4 ms·
I am curious what workloads Citadel Securities is running on these TPUs? Are you telling me they need the latest TPUs for market insights?
by nickandbro 6mo ago
I am curious what workloads Citadel Securities is running on these TPUs? Are you telling me they need the latest TPUs for market insights?
- vibe42 6mo agoTraining their own, closed, internal models on their own data sets? Probably a good way to squeeze out some market trading signals.
- nickandbro 6mo agoReminds me of when hedge funds started laying increasingly shorter fiber-optic cable lines to achieve the lowest possible latency for high-frequency trading.
- written-beyond 6mo agoI thought these TPUs were primarily used for inference?
- vlovich123 6mo agoTPU8t is for training. But even still, once you’ve trained, you need to run the model too. And these kinds of models already have a huge latency hit so there’s not much hurting running it away from the trading switches.
- knowaveragejoe 6mo agoAs the article states, there's both training and inference dedicated chips.
- dist-epoch 6mo agoNot Citadel, but Jane Street is training LLMs for trading: https://www.janestreet.com/join-jane-street/machine-learning/ https://www.janestreet.com/join-jane-street/machine-learning... > We build on the latest papers in LLMs, computer vision, RL, training libraries, cuda kernels, or whatever else we need to train good models. > We invent our own set of architectures and optimizations that work for trading.
- brcmthrowaway 6mo agoTruly an epic company.