For years Jane Street and Hudson River Trading have fought over speed, measured in nanoseconds. Today the race has moved to computing power: the two trading giants are competing for AI compute, and access to the most expensive technology of the moment has become the stake.
The largest firms are taking different routes. Some rely on cloud contracts, others back chip startups, and others build their own data centers. According to Bloomberg, the reason is straightforward: these operators increasingly rely on machine learning to predict market moves, and without computing capacity they cannot. One executive estimates that compute spending is now ten to twenty times higher than before ChatGPT launched.
Image: janestreet.com — https://www.janestreet.com/culture/street-view/a-look-inside-our-newest-data-center/
The figures reported by the sources show the scale. Hudson River spends about $1 billion a year on AI, according to AI chief Iain Dunning, and has signed a multiyear deal with CoreWeave. Jane Street took a more direct route: it invested $1 billion in CoreWeave and committed $6 billion to its computing services. According to one source, its Texas data center reportedly runs more than 4,000 Nvidia GPUs. The firm also led a funding round for chip startup Etched and became the first customer for its AI chip, which is designed to deliver more performance at a lower voltage. XTX Markets is building a data center in Finland.
The bet is risky, and the players themselves acknowledge it. The industry sees rising costs and fast-aging chips as real dangers. For high-frequency traders, however, investing is non-negotiable if it secures greater processing capacity and faster chips: these firms profit from tiny price differences, and speed remains the decisive edge. Jan Boomaars, CEO of Optiver, argues that doing nothing is still more dangerous.
The story fits into a wider race for the physical infrastructure of AI, which AlexTech has already followed: NVIDIA's move to turn compute into a financial asset class and Robinhood's plan to bring AI trading agents to 29 million clients show finance entering the AI game from several sides. The central issue, though, remains compute: in Wall Street's race, whoever lacks processing capacity risks falling behind.
No comments yet. Be the first!