Street’s quant firms have long been among the most lucrative destinations for the world’s top scientific minds. The AI boom has given those mathematicians and physicists another set of deep-pocketed suitors — and sent some of finance’s most secretive employers on an increasingly public recruitment drive.

Take the jet-setting schedule of Charles Martinez, who leads academic outreach for G-Research, a London-based quantitative trading firm. A physicist with a doctorate himself, he attends AI conferences around the world, visits universities from Boston to the Arctic Circle and hosts pub quizzes where math replaces trivia. The pitch, in essence: help build models to predict the next move in financial markets rather than the next token in an AI system — even if it means toiling in the shadows.

“Your friend in OpenAI will be, ‘Okay, I did this, I did that,’ and you’ll be like, ‘Yeah, I did something, but I can’t tell you,’” said Martinez. Money is hardly scarce in quant finance. First-year compensation packages for machine-learning and related PhDs typically range from $450,000 to $650,000, but can go as high as $800,000, according to one recruiting firm. But AI labs and startups offer those same researchers a different proposition: prestige, visibility and, in some cases, the prospect of a windfall from owning a stake in a fast-growing technology company.

In response to the competition, some quant firms are pushing deeper into academia themselves. Qube Research & Technologies has launched QRT Labs, a program sponsoring more than 70 PhD and postdoctoral researchers. Jump Trading funds academic fellowships. Abu Dhabi Investment Authority has started a quant research residency for PhDs. Capital Fund Management in Paris runs a machine learning lab designed in part to resemble a university research group.

The payoff isn’t necessarily an immediate trading signal. University sponsorships can raise a firm’s profile among potential recruits, while internal labs can help attract scientists and give them room to pursue longer-term research that may eventually improve trading models.

G-Research is making a similar bet. This summer, it announced the largest private-sector investment in the European Laboratory for Learning and Intelligent Systems, a network of machine-learning researchers, with funding for 36 of them. The firm, which employs about 1,000 people, also funds 43 PhD students across eight British universities.

There’s no expectation that the sponsored researchers will eventually join G-Research or do work applicable to finance, Martinez stresses. The payoff is reputational: raising the firm’s profile among potential recruits and giving its brand what he calls a “warm fuzzy feeling.”Even with the intense competition, he reckons G-Research can go toe-to-toe with the AI labs.

The way its quants “run their groups is like the way a university would run a PhD group,” he said from the firm’s London headquarters. “The second thing is that we benefit from one of the most powerful computing resources in the world. So whatever you want to test, you can do it.”XTX Markets Ltd., JPMorgan Chase & Co., Hudson River Trading LLC, Optiver Holding BV and CFM are among other firms with dedicated AI labs, though their mandates vary.

CFM, the roughly $35 billion Paris-based quant hedge fund co-founded by a physics professor, has built a lab designed to preserve some of the freedoms of academic research. To lead it, the firm hired a New York University math professor who now also teaches in Paris. He works with two full-time researchers and a rotating group of postdocs.

“They are not under the pressure to build alpha models,” said Chief Product Officer Laurent Laloux. “They have complete academic freedom in the sense that they are fully immersed in the academic system, participating and contributing to research, publications and conferences and also bringing good research internally.”Quants have long been the most scholarly of Wall Street’s lot. Many of their models have academic origins, and the work often draws on statistical and engineering skills honed in science PhD programs. That relationship has only deepened as systematic firms deploy more machine learning, a field whose breakthroughs have often emerged from universities.

At CFM, the lab gives its researchers the time and space to “do this longer-term thinking to ensure that we are able to grasp the cutting edge of academia and understand how it can be applied to systematic investing,” said Laloux.

Leading machine-learning conferences such as NeurIPS and ICML are now dotted with booths from quant firms handing out swag, hosting parties and explaining finance 101 to students. Of the top-tier sponsors at the NeurIPS conference this December, nearly half are financial firms.

The likes of HRT, XTX and Jane Street also operate their own data centers packed with GPUs — another draw for the top whizzes.

“It isn’t just about compensation,” said Tom Chambers, partner specialized in AI at Maven Partnership, a recruitment firm. “It’s about commitment to compute and infrastructure.”The competition is evident in the revolving door of talent. HRT, known for its longstanding AI lab, has lost several employees to OpenAI.

Two Sigma quants have gone to Anthropic. More occasionally, the traffic moves the other way.

A few Google DeepMind scientists have joined Jump and Citadel Securities, for instance. At Jump, the long-established high-frequency trading firm, Chief Technology Officer Alex Davies says it has always had academic partnerships but decided to publicize them more as it tries to appeal to younger researchers, who are more likely to investigate potential employers themselves rather than go through recruiters.

“The competition for talent is as challenging as I’ve ever known it,” said the 15-year Jump veteran. “AI labs are clearly the biggest competitor right now.”In one respect, Jump is also starting to sound more like those labs. Some employees take the title “member of technical staff,” a deliberately vague label popularized by OpenAI and Anthropic. Davies says he likes that the term erases the traditional distinction between researchers and engineers — a line that has become blurrier with AI-assisted coding.

Both quant firms and AI labs train machine learning models to make predictions. But Davies thinks financial markets hold a particular appeal for researchers drawn to difficult problems.

“If your problem is to perfectly translate language, people don’t make the language harder to confuse the bot,” said Davies. “Even if you perfectly solve prediction in finance, by rolling that out, markets will change underneath you, and you haven’t solved it any more.”