The Quant Revolution

How artificial intelligence is transforming the most profitable corner of global finance and what it means for private equity.

In 2025, a relatively unknown trading firm generated $39.6 billion in revenue. It beat JPMorgan, the world’s largest bank with 320,000 employees and a century of institutional history, by 11%. It did this with 3,500 people. That works out to over $11 million in revenue per employee and roughly $9 million in profit per head, the highest ratio of any company with more than 1,000 employees anywhere on the planet. For context, Nvidia, the defining business of the AI era, generates $2.9 million in profit per employee. The firm is called Jane Street, and outside of quantitative finance, almost nobody has heard of it.

Jane Street does not manage client money, issue research, or advise on deals. It trades - systematically, algorithmically, at extraordinary scale across more than 200 exchanges worldwide. It is the purest expression of what quantitative trading, at its apex, looks like: a small group of mathematicians and engineers, an enormous amount of computing infrastructure, and a set of models that find and exploit patterns in markets faster and more precisely than any human could. The firm’s results are not an anomaly. They are the logical endpoint of a revolution that began over four decades ago, in a very different setting.

The Origin Story

The story begins in 1978, when a mathematician named Jim Simons walked away from his tenured chair at Stony Brook University and opened a trading firm in a Long Island strip mall. Simons was not a financier. He was a code-breaker, a geometer, a winner of the Oswald Veblen Prize - one of the highest honours in mathematics. He had no particular theory about markets. What he had was a conviction: that financial prices, like any complex system, contain patterns that mathematics can detect. He called the firm Monometrics. It would later be renamed Renaissance Technologies.

The early years were humbling. Simons experimented with lunar cycles and sunspot activity before abandoning qualitative hunches entirely and committing to a single principle - let the data speak, and let the computers act on what it says. He recruited not traders but scientists: cryptanalysts, physicists, astronomers, statisticians. Wall Street experience, he believed, was not an asset. It was a liability. The herd mentality of finance was precisely what his models were designed to exploit. By 1988, Renaissance had launched the Medallion Fund, and by the time Simons retired in 2009, it had generated average annual returns of 66% - a track record that has never been matched, by anyone, in the history of financial markets.

What Simons proved was not merely that mathematics could beat intuition. He proved that markets are not fully efficient - that exploitable patterns exist, invisible to the human eye, detectable only by machines operating at a scale and speed no analyst could replicate. That thesis, once demonstrated, could not be ignored.

Through the 1990s and 2000s, a new class of firm emerged in Renaissance’s wake: D.E. Shaw, founded by a Columbia computer scientist; Two Sigma, built explicitly as a technology company that happened to manage money; Citadel, which assembled an army of mathematicians and grew into one of the most powerful trading operations in the world. Each took Simons’ founding insight and pushed it further with more data, more computing power, faster execution, broader asset classes.

By the 2010s, quantitative and algorithmic strategies had ceased to be a niche and become the dominant force in global markets. Today, systematic funds control an estimated 35% of all US equity trading volume. The question is no longer whether mathematics beats intuition in financial markets. That debate is settled. The question now is what happens when artificial intelligence enters the equation.

What Quant Trading Actually Is And What It Looks Like in Practice

To understand what AI changes, it helps to first understand what quantitative trading actually is.

At its core, it is the replacement of human judgment with statistical inference. Rather than an analyst studying a company and forming a view, a quantitative fund builds mathematical models that identify relationships between prices, assets, market signals and future returns. The models then execute trades based on what those models predict. The human role shifts from making individual investment decisions to designing, testing and refining the models themselves. Emotion, narrative, conviction: none of it enters the process. Only data.

In practice this encompasses a wide range of strategies.

Statistical arbitrage exploits pricing discrepancies between related securities, betting that temporary divergences will revert to their historical mean.

Factor investing systematically tilts portfolios toward characteristics - value, momentum, quality, low volatility - that academic research has shown to predict returns over time.

High-frequency trading operates at microsecond timescales, capturing tiny margins at enormous volume. Trend-following identifies and rides sustained directional moves across asset classes.

What unites all of them is the same underlying logic: find a signal in the noise, size it correctly, and execute with absolute discipline before the market closes the gap.

What separates the best quantitative funds from the rest, however, is not the sophistication of their models alone. It is the quality and originality of the data they feed into them. This is where the field has undergone its most creative evolution and where some of its most striking stories emerge.

Consider crude oil.

The global oil market is notoriously opaque: official government inventory reports are published weekly at best, are frequently revised, and tell traders only what was true several days ago. A Palo Alto startup called Orbital Insight found a different way in. Crude oil is stored in vast cylindrical tanks equipped with floating roofs that rise and fall with the oil level. When the roof drops as oil is drawn out, the tank's outer wall casts a crescent-shaped shadow onto the sunken surface. By processing satellite imagery of the roughly 25,000 such tanks it monitors globally, and calculating the precise geometry of those shadows against the sun's angle, Orbital Insight can estimate oil inventory levels with remarkable accuracy - weeks before any official report is published. Traders who processed this data in early 2025 could position in oil futures up to 22 days ahead of official EIA figures. The information was always there, written in shadow geometry from space. It took a mathematician, not an oil analyst, to see it.

Perhaps the most vivid real-time illustration of the same principle is playing out in the Strait of Hormuz right now.

Since late February 2026, the strait, through which roughly 20% of the world's seaborne oil and one fifth of its liquefied natural gas normally flows, has been effectively closed following the outbreak of conflict in the region. Daily vessel transits collapsed from a February average of 135 ships per day to just seven by March 3rd. For most market participants, this was a crisis that announced itself through news headlines. For quantitative funds equipped with maritime intelligence platforms like Windward and Kpler - which aggregate AIS vessel tracking data, satellite imagery and real-time insurance pricing - the picture was visible far earlier. These systems were detecting the clustering of vessels off Fujairah, tracking ships going dark by disabling their AIS transponders, and monitoring the rerouting of Saudi crude through the Red Sea corridor, all before oil prices had spiked 20% and shipping futures had hit their daily trading limits. The data did not predict the conflict. But it told those watching closely that something had already changed — before the market had finished reading the news.

A third category of alternative data operates not in physical space but in the vast stream of human language.

Natural language processing algorithms now ingest millions of social media posts, news articles, earnings call transcripts and central bank statements, converting them into quantifiable sentiment signals in real time. The underlying insight is that markets are not moved by fundamentals alone - they are moved by fear and greed, by the collective emotional state of their participants.

By vectorising keywords across platforms like Reddit and X and mapping the resulting signals onto a fear-greed spectrum, algorithms can measure market psychology with a precision that no human analyst could sustain. The GameStop episode of 2021 was an early and dramatic demonstration of what happens when retail sentiment moves faster than institutional models can track. A Reddit-driven short squeeze that cost certain hedge funds over $70 billion in days. Sophisticated funds drew the lesson in reverse: if crowd sentiment can move markets that violently, then monitoring it systematically is not optional. It is alpha.

The fourth example is perhaps the most instructive and for investors focused on India, the most familiar.

By 2025, India had become the world's largest derivatives market, accounting for an extraordinary 61% of global equity options contracts by volume. The structural reason is telling: on a single Bank Nifty expiry day in January 2024, options notional turnover reached $1.26 trillion against just $3.6 billion in the underlying cash market - a ratio of 350 to one. That extreme imbalance between the derivatives market and the cash market on which it settles created an arbitrage opportunity that no human trader would have had the speed or scale to exploit. Jane Street did.

Over 18 expiry days between January 2023 and March 2025, the firm allegedly accumulated large positions in Bank Nifty constituent stocks in the morning - at times representing 15 to 25% of total market volume in select stocks - moving the index, then unwinding those positions later the same day to profit from the corresponding options positions. On January 17, 2024 alone, it made $84.5 million in a single session. Cumulative profits across the period reached an estimated $450 to $565 million.

The strategy only came to light accidentally. In April 2024, Jane Street sued a rival firm, Millennium Partners, in a New York court for allegedly stealing its India trading algorithms. The lawsuit exposed the strategy to India's market regulator, SEBI, which launched an investigation. On July 3, 2025, SEBI issued a trading ban and froze $565 million in Jane Street assets, concluding that the trading pattern had "no economic rationale other than the intent to manipulate." Jane Street maintains the trades constituted legitimate arbitrage. The legal outcome remains contested.

What is not contested is the impact: following the ban, trading volumes in Indian equity index options fell by a third - a measure of just how dominant a single algorithmic participant had become in one of the world's fastest-growing markets. It is a vivid illustration of the double-edged nature of quantitative dominance: the same precision that generates extraordinary returns can, when it outpaces the market structure it operates in, draw the attention of regulators with equal precision.

Quant Trading in the Age of AI

For most of its history, quantitative trading operated on rules that humans could, at least in principle, understand and audit. Buy when momentum exceeds a threshold. Sell when volatility breaches a limit. The models were sophisticated, but their logic was legible. What modern artificial intelligence introduces is something categorically different: the ability to identify patterns of extraordinary complexity that resist human interpretation entirely, operating across datasets so large and variables so numerous that no analyst could hold the relationships in their head. The change is not incremental. It is structural.

The scale of commitment from the industry’s leading firms reflects this.

Jane Street has deployed tens of thousands of high-end GPUs and more than one exabyte of storage to support its neural network models. In early 2026, it committed approximately $6 billion to CoreWeave’s AI cloud platform and made a separate $1 billion equity investment in the same company, securing access to next-generation compute infrastructure including Nvidia’s latest Vera Rubin architecture. The firm has stated publicly that deep learning is, in its view, the future of quantitative trading — not a tool within it, but the direction the entire field is moving.

Bridgewater Associates, the world’s largest hedge fund by assets under management, established its Artificial Investment Associate laboratory in 2023, structuring it so that AI serves as the primary decision-maker within the fund, with human professionals overseeing risk management, data acquisition and execution.

Citadel, which manages $50 billion and generated a record $12.2 billion in trading revenue in 2025, has built what amounts to a world-class technology company alongside its investment operation. These are not experiments. They are strategic bets on the permanence of AI as the dominant competitive variable in markets.

What does AI actually enable that previous generations of quantitative models could not?

Several things, each meaningful in its own right. Deep learning models can detect non-linear relationships in price and volume data that classical statistical models will systematically miss. Large language models can parse the precise wording of a Federal Reserve statement, an earnings call transcript or a geopolitical communiqué in milliseconds, extracting sentiment and forward signals that a team of analysts would take hours to process.

Reinforcement learning allows trading agents to be trained through simulated market environments, developing strategies that adapt dynamically to changing conditions rather than breaking when reality diverges from historical patterns. And increasingly, multi-agent systems, networks of specialised AI models collaborating across the investment pipeline, are beginning to automate not just execution but the research process itself.

The performance gap this creates is measurable. In 2024, funds with advanced AI capabilities outperformed traditional quantitative funds by four to seven percentage points on average. AI-driven strategies accounted for over 40% of hedge fund trading volumes that year. Systematic funds gained nearly 12% in the first half of 2025, roughly double the returns of traditional stock-pickers over the same period. Bridgewater’s Pure Alpha fund gained 17%.

These are not outliers.

They represent a consistent pattern: in volatile, information-rich markets, systematic AI-augmented strategies are outperforming discretionary ones with increasing regularity.

None of this means the edge is guaranteed or permanent. The same forces that are concentrating returns among AI-capable funds today are quietly laying the groundwork for the next phase of the industry’s evolution - one that raises more fundamental questions about where competitive advantage in quantitative trading will ultimately reside.

The Future: Data and Research Quality as the Last Moat

There is a seductive assumption embedded in the current AI arms race in quantitative finance: that the firm with the most sophisticated model wins. It is an assumption worth interrogating. Because if the history of quantitative trading teaches anything, it is that algorithmic edges are temporary. Data edges are not.

This phenomenon has a name in the industry: alpha decay.

When a strategy works, capital flows toward it. As more funds adopt the same approach, the pricing inefficiency it exploits gets arbitraged away, and the returns compress. The momentum factor, one of the most robust and widely studied signals in quantitative finance, generated returns of approximately 10% annually through the 1990s. Today that figure is closer to 2%.

The decay is not random. Research shows it follows a hyperbolic curve, accelerating sharply once a strategy becomes widely known, and correlating directly with the growth of capital allocated to it. Crucially, this process has been getting faster. Academic analysis shows that crowding accelerated significantly post-2015, coinciding with the democratisation of factor investing through ETFs - and there is every reason to expect AI democratisation to accelerate it further still.

As the largest model providers make increasingly powerful AI available at decreasing cost, the models themselves will cease to be a meaningful differentiator. A fund running a transformer architecture in 2030 will be in roughly the position of a fund running a standard momentum strategy in 2010: using a tool that is necessary to compete, but not sufficient to win. The algorithm will be the price of entry, not the source of edge.

What cannot be so easily commoditised is data. Not the publicly available data that everyone can purchase from the same vendors, but the proprietary, difficult-to-replicate data pipelines that the best funds have spent years and hundreds of millions of dollars constructing.

Renaissance Technologies’ real moat was never its model architecture. It was forty years of tick-by-tick price data going back to the 1960s, combined with weather patterns, satellite imagery and shipping records that no competitor had thought to collect, integrated into a single unified model that compounded its advantage with every passing year. That forty-year head start does not compress when a better algorithm becomes available. It widens.

The Hormuz example makes this concrete. When the strait began closing in late February 2026, the AIS vessel tracking data that Windward and Kpler aggregate was, in principle, accessible to any fund willing to pay for it. But data access alone does not generate alpha.

What generated alpha was the research framework that told a fund’s analysts which signals to monitor, what behavioural patterns - vessels going dark, clustering off Fujairah, rerouting through Iranian territorial waters - were meaningful precursors to a supply shock, and how to size a position in oil futures before the market had finished reading the headlines. That framework is not a model. It is accumulated intellectual capital: the product of years of research into how maritime disruption transmits into commodity pricing. It cannot be downloaded.

This points toward what we believe will define the next era of quantitative investing: not the quality of the algorithm, but the quality of the question. The funds that will generate sustained alpha in a world of commoditised AI will be those with the sharpest instinct for which signals matter, the most original thinking about where unexploited patterns might exist, and the data infrastructure to pursue those hunches at scale. The best quant funds of the next decade will look less like technology companies and more like the original Renaissance: a group of brilliant people asking questions that nobody else has thought to ask, about data that nobody else has thought to collect.

There is one important counterargument to this thesis. The same multi-agent AI systems that are beginning to automate trade execution are also beginning to automate research itself - generating and testing hypotheses with minimal human input. If AI can eventually replicate not just the analytical process but the creative instinct that identifies which questions are worth asking, then even the research moat may face pressure in time. That inflection point is not imminent. But it is worth watching.

The Private Equity Opportunity

For private equity, the rise of quantitative trading presents opportunities that operate on three distinct levels - each requiring a different kind of engagement, and each carrying a different risk profile.

The most PE-native entry point is the General Partner (GP) stakes model.

As quantitative and multi-strategy funds scale, they face a structural capital problem that private equity is well positioned to solve. Building a competitive quant operation - proprietary data pipelines, computing infrastructure, talent capable of commanding seven-figure compensation - requires significant balance sheet investment that performance fees alone cannot reliably fund, particularly through lean periods or aggressive expansion phases.

The established response, pioneered by Goldman Sachs’ Petershill programme, Blue Owl’s Dyal platform and Blackstone’s Strategic Capital group, is the minority stake acquisition: purchasing 10 to 20% of a fund’s management company, typically targeting managers with $2 to 15 billion in AUM and a four to five year track record, in exchange for permanent capital that the GP can deploy into infrastructure and talent. Return targets in this space have historically centred around 20% IRR, underpinned by a share of management fee income that provides stability regardless of fund performance.

For PE firms with the relationships and diligence capability to identify tomorrow’s quant winners before they become obvious the entry economics are significantly more attractive than waiting for the category to mature further.

The second opportunity is in the infrastructure layer that quantitative trading depends on but cannot build alone.

The alternative data market - the ecosystem of satellite imagery providers, maritime intelligence platforms, NLP data vendors and AI compute suppliers that feeds the strategies we have described - is projected to reach $135 billion by 2030. PE-backed data centre M&A alone hit $18 billion globally in 2024.

These are picks-and-shovels plays: businesses whose revenues grow with the expansion of quantitative trading broadly, without the performance risk of backing any individual fund. Kpler, Vortexa, Orbital Insight and their peers are the vendors that every sophisticated quant fund depends on. Owning the infrastructure layer is a more durable position than owning any single strategy built on top of it.

But perhaps the most consequential opportunity is also the least obvious - and it has nothing to do with investing in quantitative trading at all. It is about bringing quantitative thinking into private equity itself.

The core lesson of the last four decades in public markets is that systematic, data-driven decision-making consistently outperforms intuition-based judgment over time - not because human judgment is worthless, but because it is inconsistent, subject to bias, and unable to process information at the scale that modern markets require.

That lesson has not yet been fully absorbed by the private markets industry. Deal sourcing remains largely relationship-driven. Sector selection is still dominated by narrative and conviction. Portfolio monitoring relies heavily on periodic reporting rather than continuous signal detection.

Some of the industry's most forward-thinking firms are beginning to change this. EQT, the Swedish PE giant, has operated an internal platform called Motherbrain since 2018. It is a proprietary system that ingests and analyses public and web data sources using AI and machine learning to identify potential investment targets before they surface through conventional channels. The explicit goal is to find companies earlier than a banker's shortlist would allow, and to do so systematically rather than through the happenstance of relationships. It is, in essence, the application of quantitative signal-detection to private markets deal flow - the same underlying logic that Renaissance applied to securities prices, directed instead at privately held companies.

The firms that will have a structural advantage in private equity over the next decade are those that follow this lead and go further. This means building proprietary datasets on sector dynamics, talent flows, and competitive positioning that inform investment theses before a process begins. It means monitoring portfolio companies not through quarterly board packs but through the same kind of real-time data signals that quant funds use to front-run earnings announcements. It means treating the question of which sectors to enter not as a matter of conviction but as a data problem.

This is not a distant possibility. The tools are available today, and the cost of accessing them has fallen dramatically. The question is not whether data-driven private equity will become the norm. It is which firms will build that capability first - and capture the advantage that always accrues, in any market, to those who saw the signal before everyone else.

The firms that will have a structural advantage in private equity over the next decade are those that begin applying the quantitative discipline - the rigorous data sourcing, the systematic signal generation, the model-driven hypothesis testing - that has already transformed public markets. This means using alternative data to identify high-growth companies before they appear on a banker’s shortlist. It means building proprietary datasets on sector dynamics, talent flows, and competitive positioning that inform investment theses before a process begins. It means monitoring portfolio companies not through quarterly board packs but through the same kind of real-time data signals - foot traffic, hiring velocity, supply chain activity - that quant funds use to front-run earnings announcements.

This is not a distant possibility. The tools are available today, and the cost of accessing them has fallen dramatically. The question is not whether data-driven private equity will become the norm. It is which firms will build that capability first - and capture the advantage that always accrues, in any market, to those who saw the signal before everyone else.

The quant revolution's lesson is simple: in any market, the edge belongs to those who trusted the data first

Disclaimer: This Market Note is produced by NEEM Group for informational purposes only and does not constitute investment advice or a solicitation to buy or sell any security. The views expressed reflect NEEM’s analysis at the time of writing and are subject to change. Readers should conduct their own due diligence before making any investment decisions.