A Five-Part Series on AI Drug Discovery Valuation Challenges
Introduction
I have no idea.
That was my brilliant, insightful, memorable, put this on a t-shirt answer to the question posed to me regarding AI and drug candidate valuation.
I’ve considered this question in the past. But my answer has a generally been a negative one. After all, does it really matter where the candidate came from?
Does saving a few months and a few hundred thousand dollars really impact the value of a drug candidate?
While this may have been the case a few short years ago, it is simply no longer true (if it ever was).
Today, AI drug discovery tools and technologies are generations ahead of where they were only a few short years ago.
And the data suggest that the industry (both Licensees and Investors) feel the same.
In 2025, AI-driven drug discovery deals totaled over $43 billion in biobucks across 114 transactions, up from $11.8 billion across 84 deals in 2024 (DealForma). Companies like Isomorphic Labs, Recursion, and XtalPi are signing partnerships worth billions with Big Pharma. Meanwhile, AI-native biotech companies are fetching nearly 100% valuation premiums over traditional biopharma valuations.
But a fundamental question emerges for BD&L teams, “Do AI-discovered drug candidates actually warrant different valuations than traditionally discovered assets?”
Or are we simply witnessing a market cycle where buzzwords drive premiums that may not reflect underlying value?
Over this five-part series, we’ll examine some of the unique valuation challenges that AI-discovered candidates present for licensing professionals.
We’ll explore the valuation premium question, IP ownership ambiguity, quantifying AI’s contribution, platform versus molecule economics, and data quality considerations.
Part I: The Valuation Premium Question
Over the past 12 months, VC invested $3.2 billion across 135 AI drug development startups, paying significantly higher multiples than for traditional discovery companies.
This suggests investors are betting that AI discovery platforms will generate more and larger licensing deals than their conventional counterparts.
But is this true?
On the one hand, there is some truth to this idea.
First, speed to clinic matters…a lot. If AI truly reduces discovery timelines by 40-60% as some industry analyses suggest, the net present value of earlier cash flows and extended patent life could be substantial.
Second, if AI platforms can identify candidates with superior safety or efficacy profiles by screening vastly larger chemical spaces, the resulting Probabilities of Success and Likelihood of Approval should be higher, thereby justifying premium pricing.
However, the evidence remains mixed. While deals like XtalPi’s $6 billion partnership with DoveTree or Helixon’s $1.7 billion licensing agreement with Sanofi demonstrate market willingness to pay extraordinary sums for AI-discovered assets, the deal structures themselves suggest uncertainty.
The XtalPi-DoveTree deal, for instance, includes only $51 million upfront against $5.9 billion in potential milestones. This Upfront/Deal Value ratio that suggests the parties are hedging heavily on unproven value rather than paying for demonstrated superiority.
As one industry analysis notes, the shift toward milestone-based structures that share risk has become a defining feature of 2025-2026 AI licensing deals. This allows licensees to cap downside while giving licensors a route to full value if the science delivers.
However, is this backloading driven by uncertainty about AI-derived candidates specifically, or simply standard practice for early-stage assets regardless of discovery method?
We think it is a little of both.
The shift toward contingent payments linked to trial outcomes or sales thresholds is common across all early-stage licensing deals, AI or otherwise.
However, the particularly dramatic ratios in some AI deals (with Upfronts representing less than 1% of total potential value) may reflect additional uncertainty about whether computational discovery actually produces superior candidates.
If the relatively modest upfront payments in AI deals are indeed driven partly by uncertainty about the AI contribution’s value, we should (in theory) see Upfronts rise over time as clinical data validates (or fails to validate) the computational approach.
If AI-discovered candidates consistently demonstrate faster development timelines, higher clinical success rates, or superior commercial performance, licensees should become willing to pay more upfront for computationally-derived assets.
Conversely, if AI candidates perform no better than traditional discovery, the “AI premium” in deal economics should disappear entirely.
Thus, to rephrase the question, “Should licensing terms for AI-discovered candidates include performance-based adjustments that allow valuation to track with demonstrated advantages, rather than paying the premium Upfront based solely on discovery method?”
We think the answer depends on resolving several other valuation challenges first, starting with the murky question of who actually owns the IP when AI generates a molecule.
We’ll tackle that question in Part II of this Series.
Key Takeaways
- AI-native biotechs command valuation premiums in venture markets, but this doesn’t automatically justify premium licensing terms.
- Deal structures with dramatic milestone backloading (Upfronts <1% of total value) suggest market uncertainty about AI’s actual value contribution…only time will tell if this is the case.
- Premium pricing may be justified if AI candidates demonstrate superior speed, safety, or efficacy. But this requires clinical validation and approvals, not just computational predictions.
- BD teams
shouldmust consider performance-based deal structures that tie AI premiums to demonstrated advantages rather than paying upfront for “AI-ness”.
Coming Up in Part II: IP Ownership and Inventorship Ambiguity
When an AI platform generates a novel molecular structure, who is the inventor?
The AI company, the pharma partner, the data scientists, or the medicinal chemists who validated the output?
This legal ambiguity creates IP risks that directly impact asset valuations and licensing negotiations in ways that don’t exist with traditional discovery.
In our next post, we’ll explore how inventorship uncertainty is shaping deal structures and what BD teams can do to mitigate these risks.
NB: LLMs were used for some of the research aspects of this post.