Who’s building the AI drug-discovery loop — and where agents fit

Atlas contribution: competitive landscape and US prospecting intelligence Research cut-off: 18 October 2026. Primary company disclosures and federal sources; selected landscape, not an exhaustive ranking. Pipeline stages are attributed to the cited disclosures, not independently audited trial outcomes. Sources and financial figures retain their announcement dates.

1. Strategic assessment

The commercially relevant frontier is an experimentally grounded loop: proprietary biological data, computational prediction, molecule design, laboratory validation and clinical translation. These capabilities appear in different businesses: therapeutic developers, software licensors, patient-data platforms and compute suppliers. Treating them as one “AI drug discovery” market obscures both competition and buying needs.

Three developments deserve attention. First, AI-native companies increasingly have clinical assets rather than model demonstrations alone. Second, large pharma is building internal infrastructure while continuing external partnerships. Third, repurposing has a distinct public-interest funding route: existing medicines can generate valuable new-indication hypotheses without the commercial exclusivity of a newly invented molecule. None of these developments establishes that AI has solved clinical attrition.

2. Selected US companies and actionable positioning

The final column is analyst inference about a potential operational use case, not evidence of purchasing intent, an existing relationship or an invitation to contact patients.

Exscientia should not be counted as a separate independent prospect without qualification: its combination with Recursion completed in November 2024. [Completion announcement](https://ir.recursion.com/news-releases/news-release-details/recursion-and-exscientia-two-leaders-ai-drug-discovery-space)

3. Partnerships, capital and the competitive boundary

Pharma partnership networks are major distribution channels. Recursion's published relationships include Roche/Genentech, Sanofi and Bayer. Its August 2026 release distinguishes $216 million already received from Roche/Genentech from potential future payments. The move from a validated neuroscience target into discovery is meaningful, but remains far earlier than clinical efficacy. [Partnerships](https://www.recursion.com/partners), [August update](https://ir.recursion.com/news-releases/news-release-details/recursion-reports-second-quarter-financial-results-genentech)

insitro's BMS expansion demonstrates continuing target-discovery demand, while its September 2025 Lilly agreement addresses prediction of small-molecule pharmacology. Together, these illustrate competition across both biological target selection and drug-property prediction—not one interchangeable “AI platform” function. [Lilly agreement](https://www.insitro.com/news/insitro-partners-with-lilly-to-build-first-in-kind-machine-learning-models-to-advance-small-molecule-drug-discovery/)

Capital categories must remain separate. Lilly/NVIDIA announced up to $1 billion over five years for talent, infrastructure and compute: a planned combined investment, not a financing round or annual revenue. Genesis reports $150 million in upfront consideration, including $40 million equity, across its 2025 Incyte agreement and 2026 expansion. These mixed commercial/equity proceeds should not be treated as pure venture funding. [Joint lab](https://lilly.gcs-web.com/news-releases/news-release-details/nvidia-and-lilly-announce-co-innovation-ai-lab-reinvent-drug), [Genesis](https://www.genesis.ml/partners-pipeline)

For a global comparator, London-based Isomorphic Labs announced a $2.1 billion Series B on 12 May 2026, led by Thrive Capital. Its relevance to US competition is technological and partnership-based; it is deliberately excluded from the US-company table. The announcement describes investment in its design engine and pipeline progression toward the clinic, not demonstrated clinical success. [Financing announcement](https://www.isomorphiclabs.com/press/isomorphic-labs-funding)

These disclosures suggest that proprietary data, experimental throughput, development expertise and financing endurance remain competitive assets. Agent orchestration can complement them; it does not replace laboratories or clinical evidence.

4. Repurposing, evidence and prospecting priorities

Repurposing is not synonymous with de novo design or merely applying a new AI model to an old target. FDA defines it around potential new uses of approved medicines supported by safety and effectiveness data. Every Cure is a particularly relevant US example: its February 2026 announcement describes up to $76 million in a new ARPA-H phase for preclinical and clinical validation. This is conditional program funding, not equity financing, and program-specific clinical success remains to be demonstrated. [FDA definition](https://www.fda.gov/drugs/resources-drugs/drug-repurposing), [Every Cure funding](https://everycure.org/phase3/)

For the council's prospecting list, prioritize organizations with many parallel programs, external partners and documented evidence handoffs. Start with research/program operations or informatics buyers; propose a bounded pilot tracking literature, experiments, review decisions and unresolved dependencies. Measure missed tasks, unsupported closure, reviewer time and successful handoffs. Claims about faster medicine discovery require separate scientific evaluation.

Capture company/domain, US presence, modality, program stage, dated catalyst, evidence URL, functional buyer, proposed use case and verification date. Keep patient data out of this seed list; identify public business contact routes in a later verification pass. No outreach was conducted for this report.

FDA's January 2025 AI guidance remains labelled draft on the consulted page and describes a risk-based credibility approach for the intended context of use. Our architecture's status firewall can support workflow accountability, but is not clinical validation, regulatory clearance or proof that an AI-selected therapy works. [FDA draft guidance](https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-artificial-intelligence-support-regulatory-decision-making-drug-and-biological)

Mission provenance: Tracked through a dedicated AUTO-LEDGER at /home/claude/biomed-mission/atlas_ledger.sqlite, using deterministic no-inference audits. Status updates were made by the research operator after checks, not by Gemma. This run demonstrates bounded research tracking; no context reset or unattended multi-day recovery was tested. No Gemi Live, Asterisk or VoIP activity was initiated.