Weekly Tech+Bio Highlights #75: Lilly's Supercomputer, FDA's One-Trial Shift, and China Deals Get Pricey
Generate:Bio's $400M IPO, federated ADMET modeling across five pharma companies, Gilead's $7.8B CAR-T buyout, and an $80M AI-enabled brain health clinic network
The past week was notable on both policy and infrastructure fronts, with the FDA formalizing a one-pivotal-trial default and emphasizing mechanistic, real-world, and model-based confirmative evidence, and Eli Lilly bringing online its in-house AI supercomputer in Indianapolis to support large-scale biology and chemistry models.
A newly launched U.S. brain health clinic network also raised $80M to bring psychiatric and neurological care, FDA-regulated clinical trials, and an AI-based coordination platform under one roof, combining in-person services with research infrastructure, with one site named after the late Dr. Nolan R. Williams, known for co-developing the FDA-cleared neuromodulation system.
According to the latest report, China biopharma licensing is no longer “cheap”. Generate:Biomedicines went public, raising $400M in its IPO and debuting at a valuation above $1B, though shares traded down after listing. Also, IQVIA said it will buy selected Charles River discovery assets to add in-vitro capabilities and a machine-learning small-molecule platform to its offering.
Hi! This is BiopharmaTrend’s weekly newsletter, Where Tech Meets Bio, where we explore technologies, breakthroughs, and cutting-edge companies.
If this newsletter is in your inbox, it’s because you subscribed, or someone thought you might enjoy it. In either case, you can subscribe directly by clicking this button:
This newsletter reaches over 11K industry professionals from leading organizations across the globe. To reserve your sponsor slot in one of the upcoming issues, contact us at info@biopharmatrend.com
🤖 AI x Bio
(AI applications in drug discovery, biotech, and healthcare)
🔹 Pharma-scale AI compute — Eli Lilly brought online its in-house AI supercomputer in Indianapolis with 1,000+ GPUs delivering ~9,000 petaflops, built to train biology/chemistry/genomics foundation models and support drug discovery and clinical development using ~700 TB of genomics data.
🔹 Shared standards for AI in biology — the international BEACON consortium, coordinated by the open-science nonprofit Conscience, launched to unify benchmarking across AI-driven disease biology and small-molecule discovery via community challenges, shared datasets, and an open validation platform, with a public debut at a Barcelona symposium on March 18–19, 2026.
🔹 Community-built AMR datasets — The Align Foundation and Google DeepMind launched a global effort to define AI-ready antimicrobial-resistance datasets and evaluation benchmarks.
🔹 Privacy-preserving ADMET modeling — Lundbeck, Orion, Recursion, Servier, and another drugmaker put ~80% of their internal ADMET data into a federated-learning network run by Apheris to improve absorption/metabolism/toxicity prediction without sharing raw proprietary datasets, initially for small molecules.
🔹 Scaling tissue AI models — Bruker and Noetik expanded their partnership to generate larger spatial single-cell human tissue datasets (building on 3,500+ patients) to train foundation models for translational and therapeutic use.
🔹 AI spots early diabetes tissue signals — Researchers at the German Center for Diabetes Research trained explainable deep-learning models on gigapixel pancreas slides from living donors to distinguish type 2 diabetes and pinpoint subtle structural changes (islet cell patterns, nerve fibers, and fat–islet proximity) that may serve as biomarkers.
🔹 AI to speed clinical trials — Evinova signed clinical-development AI partnerships with Astellas and AstraZeneca (after BMS) to use shared operational data to optimize trial design and execution, claiming ~5–7% cost savings per study.
🔹 AI planning for brain drug delivery — Brainlab and Precision NeuroMed partnered to build an AI-enabled planning platform for direct brain infusion to standardize patient-specific delivery, starting with recurrent glioblastoma and expanding to other CNS diseases.
🔹 Bo Wang, SVP and Head of Biomedical AI at Xaira Therapeutics, argues that diffusion models may be better suited than autoregressive models for biological foundation models, because biology (proteins, gene expression, regulatory DNA) is inherently bidirectional and high-dimensional rather than sequential like language. Separately.
💰 Money Flows
(Funding rounds, IPOs, and M&A for startups and smaller companies)
🔹 China licensing prices surge — Evaluate data show average upfront payments in Western–China biopharma licensing jumped 230% from $52M (2022) to $172M (2026 YTD) as deal volume and total upfront value rose, signaling China assets are no longer “cheap.”
🔹 Big AI-protein IPO, rough debut — Generate Biomedicines priced 25M shares at $16 to raise $400M to fund AI-designed protein medicines, led by two Phase 3 severe-asthma trials (~1,600 patients), although the stock fell ~21% after trading, valuing it at ~$1.61B (February 27th).
🔹 AI + in-vitro scale-up IQVIA will buy selected Charles River discovery assets for ~$145M (plus up to $10M), adding five in-vitro sites and a machine-learning small-molecule platform backed by 20+ years of data linked to 100+ clinical-trial molecules; closing Q2 2026.
🔹 No-code AI modeling for bench scientists — Tamarind Bio raised $13.6M Series A to scale a cloud platform used by ~100 biotechs (including “8 of the top 20 pharmas”) that lets lab teams run 200+ AI and simulation models for protein and small-molecule research without command-line or infrastructure setup.
🔹 Digital-cell models expand into immunology — Turbine raised $25M Series B to scale its AI “virtual cell” platform beyond oncology, launching an immunology collaboration with an undisclosed top-10 pharma and building on prior work with AstraZeneca, after supporting 30+ partner discovery programs.
🔹 Funding for organ-targeted gene delivery — BreezeBio raised $60M Series B to advance non-viral, tissue-targeted nanoparticle delivery and push its first genetic-medicine programs toward the clinic, starting with IND-enabling work in type 1 diabetes.
🔹 Recursion reported a $4M progress payment from Sanofi (total $134M to date) and $754M cash extending runway to early 2028, while saying the joint immunology-oncology small-molecule portfolio could grow to up to 15 programs. Separately, recent filings show NVIDIA sold its Recursion shares by Dec. 31, 2025 while ARK bought more.
🔹 Quantum-chemistry drug discovery funding — Ten63 Therapeutics raised $45M led by Chugai Venture Fund and the Gates Foundation to advance a quantum-chemistry-based small-molecule discovery model aimed at “undruggable” oncology targets, including MYC, starting with cervical cancer.
🔹 Record ALS research push — U.S. Congress approved a record $315M for ALS research in 2026.
🔹 AI deal for genetic epilepsy targets — Angelini Pharma struck a multiyear partnership with Quiver Bioscience worth up to $120M in milestones (plus royalties) to use AI-driven genomics and safety modeling to identify new therapies for rare pediatric genetic epilepsies.
🔹 DMD gene therapy launch in Japan — Sarepta’s Duchenne gene therapy began commercial sales in Japan (via Chugai/Roche) for ambulatory children 3 to <8 years, triggering a $40M milestone payment to Sarepta.
🚜 Market Movers
(News from established pharma and tech giants)
🔹 Pfizer and gene editing — Pfizer exercised its option for global rights to Beam’s liver-targeted in vivo base-editing program (from a $300M upfront collaboration with up to $1.05B milestones), while Beam also disclosed a $500M Sixth Street loan extending runway into mid-2029.
🔹 Big CAR-T buyout — Gilead agreed to buy Arcellx for $7.8B ($115/share + $5 CVR) to take full control of a BCMA CAR-T for relapsed/refractory multiple myeloma under FDA review with a Dec. 23 decision date.
⚙️ Other Tech
(Innovations across quantum computing, BCIs, gene editing, and more)
🔹 Organoid-informed gene therapy enters clinic — Mahzi Therapeutics dosed the first patient in a Phase 1/2 trial of an AAV9 gene-replacement therapy for Pitt Hopkins syndrome (TCF4 mutations), enrolling ~12 participants across five sites in the U.S., Israel, and Spain.
🔹 Protein degradation via polymer “bridges” — Northwestern researchers created protein-like polymers that bind hard-to-drug cancer drivers and recruit the cell’s disposal machinery to degrade them, showing selective target removal and tumor-growth slowing in mouse models.
🔹 Single-molecule proteomics debuts — Nautilus unveiled its Voyager benchtop at US HUPO 2026 to map up to 10B intact proteins/proteoforms per run.
🔹 Rett gene therapy fast-tracked — Neurogene’s one-time Rett syndrome gene therapy received FDA Breakthrough Therapy status based on interim Phase 1/2 functional improvements, with registrational trial dosing expected to finish in Q2 2026.
🏛️ Bioeconomy & Society
(News on centers, regulatory updates, and broader biotech ecosystem developments)
🔹 FDA one-trial default draws caution — FDA made a single pivotal trial the default for approvals beyond oncology/rare disease, but experts warn the change may mostly shift discretion and increase sponsor risk without clearer guidance. See how this could affect AI-enabled evidence packages in our recent coverage “FDA Shifts Drug Approval Policy: What Does This Mean for AI-Enabled Therapies?”
🔹 AI pathology distribution in the EU — Epredia signed an EU distribution deal to add Mindpeak’s on-prem AI image-analysis modules to its digital pathology offering.
🚀 A New Kid on the Block
(Emerging startups with a focus on technology)
🔹 AI-enabled brain health clinics — Salma Health launched with $80M Series A (Mubadala Capital and ARCH) to build an integrated U.S. network combining psychiatry/neurology care, FDA-regulated trials, and an AI care-coordination platform, starting with clinics in San Diego, Orange County, Fremont, and Berkeley.
New AI-Native Biotechs: 2026 Watchlist
We’ve updated our AI-Native Biotechs Watchlist with fresh context on the FDA’s newly formalized “one pivotal trial” default and its explicit emphasis on mechanistic, real-world, and model-based “confirmative evidence” around that study. In the article’s framing, that shifts more weight onto things like validated biomarkers, external controls, real-world evidence, and computational work such as modeling, in-silico simulations, and synthetic comparators.
We scan a batch of newly launched startups built around those computational capabilities, spanning areas like RNA design, biomolecular modeling, regulatory DNA engineering, data imputation, lab automation, and structure-based drug discovery, with company-by-company notes on what each is building and what it reports so far.
Why AI Won’t Automatically Accelerate Clinical Trials
Asimov Press just published a response to Dario Amodei’s “trials will take one year” speculation, arguing that better drug design does not automatically compress clinical trial timelines (Ruxandra Teslo with Asimov Press, Feb. 27, 2026). The piece separates two variables that often get blurred: whether a candidate works (trial success rate) versus how long it takes to run the experiment once it starts (trial speed).
It spends most of its time on why speed is bounded by things like recruitment, site operations, shipping logistics, regulator requirements, and biology itself, including the time it takes to observe endpoints.
The most concrete illustration is osteoporosis: despite a strong preclinical model (the ovariectomized rat) and high Phase III success reported as 83.7%, the author argues the field still struggles because Phase III trials tend to be enormous (10,000–16,000 participants), long (three to five years), and costly ($500M to $1B), largely due to fracture reduction being a slow, low-frequency endpoint.
The article also points to a 2015 oncology R&D analysis linking surrogate endpoint acceptance to investment patterns, and it frames better surrogate endpoints and regulatory process changes as the real levers, with AI helping “at the margins” via drafting, site selection, matching, and data workflows. It proposes a CTN-like approach (Australia’s Clinical Trial Notification framework) as an example of how early-phase work might move faster than the U.S. IND default, while still keeping safety oversight.
Read also:
Three Big Ideas in Aging Research That Could Shift the Therapeutic Landscape