High-frequency trading firms like Jane Street and Hudson River Trading are no longer just racing for nanoseconds — they’re now competing for compute power in the AI era. These market-making giants are investing billions in data centers, chips, and infrastructure to fuel machine learning strategies that predict market behavior. One executive estimates spending on compute has surged 10 to 20 times since the launch of ChatGPT in November 2022. Firms are building their own data centers, partnering with tech giants, and acquiring startups to secure faster processing capacity. Hudson River operates a data center in a former olivine mine in western Norway, powered by Dell and Nvidia hardware, and cooled by nearby fjord water. While the costs are climbing and advantages may be short-lived, most high-speed traders view compute investment as essential to avoid obsolescence. As Jan Boomaars, CEO of Optiver, notes, staying competitive in systematic trading requires being in the AI game.
The shift reflects broader industry trends. AI-related IPOs have faced setbacks, and infrastructure costs are rising. Yet for HFTs, access to cutting-edge chips and data centers remains a nonnegotiable priority. Their edge comes from exploiting tiny price differences at high speed, increasingly guided by machine learning models that demand vast computational resources.
Despite risks, firms continue expanding. By one estimate, expenditure on computing power has grown as much as 20-fold in four years. The race for compute is now as critical as the race for speed.
Source: Bloomberg Markets · Summarized by HeadlinesBriefing