AI-Native Drug-Discovery Competitive Positioning for a Top-20 Pharma

Replacing a fragmented partnership posture with a structured build-buy-partner framework

Client context

The client is a top-20 pharma R&D function with multi-decade discovery capabilities and three active partnerships across AI-native operators. The partnerships were established at different times under different leadership, with different commercial structures and overlapping scope. Internal sentiment ranged from "AI is the next platform" to "the clinical evidence is not there yet"; the partnerships reflected that divergence rather than resolving it.

The category had shifted decisively over 18 months. Insilico's rentosertib became the first AI-discovered drug with peer-reviewed Phase IIa proof in June 2025. The Recursion-Exscientia merger closed in November 2024, signalling category consolidation. The NVIDIA-Lilly US$1 billion co-innovation lab announcement (January 2026) reframed compute substrate as a multi-decade strategic relationship rather than a procurement line item. Cumulative AI-native biotech VC funding reached approximately US$19.9 billion across 2024–25.

By Q1 2026, the executive committee had three concerns. First, was the existing three-partner posture coherent or accidental? Second, were the capability gaps real, and which operators actually closed them? Third, what posture should the FY2026 R&D capital allocation cycle reflect?

We were brought in to provide the structured read.

The decision context

Three connected decisions framed the engagement:

• Capability posture: build internally (in-house AI), buy (acquire an AI-native), or partner, and at what scope across target identification, lead optimisation, and clinical biomarker discovery?
• Partner consolidation: rationalise the three existing partnerships into a coherent posture, or expand to a fourth?
• Investment thesis: what evidence threshold should the FY2026 capital allocation cycle apply, given the 15–20 AI-discovered Phase II readouts expected in 2026–28?

The brief was scoped to produce an investment-committee-ready framework before the FY2026 capital cycle.

Scope and methodology

The review combined four workstreams: operator capability mapping, clinical-evidence assessment, deal-structure benchmarking across 22 transactions, and synthesis into a build-buy-partner framework anchored to the client's three open decisions.

What we found

Clinical evidence is no longer speculative, but it is not uniform. Insilico's June 2025 Phase IIa proof shifted the category baseline. Phase I success rates for AI-discovered drugs reached 80–90% (versus 50–65% historical baseline) across the cumulative 2015–24 sample of 75 AI-discovered drugs entering trials. Phase II evidence remained limited, with approximately 40% success rate against a small sample. The category thesis is conditionally validated, gated on the 2026–28 readout cluster.

Operator capability segments more cleanly than partnership-shortlist exercises typically reveal. The nine operators we mapped split into three structurally distinct capability groups: physics-based (Schrödinger, Isomorphic), data-and-platform (Recursion, Tempus, Generate), and target-discovery-focused (Insilico, Atomwise, Iktos, Owkin). The client's three existing partnerships covered two of the three groups; the third was structurally absent. The fragmentation was real, not interpretive.

Deal-structure benchmarks reveal the bankability tier. Across 22 transactions teardown, structurally durable partnerships shared three features: upfront payments above US$50 million (signalling multi-decade commitment), milestone potential above US$1 billion (enabling pharma capital recycling), and IP allocation that left lead optimisation with the AI-native (preserving operator scale economies). The Schrödinger-Novartis (US$150 million upfront, US$2.3 billion milestones), Isomorphic-Lilly (US$45 million upfront, US$1.7 billion milestones), and Generate-Novartis (US$65 million upfront, over US$1 billion in milestones) deals anchored the bankable tier.

Strategic outcomes

The review produced four shifts in the FY2026 R&D capital cycle:

• Three-track posture. Internal in-house AI built for target identification (highest defensible capability). Partnership for lead optimisation, expanded with one existing operator on a multi-year commitment. Acquisition evaluated for clinical biomarker discovery, contingent on Phase II readout cluster.
• Partnership rationalisation. Two existing partnerships expanded to multi-year commitments at the bankable-tier deal structure benchmark. One existing partnership flagged for orderly wind-down on capability-overlap grounds.
• Investment thesis. FY2026 capital allocation set at a base case sized to the conditional clinical thesis, with explicit gates tied to the 2026–28 Phase II readout cluster. A reserve allocation released contingent on the readout outcomes meeting the historical industry baseline or above.
• Substrate posture. NVIDIA BioNeMo formally adopted as the compute-substrate standard, mirroring the Lilly co-innovation template.

Why this case matters

A pattern keeps recurring across pharma R&D engagements with platform-shifting categories:

• Partnership postures accumulate transaction-by-transaction and become illegible without periodic structured rationalisation
• Clinical-evidence thresholds move faster than internal capital-allocation frameworks update; the 2025 Phase IIa proof reset the baseline that 2024 capital cycles had not modelled
• Operator capability segments are typically over-merged in shortlist exercises; the genuine three-group split changes which operator does what

The fix is not more partnership analysis. It is structured competitive intelligence tied to specific R&D capital decisions, with findings translated into build-buy-partner framework rather than scoring.

How we helped

We designed and ran a six-week competitive-intelligence review tailored to the client's three open R&D capital decisions, combining operator capability mapping across nine AI-native operators, clinical-evidence assessment against the 2026–28 readout cluster, deal-structure benchmarking, and a build-buy-partner framework. The output was scoped to be capital-committee-ready, not comprehensive, and was carried directly into the FY2026 R&D capital allocation cycle.

If you are weighing AI-native partnership posture, build-buy-partner decisions across discovery capabilities, or investment thesis sizing against a clinical-evidence cluster, contact us to discuss how structured competitive intelligence can support your R&D capital cycle.