Jane Street Group is preparing for a version of quantitative trading where the scarce asset is no longer data alone. It is electricity, cooling, fiber, and access to the latest GPUs.
The firm already operates tens of thousands of high-end GPUs. It has committed billions of dollars to external cloud capacity and is now considering a self-financed 200 megawatts (MW) data center large enough to move Jane Street into hyperscale territory. The strategy is not to replace its ultra-fast trading systems with AI. It is to own more of the infrastructure behind both.
Jane Street Wants a 100–200 MW Facility of Its Own
Bloomberg reported in June that Jane Street had started early discussions with companies in technology, finance, and crypto about developing a 100–200 MW data center. A final location has not been selected, and the firm has not publicly confirmed the project.
The proposed facility would mainly support Jane Street’s own workloads, including models used to forecast asset prices. The firm reportedly wants to increase its total computing capacity tenfold, moving from tens of thousands of GPUs towards hundreds of thousands. That is infrastructure on the scale normally associated with major AI laboratories and cloud platforms like Microsoft or Google.
One Texas Data Center Already Holds 4,032 GPUs
The scale is easier to understand from what Jane Street has already built. In May, the firm opened one of its Texas data centers to podcaster Dwarkesh Patel. Ron Minsky, who co-heads the firm’s technology group, and Dan Pontecorvo, who runs physical engineering, walked Patel through it rack by rack.
The facility contains 4,032 GPUs spread across 56 racks. It uses liquid cooling and roughly 8,000 kilometres of fibre. The most striking detail was not the number of chips. It was how little floor space they occupied compared with how much power they consumed. The building still had room for more racks, but its grid connection was already becoming the limiting factor.
Jane Street Has Also Committed $7 Billion to CoreWeave
Jane Street is not relying on owned data centers alone. CoreWeave announced in April that the trading firm had signed a roughly $6 billion cloud agreement and invested another $1 billion in CoreWeave shares at $109 each.
The agreement gives Jane Street access to AI computing capacity across several locations, including systems based on Nvidia’s Vera Rubin platform. The combination matters. CoreWeave gives the firm access to new chips without waiting for its own buildings to be completed. Owned facilities give it more control over networking, power and long-term operating costs. Jane Street is building both because neither is sufficient on its own.
Nanoseconds and AI Models Do Different Jobs
A caveat worth stating plainly is that Jane Street is not handing every trading decision to a large AI model.
During the Texas visit with Patel, Minsky explained that the firm works across very different timeframes. Some systems must process market data and respond in under 100 nanoseconds. Larger machine-learning models are used for slower work, including research, pricing and identifying patterns across large amounts of data. AI helps decide what an asset may be worth. The low-latency systems still handle the race to execute.
Power Is Becoming Part of the Trading Edge
The problem is no longer just finding enough GPUs. It is finding somewhere to run them.
“…you just can’t get the amount of power like you cannot wire in enough thunderbolts into the same data center to power all the things you need. You need to get the data centers built all over the place,” Minsky said.
His answer is distribution. That is why Jane Street expects to spread its computing infrastructure across several locations instead of forcing everything into one building.
That changes the competitive map. Quant firms have always competed for better data, faster connections and stronger researchers. Jane Street’s buildout adds electricity and data center capacity to that list. The firm is still chasing speed. It just needs far more power to do it. The next advantage in trading may still be measured in nanoseconds, but it will increasingly be built in megawatts.
Why This Race Is Bigger Than Jane Street
Jane Street isn’t alone in treating AI infrastructure as trading strategy rather than back-office IT. Hudson River Trading runs its own data center too, training models on more than 100 terabytes of market data that one of its researchers has compared to training a frontier language model. HRT posted a record $12.3 billion in trading revenue last year. Jane Street posted $39.6 billion, beating JPMorgan‘s entire trading division with a fraction of the headcount.
What’s changed is the assumption that speed alone wins. Jane Street is betting the next edge belongs to firms that can run nanosecond execution and slow, deep research at once, without either one slowing the other down. Whether 200 megawatts proves that, or is just the opening bid, is the part still playing out.
Author: Richardson Chinonyerem
The editorial team at #DisruptionBanking has taken all precautions to ensure that no persons or organisations have been adversely affected or offered any sort of financial advice in this article. This article is most definitely not financial advice.
See Also:
Jane Street Hit with Terra $40B Insider Trading Suit | Disruption Banking
