@0xCheshire: Jane Street just released inside views of its Texas AI training center: 4,032 GPUs, 8,000 kilometers of fiber optic cable, and a fully deployed liquid cooling system because air cooling couldn't keep up. But what's truly stunning is the origin of this computing behemoth. Technical lead Ron Minsky recalls...
Summary
Jane Street revealed inside views of its AI training center in Texas, housing 4,032 GPUs, 8,000 kilometers of fiber optics, and a full liquid cooling system, while recounting the 20-year evolution from a humble start with six Dell hosts to today's extreme trading system.
View Cached Full Text
Cached at: 05/16/26, 01:18 PM
Jane Street just released internal footage of its Texas AI training center: 4,032 GPUs, 8,000 kilometers of fiber optic cable, and a full liquid cooling system because air cooling couldn’t keep up.
But what’s truly stunning is the origin of this computing behemoth.
Technical lead Ron Minsky recalls that their first compute cluster was just six Dell machines stacked at the end of a desk. Back then, to be able to pull the plug quickly when something failed, the equipment was left exposed right next to the traders.
The most absurd incident: a cleaner, while vacuuming the carpet during the day, casually unplugged the power cord from a trading system that was running live.
From six Dell computers and a vacuum cleaner accident to a liquid-cooled data center that can execute trades in 100 nanoseconds.
This is the ultimate evolution of a top-tier Wall Street quantitative firm, spanning 20 years.
Cheshire🔔|Crypto+AI Plus (@0xCheshire): Last year, a team of 3,500 math geeks quietly pocketed $39 billion.
They don’t care about stock fluctuations at all! Because Jane Street is essentially a precisely operating math factory.
While the masses are frantically betting on market directions, they are relying on complex algorithms to mercilessly harvest market inefficiencies in milliseconds.
Similar Articles
@zostaff: 20 years ago Jane Street's entire compute cluster was six Dell boxes stacked on the floor at the end of an office row. …
Jane Street allowed Dwarkesh Patel to tour their new Texas data center with 4,032 GPUs, each rack pulling 140 kilowatts, highlighting the massive scale and unique networking choices.
@yminsky: We gave @dwarkesh_sp a tour of one of our new GPU-filled data-centers. Much fun!
Jane Street's tech lead Ron Minsky and physical engineering team lead Daniela Corvo gave a tour of a Texas data center retrofitted for high-density liquid-cooled GPU training, detailing liquid cooling systems, power distribution challenges, leak detection, and the opportunity cost of compute resources.
@FinanceYF5: OpenAI just broke ground on a 1GW AI data center in Michigan. The project, called "The Barn," is the previously blank Midwest site of the "Stargate" plan, with a total investment of approximately $16 billion, making it the largest single project in Michigan history.
OpenAI has broken ground on a 1GW AI data center named 'The Barn' in Michigan, with a total investment of about $16 billion. It is the Midwest site of the Stargate plan and the largest single project in the state's history.
@Damir_Akaza: Jane Street, one of the richest and most secretive firms in the world, paid him between $330,000 and $600,000 a year, a…
A tweet highlights how a math genius built an AI system capable of trillions of operations per second for Jane Street, earning a huge salary, and promotes a lecture explaining the methodology.
@xiaomovps: After a company starts using AI, they quickly hit several hard problems: whether data can be externalized, whether costs can be controlled, and whether to build internal models themselves. This article documents a very real weekend operation—remotely connecting to the company's DGX Spark and hands-on running Ling-3.0-flash. Not stopping at 'can it run', but…
This article documents the complete process of the author remotely connecting to the company's DGX Spark server on the weekend to successfully deploy the Ling-3.0-flash model, including selection, deployment, performance testing, and integration with development tools, and shares insights on local deployment as a controllable intermediate state.