Jane Street keeps FPGA teams

Jane Street’s own hiring and engineering pages show the trading firm still maintains dedicated hardware teams for ultra-low-latency trading and larger machine-learning infrastructure. As of May 17, 2026, the firm was advertising FPGA engineer, ASIC engineer and ASIC physical design roles, alongside data-center jobs tied to GPU hardware, fiber cabling and liquid cooling. The company’s public materials do not support some of the more precise figures circulating on social media about its compute footprint. Jane Street’s machine-learning page says the firm uses “tens of thousands” of high-end GPUs and more than exabyte-scale storage, but it does not publish a count of 4,032 GPUs, 56 racks or 8,000 kilometers of private fiber on the pages reviewed. That leaves a narrower, verifiable story: Jane Street is publicly signaling that it still invests in both bespoke trading hardware and large-scale research compute. (janestreet.com) The evidence comes from the firm’s own recruiting language, engineering pages and technical materials. ### Where is the evidence that Jane Street still runs FPGA teams? Jane Street is currently hiring FPGA engineers in New York and London, according to job pages on its website. (janestreet.com) The New York posting says the firm wants engineers to “build our FPGA-based applications” and describes work on “ground-up design and implementation of new FPGA applications,” using OCaml, Hardcaml, Verilog and C. (janestreet.com) Jane Street’s performance engineering page also says the firm uses “FPGA accelerators” where CPUs alone are not enough. That same page says the company leads development of Hardcaml, its open-source hardware design library, and ties that work directly to performance-sensitive systems. A November 2025 Jane Street blog post on its “Advent of FPGA” challenge adds a more public marker of an active hardware culture. (janestreet.com) The post was written by Jane Street hardware engineers and promoted Hardcaml-based and other RTL submissions, with judging by the firm’s hardware team. ### Is this only FPGA work, or does Jane Street still have ASIC capability too? Jane Street is also advertising ASIC roles in New York and London. (janestreet.com) An ASIC engineer posting says candidates should be ready to work on “both FPGA-based and ASIC-based technologies,” while an ASIC physical design role in New York says the engineer would join the firm’s “Ultra Low Latency team.” The ASIC physical design listing is especially specific about where that work sits. (blog.janestreet.com) It says the role would collaborate with teams across “trading, networking, and research infrastructure,” suggesting the hardware effort is tied into execution systems and internal compute platforms rather than standing apart from them. ### What can actually be verified about Jane Street’s GPU and data-center buildout? (janestreet.com) Jane Street’s machine-learning page says the firm believes deep learning is “the future of quantitative trading” and says its ML team builds both models and the infrastructure for training and inference. The same page cites “tens of thousands of high-end GPUs” and says researchers, engineers and traders work together on models and production systems. (janestreet.com) Separate data-center job postings fill in some of the infrastructure detail. A Chicago data-center engineer listing asks for experience maintaining GPU hardware, designing and troubleshooting copper and fiber cabling, and working with liquid cooling systems. A New York mechanical engineer listing says the role would support “ML/compute cluster” infrastructure and cooling systems for “high-density GPU and trading systems.” (janestreet.com) ### What does Jane Street itself say these systems are for? Jane Street’s machine-learning page says its neural-network models drive trading strategies and that the team builds the infrastructure needed for training and inference. The page also says financial markets produce a torrent of noisy data and require “ultra-low latency systems” just to process it. The performance engineering page describes the same dual track from another angle. (janestreet.com) It says engineers work both on shaving nanoseconds from trading systems and on optimizing training loops for ML models, while also maintaining distributed systems with high throughput, low latency and reliability guarantees. Jane Street’s website continued to list FPGA, ASIC and data-center roles on May 17, 2026, and its engineering pages remained live with details on machine learning, performance engineering and hardware tooling. (janestreet.com 1) (janestreet.com 2) (janestreet.com 3)