture_system [2] https://signals.numer.ai


  👤 fiat_fandango Accepted Answer ✓
I mean... you should dig a bit deeper. Jane Street (a top quant fund in NYC) has done exceptionally deep work with Ocaml and more specifically building pipelines AND custom hardware to run Ocaml on FPGA's and ASIC's.

👤 xodjmk
It's called 'Financial Technology', but if you really want to be cool, you can say 'Fintech'.

👤 version_five
Tldr, I think risk and laziness

I actually work in a very tenuously related field, and I've seen that their is a reluctance for people whose world is basically "what you can do on a computer" to actually try and get their hands dirty and get some data themselves, vs continuing to churn over whatever data is already out there- that's the laziness

The other important thing is the risk. It's easy to analyze existing dataset X and decide if it has predictive power. Throwing up a network of sattelites or sensors or whatever without clear knowledge that they are going to help is a massive risk that's probably beyond most asset managers.

Third, if someone is successfully doing something like what you're mentioning, they would be best served by doing it quietly, so there may be lots of cools stuff happening that we just aren't aware of