AI and Biology Are Converging — Again, but With Different Capital Discipline

A new wave of AI-driven drug discovery emphasizes infrastructure over promises, open science over secrecy, and measured ambition over sweeping claims. Investors are placing smaller bets on tools and platforms to redefine biomedical research and its commercial potential — signaling a shift not just in funding patterns, but in how society approaches the business of breakthroughs.

The first generation of AI-biotech startups promised disruption. Sweeping claims heralded technology that could compress drug development timelines, reduce costs, and cure previously untouchable diseases. Yet, many companies underdelivered. Investors poured billions into therapeutic ventures that struggled when AI collided with the complexity of biology. This time appears different.

Boltz, Aurora Therapeutics, Topos Bio, and others have collectively raised less than one hundred million dollars, a paltry figure compared to early attempts to pivot AI into drug development. But their focus is narrower: building foundation models that researchers can reliably integrate, and tools that enable faster iterations while reducing operational overhead. As Dylan Reid, managing director at Zetta Venture Partners, observed, the conventional biotech model — which often involves raising hundreds of millions for one major shot at creating a breakthrough drug — seemed “super disconnected” from how technology evolves.

“We believe in a future where every scientist can iterate at the speed of inference,” Gabriele Corso, co-founder and CEO of Boltz, wrote in a statement. The $28 million seed round backing his company is emblematic of the new approach, emphasizing platform infrastructure over single-shot therapeutic bets. Corso’s team has built open-source AI models for biomolecular structure prediction, hoping to democratize the tools needed to push drug discovery into higher gear. He envisions a world where scientists prioritize end-to-end systems: “tools that are reliable, scalable, and require minimal operational overhead.”

This shift is happening against a backdrop of heightened skepticism. Earlier AI-drug startups were pitched as engines of radical disruption, producing press-ready claims that often failed to deliver. The new generation seems to emphasize discipline over hype. For example, Topos Bio is focused entirely on modeling the dynamics of intrinsically disordered proteins — a thorny frontier in drug discovery long considered "undruggable" by traditional methods. CEO Ryan Zarcone calls these structures “one of the last major frontiers,” explaining that their Topos-1 platform operates by modeling protein dynamics as ensembles rather than static snapshots. With ten point five million dollars in funding, it is a restrained pitch that acknowledges the limits of what any individual platform might achieve.

Corso’s critique of earlier models revolves largely around access: while early AI platforms were typically paywalled or closed-loop, Boltz aims to operate openly, mirroring a culture more common in academic settings than commercial ones. “We did our PhDs at MIT CSAIL, where patenting is rare and research is typically shared openly,” Corso stated. That ethos underpins why Boltz operates as a public benefit corporation, with open-source development prioritized over proprietary tools.

But even among this more technically rigorous cohort, the debate over AI’s promises continues. Ron Alfa, CEO of Noetik, still frames his company’s agreement with GSK as transformative. “Processing hundreds of tumor samples per week now to train the largest cancer foundation models. AI will cure cancer,” Alfa posted on X. Yet his five-year licensing deal for foundation models illustrates the same platform-first mentality driving competitors. Rather than leaping to results, capital is flowing toward systems that refine molecular prediction, optimize iterative processes, and enable researchers to make smarter bets during discovery — leaving clinical development timelines intact.

“We’re shifting from pilot to platform,” states a 2026 report from Benchling, which identifies growth in investments tied to infrastructure AI as opposed to therapeutic AI. It’s a pivot that reflects realism about biology’s difficulties and the bottlenecks AI cannot yet bypass. While some voices frame reductions in time and cost as compression on unprecedented scales (“down to five years and $200 million per drug”), others note that discovery gains cannot skip clinical trials or regulatory hurdles. As Aadit Sheth wrote on X, AI may primarily affect “what trials we run, making sure they’re the right ones.”

Founders like Jennifer Doudna aren’t ignoring these complexities either. Aurora Therapeutics, her company focused on personalized CRISPR-based therapies, raised $16 million for bespoke gene-editing tools targeting rare diseases. Still, the funding remains modest compared to earlier generational rounds hyped for their supposedly transformative impact.

For investors, the financial discipline shaping these bets reflects not cynicism but realism. The biotech vision powered by AI now revolves around modular systems solving specific problems rather than holistic transformation. Platforms allowing researchers to iterate faster or identify better candidates resemble moves seen in software development more than pharma’s usual playbook. The second-order effects of this trend will likely ripple outward, influencing trends in intellectual property, education, and industry norms as technical collaboration becomes the ultimate differentiator.

Yet questions remain unanswered. What unintended consequences will arise as the industry becomes increasingly tool-driven? Who owns the intellectual scaffolding for models trained openly but deployed commercially? How will power dynamics shift if infrastructure AI becomes dominated by entities willing to license models for specific sub-populations rather than broadly intended cures? Translational concerns loom especially large: how far can AI systems compress biological time before human trials inevitably reassert themselves as slow bottlenecks?

For now, what’s clear is that ambition alone won’t drive the convergence between AI and biology. The new cohort’s efforts emphasize knowledge-sharing, high technical standards, and realistic constraints. Whether such discipline recalibrates public trust remains unanswered, but from investors to founders, messaging now steers toward tools not cures, infrastructure not promises.