AI drug discovery in 2026: 15–20 AI-born drugs race into Phase III — how can ordinary investors participate?
For AI drug discovery, 2026 is the year of validation: 15–20 AI-designed or AI-discovered programs are expected to enter pivotal Phase III trials, Isomorphic Labs raised $2.1B, and Lilly and Novartis are betting big. We map the public-market entry points — from Tempus (an ARK add) to Recursion to the picks-and-shovels play, Nvidia — plus the risks unique to this lane.
Why 2026 is AI pharma’s year of validation
For a decade, “AI-designed drugs” lived mostly in papers and seed rounds; in 2026 they face the hardest referee — pivotal Phase III trials. Industry trackers expect 15–20 AI-designed or AI-discovered programs to enter Phase III this year; DeepMind spinout Isomorphic Labs (born from the AlphaFold team) closed a $2.1B raise in May, and founder Demis Hassabis has publicly committed to putting its first AI-designed candidate into Phase 1 within the year. When data lands, the valuation logic flips from narrative to pipeline.
Big pharma has already voted with its wallet
Eli Lilly is the most aggressive: over $1.7B in milestones with Isomorphic plus an up-to-$1B AI-infrastructure commitment with Nvidia; Novartis has ~$1.2B with Isomorphic, expanded in 2025. The math is easy for big pharma — a traditional drug averages 10 years and $2B+, so if AI compresses preclinical work toward ~18 months, the saving is not just money but patent runway.
Public-market entry points: three layers
The purest names are mostly private (Isomorphic, Chai), so public exposure comes in three layers: (1) the data & trial-matching layer — Tempus AI (TEM), matching patients via genomic + clinical data; ARK kept adding in Q1 2026, making it the only direct AI-healthcare holding among our tracked legends; (2) the AI-pipeline layer — Recursion (RXRX), Schrödinger (SDGR), Relay (RLAY) and 2026 IPOs Generate and Eikon — maximum torque, maximum risk; (3) the picks-and-shovels layer — Nvidia (compute plus the BioNeMo platform for pharma) and the clouds — highest certainty, lowest “drug content.”
Risks unique to this lane
Unlike chips or cloud, AI pharma is ultimately judged by biology and regulators: Phase III failure rates are inherently high, and so far AI has demonstrably improved speed and cost, not yet success rates; most names are loss-making and financing-dependent, so valuations get crushed whenever rates or risk appetite tighten; and platform revenue (data, software) blends with pipeline revenue (milestones, royalties), making valuations easy to misread. Size it as a high-risk satellite, not a core holding.
Calibrate enthusiasm with consensus data
Note the proportion: among our eight tracked legends, only Cathie Wood directly holds an AI-healthcare name (TEM) — institutional consensus hasn’t formed, which means both early-mover opportunity and a lack of smart-money validation. By contrast, the memory chain and power already have multiple legends invested with real money. Run TEM through the Portfolio Check to see its gap versus strong-consensus names. Educational content, not investment advice; investing carries risk.
Sources: IntuitionLabs — Isomorphic Labs $2.1B Series B analysis · AIM Media House — 2026 is the year AI drug discovery meets clinical reality · MarketWise — 6 AI drug discovery stocks to play the Isomorphic trade · Vision Life Sciences — AI drug discovery companies: the 2026 power list · Seeking Alpha — ARK Invest Q1 2026 13F (TEM added)