Why We Invested in Variational AI:
AI-First Drug Discovery Is Here

Thursday, 19 June 2025 I Written by Jason Robertson & Adrisa Agarwal

Drug discovery is hard. Really hard.

It takes over a decade and $2.5 billion to develop a single successful drug[1]. And that’s assuming it even makes it—90% of drug candidates fail before they ever reach the market. Why? Because drug discovery today is largely a game of trial and error. Scientists have to sift through enormous chemical libraries, testing millions of molecules in search of the right match, with the most promising compounds often failing in clinical trials.

But the biggest problem? Most potential drugs have never even been conceptualized.

That’s why Nimbus Synergies invested in Variational AI, because they aren’t just searching for drugs. They’re imagining them.

Why Traditional Drug Discovery Is Due for Disruption

For decades, drug discovery has revolved around screening libraries of molecules to see if any show promise. These screens can be virtual (using computational models) or experimental (using high-throughput lab tests), but they all suffer from the same fundamental issue: they’re limited to what already exists.

The chemical space of potential drugs is astronomically large, estimated at 10⁶⁰ possible small molecules. To put that in perspective, if every star in the observable universe had a billion planets, and every planet had a billion trees, and every tree had a billion branches, and every branch had a billion leaves, there would still be fewer leaves than possible drug molecules. Traditional approaches barely scratch the surface. At best, all of our searching to date has covered ~10^10 possible small molecules in chemical space. To continue the analogy, as of today we’ve looked at the leaves of approximately 10 trees… all within the same grove, on the same planet.

Even AI-driven virtual screening methods only rank existing molecules for potential effectiveness, which doesn’t solve the problem, they just speed up the same old process. This inefficiency isn’t just costly for pharmaceutical companies, it delays life-saving treatments for patients who desperately need them.

Most drug discovery efforts fail not because researchers can’t find molecules, but because they spend years pursuing ones that ultimately don’t work. A drug candidate might bind to a target but be too toxic, unstable, or impossible to synthesize.

Variational AI does something different.

Instead of screening known molecules, Variational AI’s Enki™ platform creates new ones from scratch from the chemical space we haven’t yet explored. This is a shift from selection to generation—from finding molecules to ideating them.

Think of it like an AI-powered architect for drug design. Scientists can feed in the biological target, and Enki™ generates molecular blueprints that fit perfectly. It doesn’t just guess; it crafts precision-designed solutions, with the right balance of potency, safety, and drug-likeness.

Unlike traditional AI models that rely on massive proprietary datasets, Enki™ is trained on decades of pharmaceutical data, allowing it to generate promising compounds even in areas where little prior knowledge exists. And unlike brute-force screening, which tests molecules in bulk, Enki™ continuously refines its designs, learning from every iteration to produce better candidates in a fraction of the time.

The impact is real. Recent benchmarks from Variational AI’s research team show that just 100 AI-generated molecules are as effective as a high-throughput screen of 1,000,000 compounds. That’s not just an efficiency gain, it’s an entirely new paradigm for drug discovery.

For pharmaceutical companies, this is a game-changer. Rather than waiting years to test countless compounds, researchers can start with AI-designed molecules that are already optimized for success. That means fewer failures, lower costs, and faster timelines to get life-saving treatments to the people who need them most.

Variational AI has already gained traction with biopharmaceutical partners, working on next-generation drugs for oncology, immunology, and other high-impact therapeutic areas. Their recent oversubscribed $5.5 million financing round (which Nimbus proudly led) will help scale its AI-powered drug discovery model and expand its impact across the industry.

Why We Invested

At Nimbus Synergies, we don’t just invest in companies that speed up old processes, we invest in those that fundamentally change the game: Variational AI represents a paradigm shift in how drugs are discovered, moving from a slow, trial-and-error approach to a precise, AI-driven design process.

Beyond the technology, what really stood out to us was the team. Handol Kim, co-founder and CEO, is a three-time tech entrepreneur with over 20 years of leading ventures across AI/ML, advanced computing, and enterprise software. He’s held senior roles in both startups and public companies across Silicon Valley, Canada, and Asia-Pacific, and brings a blend of technical depth and commercial acumen. On the scientific side, Dr. Jason Rolfe, co-founder and CTO, has more than 18 years of experience developing novel generative machine learning algorithms in unconventional domains. With a PhD in machine learning and theoretical physics, his work bridges deep scientific research and real-world application. Together, Handol and Jason have assembled a team of scientists and engineers from MIT, Caltech, Google, Microsoft, and D-Wave, with a collective portfolio of 124 patents, blending scientific rigor with product vision—exactly the kind of leadership we look for in our portfolio companies.

We first met the Variational AI team several years ago when they were raising their first round of financing. From those early days, it was evident they were on the path to building a leading generative drug design platform. Recent milestones such as their partnerships with Rakovina Therapeutics and ImmVue Therapeutics represented a tipping point in the maturation of the technology and validation that the time was now to accelerate their growth. Also impactful in our conviction was the timing. Just a few years ago, generative AI in drug discovery was viewed as promising but premature—particularly by medicinal chemists, many of whom were skeptical or outright dismissive of AI’s role in their work. But that sentiment has shifted. With rising R&D costs and increasing pressure to accelerate timelines, urgency is replacing hesitation. The zeitgeist has evolved, and there’s now far greater openness among drug developers to explore AI-powered approaches. The demand for smarter, more efficient tools has never been greater—and Variational AI is already delivering. While many AI-driven drug discovery startups promise breakthrough science but struggle to show real-world impact, Variational AI is proving that its technology works. They’re generating AI-designed drug candidates for oncology, neurology, and rare diseases, and their molecules have already demonstrated competitive properties compared to those discovered through conventional methods. The pharmaceutical industry is at a turning point. AI-driven design isn’t just a competitive advantage – it’s becoming the new standard for drug discovery. With Enki™, Variational AI is leading the charge, and at Nimbus Synergies, we’re proud to be part of their journey.

Are you a biopharma company looking to accelerate drug discovery? Connect with Variational AI here.
Interested in Nimbus Synergies’ investments? Learn more about our portfolio here.

[1] “Research & Development | PhRMA.” 2025. Phrma.org. 2025. https://www.phrma.org/policy-issues/research-development.