Platforms, Not Products

AI becomes pharma’s operating system. Diagnostics unlock markets. And Eli Lilly keeps buying the future.

⚡ Quick Hits

💰 Money Moves

Eli Lilly to acquire Ajax Therapeutics (up to $2.3B) — picking up AJ1-11095, a first-in-class Type II JAK2 inhibitor in Phase 1 for myelofibrosis and polycythemia vera. Unlike approved JAK2 inhibitors that bind the active conformation, AJ1-11095 targets the inactive form — designed to deliver deeper, more durable disease control and rescue patients who lose response to Type I inhibitors. Lilly →
Eli Lilly to acquire Kelonia Therapeutics (up to $7B / $3.25B upfront) — Lilly’s second in vivo CAR-T deal of 2026 (after Orna in February). Kelonia’s iGPS lentiviral platform reprograms T-cells inside the body, eliminating leukapheresis, ex vivo manufacturing, and lymphodepleting chemotherapy — replacing all three with a single IV infusion. CAR-T moves from product → platform. Fierce Biotech →

🤖 AI Watch

Merck & Google Cloud announce up to $1B agentic AI partnership (Cloud Next ’26) — multi-year deal deploys Gemini Enterprise across R&D, manufacturing, commercial, and corporate functions, with Google Cloud engineers embedded alongside Merck teams. The first publicly disclosed deal of this scale where a top-five pharma commits to agentic AI as full-stack infrastructure, not pilots. Merck →
AWS launches Amazon Bio Discovery (AWS Life Sciences Symposium) — agentic AI platform pairing 40+ biological foundation models with an AI agent layer and direct routing of candidates to wet-lab partners (Twist Bioscience, Ginkgo Bioworks, A-Alpha Bio). Memorial Sloan Kettering reportedly compressed antibody design for pediatric cancer therapies from months to weeks. AWS →
IQVIA unveils IQVIA.ai unified agentic platform on NVIDIA (NVIDIA GTC) — combines IQVIA’s healthcare-grade data with NVIDIA Nemotron and NeMo Agent Toolkit. Already deployed by 19 of the top 20 pharma companies — a neutral counterweight to single-cloud lock-ins. IQVIA →
Novo Nordisk — OpenAI strategic partnership — announced the same week as Amazon Bio Discovery and just days before Merck/Google. With this deal, every top-10 pharma is now publicly aligned with a major AI vendor. The neutral middle is gone.
CZI Biohub launches the Virtual Biology Initiative — a global effort to create the open data foundation for AI-accelerated biology. While the hyperscalers race to commercialize agentic AI, Biohub is building the public counterweight: an open dataset to keep next-generation biology AI from being walled into one or two cloud ecosystems. PRNewswire →

📋 Policy & Payer

CMS ACCESS Model — 150+ providers and digital health companies approved — first major federal experiment in outcome-based reimbursement for technology-supported chronic care. Participants receive ~$7.50–$35 per beneficiary per month for managing hypertension, diabetes, kidney disease, anxiety, and depression — with full payment tied to measurable outcomes, not visits. Healthcare Dive →
CMS + FDA propose new fast-track Medicare coverage pathway for Breakthrough Devices — addresses the long-standing complaint that an FDA Breakthrough designation meant little because Medicare reimbursement could lag by years. Material upside for every AI-enabled diagnostic, digital therapeutic, and remote monitoring device in the pipeline. STAT News →

💊 Pharma & Diagnostics Corner

Revolution Medicines — pancreatic cancer win as “tip of the iceberg” (AACR 2026) — RAS has been the most-wanted, hardest-to-drug oncology target for four decades. A clinical signal in pancreatic cancer — historically one of the most lethal solid tumors — suggests RAS is shifting from undruggable → platform class. CRC and lung are next. Fierce Biotech →
Labcorp launches FDA-approved companion diagnostic for KEYTRUDA in platinum-resistant ovarian cancer — Agilent’s PD-L1 IHC 22C3 pharmDx now nationally available. Not a drug story — a distribution story. ~80% of ovarian cancer patients relapse; without testing at scale, eligibility never translates to treatment. Access → utilization → market expansion. Labcorp →
SERENA-6 (camizestrant) — FDA ODAC review — AstraZeneca’s oral SERD shown to deliver a 56% PFS improvement when used to switch therapy at the moment ESR1 mutation is detected in ctDNA — before radiographic progression. The FDA has flagged immature OS data and questioned whether a molecular signal alone should drive treatment shifts. Whichever way ODAC votes, this sets precedent for how regulators evaluate ctDNA-guided treatment decisions across oncology. FDA →

📡 Deep Dive

The Pharma AI Platform Wars

In a 14-day stretch in April 2026, the four biggest hyperscaler/AI vendors each planted a flag in pharma. These deals aren’t four flavors of the same thing — they reflect three distinct architectural bets on what “agentic AI in pharma” actually means.

🏛️ Four players. Three architectural bets.

1. The vertical-stack bet — Merck + Google Cloud. One pharma, one cloud, one model family, embedded engineers. Maximum integration, maximum lock-in. Best if you believe AI capability compounds within a single environment.
2. The marketplace bet — AWS Bio Discovery. A neutral platform of multiple foundation models, with the wet lab as a service routed through partners. Best if you believe the model layer is commoditizing and the moat is the experimental loop.
3. The domain-specialist bet — IQVIA.ai on NVIDIA. A pharma-native agentic platform leveraging proprietary clinical and real-world data. Best if you believe the moat is healthcare-grade data and regulatory fluency, not raw model capability.
4. The frontier-model bet — Novo Nordisk + OpenAI. Best-available reasoning, applied across the pipeline. The fourth flag in two weeks confirms: every top-10 pharma is now aligned with a major AI vendor.

🔍 Three questions to watch over the next two quarters

Does Merck’s $1B commitment translate into measurable cycle-time reductions in any single function (regulatory submissions, manufacturing changeovers, MSL knowledge management)?
Will AWS’s lab-in-the-loop architecture pull traditional CROs into reactive partnerships, or push them to build their own agentic stacks?
How will the EU AI Act (effective April 6, 2026) treat agentic systems making research recommendations? The first enforcement action against a pharma agent will reset the entire risk calculus.

👉 Agentic AI has crossed from tool to operating system for pharma. The question is no longer whether — it’s whose stack runs the next decade of drug development.

💡 Maria’s Take

Lilly Is Buying the Future. The Question Is Whether the Future Knows It Yet.

In 2026, Eli Lilly has already closed or announced — just the public ones I can keep track of:

Verve — gene editing ($1B upfront)
Ventyx — inflammation ($1.2B upfront)
Orna — in vivo CAR-T ($2.4B)
Adverum — eye disease gene therapy
CrossBridge Bio — antibody-drug conjugates
Kelonia — in vivo CAR-T ($3.25B upfront / $7B total)
Ajax — next-gen JAK2 ($2.3B)

Two of those happened in April alone. The Mounjaro money is doing what Mounjaro money does — it’s buying the next ten years.

What’s encouraging — the coherence

Lilly isn’t acquiring randomly. It’s building a stacked bet on mechanism platforms — Type II JAK binding, in vivo gene placement, lentiviral T-cell engineering, base editing.

These aren’t drugs. They’re factories for drugs.

If even two of these platforms work, Lilly owns a pipeline-of-pipelines for fifteen years.

What worries me — the math

Almost none of these acquisitions come with proof-of-concept clinical data:

Ajax — Phase 1, first POC data still pending in 2026
Kelonia — KLN-1010 has 100% MRD-negative responses, in four patients
Orna — preclinical at acquisition

Lilly is paying mid-stage prices for early-stage assets, and justifying it with platform optionality. That math works beautifully on the way up — and breaks ugly on the way down.

The quiet lesson for digital health

Watch what Lilly isn’t buying. With every dollar that Lilly puts into a CAR-T platform, there’s a dollar it’s not putting into the digital infrastructure these therapies will need to actually reach patients at scale:

Patient identification
Diagnostics
Monitoring infrastructure
Outcomes-to-payer evidence pipelines

If in vivo therapies remove manufacturing constraints, the next constraint is: Who finds the patient? Who manages delivery? Who proves outcomes to payers?

The shift

From: → Pharma buys drugs.

To: → Pharma buys platforms (drug factories).

👉Lilly is buying the future. Someone still has to teach it how to find the patient. That’s not pharma — that’s digital health infrastructure.

📚 One Resource Worth Reading

FDA ODAC briefing documents — camizestrant (SERENA-6)

The clearest current view into how FDA is thinking about ctDNA-driven regulation. Free, public, and going to be cited for years regardless of how the vote lands.

Source: FDA.gov →

← Back to Newsletter ← Issue #3

Read Issue #4 on Beehiiv →