Sanofi rents the same AI brain as AstraZeneca — and that's the story
Sanofi becomes the second major pharma in three weeks to license Owkin's K Pro. BMS is now counting the bill on its 30,000-seat Claude rollout. Plus: 'first AI drug in the clinic' unpacked, the AI-funding bifurcation, and the Onion Desk.
When a second major pharma signs the same agentic-AI platform as the first, within three weeks, the story stops being about either deal and starts being about the model. Sanofi's five-year K Pro license follows AstraZeneca's three-year license of the same Owkin platform by roughly 21 days. The technology isn't new — Sanofi has had a relationship with Owkin since 2021 — but the scope is: agentic systems that plan and execute multi-step research tasks across the full pharmaceutical value chain, deployed inside the pharma's own IT environment. Two top-ten companies converging on identical vendor infrastructure in the same quarter is a signal worth examining before the third announcement lands.
Lead — Sanofi rents the same AI brain as AstraZeneca
Sanofi signed a multi-year, five-year-licensed collaboration with Owkin for K Pro — the same agentic-AI platform AstraZeneca licensed roughly three weeks prior. Both contracts deploy agents within the pharma's IT environment so proprietary data never leaves internal systems; AstraZeneca's initial focus was competitive intelligence while Sanofi's scope covers broader co-development workflows. The licensing model represents a third option beyond traditional build-versus-buy: no acquisition, no hyperscaler dependency, recurring contract. That contrasts with the ownership plays dominating 2026 — AstraZeneca acquiring Modella AI, Lilly's $1B Nvidia co-innovation lab, BMS deploying Claude across 30,000+ staff. Q1 2026 saw 36 major pharma-AI partnerships, double the prior year, and two of the largest converged on the same platform. When licenses expire, the strategic question is whether the institutional knowledge encoded in those agents travels with the vendor.
On our radar
Two weeks after going all-in on Claude, BMS is now addressing consumption-based pricing costs — the first public acknowledgment from a major pharma of enterprise-AI cost management at scale. Deployment-at-scale is a solved problem; 'value per query' is not, and pharma currently lacks agreed methodologies for computing AI ROI. Similar challenges await Merck-Google and Novo Nordisk-OpenAI. Separately: CRDMO Quotient Sciences dosed the 'first AI drug' in Phase I following MHRA approval — except the AI optimized the formulation, not the molecule. As 'first AI drug' claims proliferate, the distinction between AI-discovered, AI-designed, and AI-formulated will increasingly separate genuine clinical milestones from marketing language. And early-stage investor Nirmesh Patel of Amino Collective named a bifurcation already visible to founders: companies with AI components versus those without are accessing meaningfully different capital. That dynamic actively incentivizes AI-washing, and the real test arrives when AI-focused startups maintain valuation premiums following clinical failures.
Quick signals
Flourish raised $500M for biology-inspired AI foundation models — a substantial bet on biology-native architectures over scaled transformers. Pfizer and Innovent signed a $10.5B oncology deal with zero AI components, underscoring that the largest capital deployments this week still targeted de-risked assets and China-originated pipelines. Nature Machine Intelligence published an explainable protein language model perspective — a regulatory signal, not a research curiosity: agencies won't accept model-generated design recommendations without explainability. Context for the ongoing approved-drugs debate: roughly 460 AI drug-discovery startups have absorbed ~$60B to date.
The Onion Desk
This week: half of pharma executives cite 'trust' as an AI-in-trials barrier; the other half trust AI's own reassurances that the concerns are unfounded. Lilly's communications team spent six hours analyzing a $1.9B deal with no AI before realizing timelines can accelerate through traditional methods like 'money' and 'scientists.' And pharma celebrated 36 Q1 AI partnerships and zero AI-approved drugs while insisting on a call that efficacy remained unsubstantiated. Full satirical dispatches on Substack.