DeepMind Alumni Turn Poker AI Into a $500 Million Trading Bet

Three former DeepMind researchers who once taught an AI system to outplay human poker professionals are now trying to do something even more commercially consequential: use the same kind of self-learning technology to generate returns in global markets. Their Prague-based startup, EquiLibre Technologies, has just raised a Series A at a reported $500 million valuation, according to people familiar with the deal, as investor appetite grows for reinforcement learning applied to finance.

The company’s pitch is straightforward but ambitious. If an AI can learn to make optimal decisions in a game of imperfect information such as poker, its founders argue, it can also learn to manage trading decisions where outcomes depend on probabilities, risk, timing and incomplete signals. Backed by a new round led by Creandum, EquiLibre is already working with one of the industry’s best-known quantitative trading firms and says its systems are moving real money at scale.

That combination of academic pedigree, frontier AI techniques and live market deployment has made EquiLibre one of the most closely watched emerging AI companies in Central and Eastern Europe. It also places the startup inside a segment of AI where technical breakthroughs can translate directly into revenue — and where the competition includes some of the most heavily resourced firms in modern finance.

From poker tables to trading desks

EquiLibre’s story starts in the world of research, not Wall Street. The company was founded by Martin Schmid, Rudolf Kadlec and Matej Moravcik, who all spent time as visiting PhD researchers at DeepMind’s first international AI research office in Edmonton, Alberta. That office became a landmark in reinforcement learning research, and the trio’s work there helped produce DeepStack, an AI system widely recognized as the first program to defeat professional players in no-limit Texas hold’em poker.

That achievement mattered far beyond the game itself. Poker is a textbook example of partial information, strategic deception and sequential decision-making, making it an ideal test bed for reinforcement learning. Systems trained in this way do not simply memorize answers; they learn by repeatedly trying actions, receiving feedback and improving through rewards. For markets, that logic is particularly compelling because the “score” is easy to define: profit and loss.

Schmid has said that trading appeals to reinforcement learning because the feedback loop is unusually clean: the model’s success can be measured in money earned or lost.

The startup’s founders have carried that idea from theory into deployment. In partnership with Tower Research Capital, a major quantitative trading firm, EquiLibre says its algorithms have been trading billions of dollars in daily volume across large U.S. equity benchmarks, including the S&P 500 and Nasdaq-linked markets. The company also says its systems expanded from crypto trading in 2025 to stock exchanges more recently.

EquiLibre claims its track record has been unusually consistent, saying its agents have not posted a negative month since launch. In trading terms, that means each month has ended in positive territory overall, though the company has not disclosed the exact size of profits or the risk profile behind those results.

Why investors are paying attention

The new financing underscores how quickly investor sentiment can shift when an AI technique moves from academic promise to commercial utility. Reinforcement learning has been around for years, but its practical relevance has expanded as better compute, better data pipelines and more capable models have made it viable in real-world environments.

For venture investors, finance offers one of the most attractive proving grounds because the market is enormous and performance improvements can be monetized immediately. That helps explain why Creandum stepped in as lead investor. The firm’s vice president, Cameron Sellers, said the round was the largest single check Creandum has ever written into one company, signaling both conviction and size of ambition.

Sellers argued that global trading represents one of the biggest addressable markets anywhere in the economy and noted that successful quant funds can generate returns that dwarf many traditional venture outcomes.

At the same time, Creandum is careful not to frame the company as a conventional financial-services startup. The firm has described EquiLibre as a research lab first, rather than a finance company. That distinction matters: the company is pitching itself as an AI builder that happens to be applying its technology to markets, not as a hedge fund that happens to use AI.

That positioning may help it attract talent, capital and optionality. It also places the company inside a broader wave of AI labs founded by DeepMind alumni, a category that has become especially hot among venture firms looking for founders with a record of building frontier systems.

A Prague base with global ambitions

Unlike many AI startups in Europe that migrate to London, San Francisco or New York, EquiLibre doubled down on Czechia. The founders chose Prague as the company’s base after returning to their home country and drawing on a network they had built through years in research and engineering circles.

That decision turned out to be strategic as well as personal. The founders were able to recruit early teammates from a mix of local contacts and expatriates who had worked at Google, DeepMind and other major tech organizations. Today, EquiLibre says it has grown to 25 employees — still small by financial-industry standards, but meaningful for a deeply technical AI company.

Schmid has suggested that Prague provides an advantage over the Bay Area because the local AI ecosystem is less prone to talent churn. In San Francisco, he argued, attention can shift quickly from one hot startup to the next, making retention harder. In Prague, the company believes it can build a more stable team while still tapping into international expertise.

The city already hosts a growing cluster of AI activity, including other startups in the same building. EquiLibre is not alone in trying to turn Central Europe into a serious AI hub, but its traction and valuation place it near the front of the region’s emerging frontier-tech story.

What EquiLibre is building with the new capital

The immediate priority for the startup is compute. EquiLibre says it intends to expand its infrastructure substantially and bring online what it expects will become one of the largest compute clusters in Central and Eastern Europe.

That goal reflects a broader reality in modern AI: even when the core algorithm is elegant, training and running it at scale requires substantial hardware. In finance, this matters even more because the system must process noisy data quickly, adapt to shifting patterns and remain robust under live conditions. Better infrastructure can become a competitive advantage, especially if the company is trying to squeeze more performance from fewer chips than rivals with larger balance sheets.

The compute race also highlights one of the biggest threats to emerging AI challengers. Larger trading firms and better-capitalized competitors can match or exceed many technical capabilities simply by spending more on infrastructure. EquiLibre says it is aiming to compensate through efficiency rather than brute force.

Schmid has said the company’s strategy is to get more out of less hardware, rather than relying on the kind of massive GPU fleets used by the biggest players.

That approach could prove essential in a sector where the cost of being wrong is immediate and measurable. If the firm’s agents fail, the losses are real. If they succeed, the upside is equally tangible.

How the funding stack appears to have evolved

Although EquiLibre and its backers have not disclosed the full size of the Series A or the startup’s cumulative funding, the company has previously completed at least two earlier rounds. Public deal data suggest the startup’s seed financing reached $10 million and valued the company at $140 million, with Blossom Capital leading that round. Early backers reportedly included Credo, a Central and Eastern Europe-focused venture firm that has also backed standout regional companies such as ElevenLabs and UiPath.

The jump from a $140 million valuation to $500 million in the latest round indicates how much investor enthusiasm has increased. It also suggests that EquiLibre has managed to convince backers not only that its technology works, but that it can keep scaling in a fiercely competitive environment.

Below is a simplified view of the company’s development and funding milestones based on the information available so far.

The reinforcement learning thesis behind the company

EquiLibre’s core argument is that reinforcement learning is especially well suited to trading because financial markets reward adaptive behavior. Unlike many machine-learning tasks that require static classification, markets constantly change, forcing a model to revise assumptions and optimize against uncertainty.

Why the technique matters

In simple terms, reinforcement learning lets an AI learn by trial and error. It makes decisions, measures outcomes and updates its behavior based on what produces the best long-term reward. In poker, the reward might be winning chips. In finance, it is return adjusted for risk.

That makes the problem both elegant and difficult. Markets are adversarial, crowded and shaped by other fast-moving participants. A model can do everything “right” and still lose because prices shift for reasons it cannot fully observe. That’s why a profitable system must not only react; it must anticipate.

Schmid has said that when EquiLibre started four years ago, reinforcement learning was still met with skepticism in trading circles. Since then, he argues, it has become much more accepted, which may give the firm a head start.

That said, the fact that a method is now popular does not guarantee competitive advantage. In finance, widespread adoption often means narrower margins, as more firms chase similar signals and strategies.

What makes markets a good fit

• Feedback is measurable and immediate through profit and loss.
• Trading can be automated at scale once systems are robust enough.
• Large markets create room for meaningful upside if models work.
• Data-rich environments make continuous learning possible.

Still, the same characteristics that attract builders also attract rivals. As soon as a promising technique becomes well known, quant firms, hedge funds and technology companies begin racing to replicate or surpass it.

Competition is already intense

EquiLibre is entering a field where some of the biggest names in quantitative finance already use advanced AI methods. Jane Street, one of the most profitable trading firms in the world, publicly says it uses reinforcement learning together with large language models and whatever other tools are needed to train effective models. It also says it operates with tens of thousands of high-end GPUs.

That scale matters. A well-funded competitor can outspend a startup on infrastructure, research, talent and experimentation. EquiLibre’s answer is not to match that capital base one-for-one, but to innovate in how efficiently it uses compute and how tightly it integrates research with live trading.

There is also a broader strategic question. If AI becomes a standard ingredient in trading, then the advantage may no longer come from having the technique itself but from having the best implementation, the most disciplined risk management and the strongest operational execution.

In that sense, EquiLibre’s challenge is similar to that of many AI startups: prove that a brilliant idea can become a durable business before better-funded rivals close the gap.

Why the founders still see a long runway

Despite the competitive risks, the founders appear convinced they are early rather than late. Their view is shaped by years of work in reinforcement learning before the field became fashionable in finance.

Schmid has said the company is motivated less by market efficiency than by the excitement of building something new. That framing suggests a research-first culture in which success is measured not only in returns but also in the novelty and technical depth of the work.

The founders see the company as a builder’s project rather than a pure trading play, with the goal of creating an AI lab that happens to operate in markets.

That mindset may also help with recruitment. Top AI researchers often want to work on technically ambitious problems. A startup that can credibly claim to be advancing reinforcement learning in live financial systems may have an edge in attracting talent over a firm that is simply optimizing a trading desk.

EquiLibre’s location in Prague could reinforce that advantage. The city offers access to highly skilled engineers, lower operational costs than major U.S. tech hubs and a sense of mission around building a global company from Central Europe.

What the round signals for AI and finance

The deal says something larger than one startup’s valuation. It reflects how venture capital is increasingly rewarding AI companies that can point to a direct line from model training to economic output. In sectors like finance, logistics and manufacturing, AI is no longer just about productivity gains; it is about making decisions that influence immediate revenue.

Reinforcement learning, in particular, may be regaining momentum because it addresses environments where static prediction is not enough. That is especially true in trading, where systems must optimize over sequences of actions instead of one-off answers.

The funding also illustrates a geographic shift. While many frontier AI companies remain concentrated in the U.K. and the U.S., EquiLibre shows that serious work can emerge from Prague, especially when founders bring global research credentials and a network of former colleagues willing to return or join remotely.

For Europe’s startup ecosystem, the message is encouraging. For global quants, it is a reminder that the next serious competitor might not come from a traditional hedge fund at all, but from an AI lab founded by researchers who once solved a famous game-theory problem.

Why this is not simply a Wall Street story

Although the company’s commercial activity centers on trading, its self-description matters. EquiLibre wants to be known for building AI systems, not for being a hedge fund in disguise. That distinction could shape everything from hiring and regulation to partner perception and long-term strategic options.

By keeping a research identity at the center, the startup preserves flexibility. It can continue to work across asset classes, move between markets and adjust its product model as the technology evolves. It also helps the firm present itself as part of the broader AI wave rather than as a niche financial service provider.

That is important because venture investors tend to value software and AI businesses differently from traditional finance operations. The “lab first” framing makes the company more legible within Silicon Valley-style venture capital, even as its actual customers and counterparties live in the financial world.

Key numbers at a glance

The bigger picture for frontier AI startups

EquiLibre’s rise fits a broader pattern in frontier AI: researchers with deep technical credentials are founding companies in areas where algorithms can be tested against measurable outcomes. Poker, games, robotics and now trading all share one crucial feature — a clear reward signal. That makes them attractive proving grounds for the next generation of AI systems.

As investors search for businesses that can justify rising infrastructure costs, startups with a direct link between algorithmic improvement and monetizable performance may become even more valuable. In that environment, a company like EquiLibre can look unusually compelling: scientifically credible, commercially active and still early in its lifecycle.

Whether the firm becomes a durable player in trading or simply a valuable technical experiment will depend on execution, compute access, risk control and the ability to stay ahead of larger competitors. But for now, the company has achieved something many AI startups struggle to do: it has moved from research reputation to market relevance.

And in a sector where credibility is everything, that may be the most important trade of all.

superintelligencenews@gmail.com