Verified
Eli Lilly
Experimenting- Pharma
- ~50,000 employees
- No AI return or user count
- LillyPod: 1,016 NVIDIA GPUs on-prem
- LLMs available company-wide
Eli Lilly deploys AI on LillyPod, an NVIDIA supercomputer with 1,016 GPUs that it owns and runs in its own facilities, where it trains drug discovery models and serves open-weight language models that Chief AI Officer Thomas Fuchs said in Jun 2026 are available across the company. Eli Lilly shares discovery models with more than 100 biotechs through TuneLab. Eli Lilly has reported no business return it credits to AI and has not said how many employees use AI.
lilly.comTheir approach to AI
- The platform
- LillyPod, an NVIDIA DGX SuperPOD with 1,016 Blackwell Ultra GPUs that Eli Lilly owns and operates, went live at a ribbon-cutting in Indianapolis on Feb 25, 2026, and TuneLab has shared Lilly's discovery models with outside biotechs since Sep 2025.
- The models
- Eli Lilly trains its own discovery models on LillyPod, runs open-weight language models there, including NVIDIA's Nemotron 3 Ultra (Jun 2026), and adds partner models such as Chai Discovery's miniprotein design suite on TuneLab (Jun 2026).
- The reach
- Thomas Fuchs said in Jun 2026 that LillyPod's language models are available across Eli Lilly with no token or budget limits, and Daniel Skovronsky said in May 2026 that every chemist uses AI; Eli Lilly has not said how many of its roughly 50,000 employees use AI.
- The catch
- Eli Lilly reports no business return it credits to AI and no count of employees using it, and Daniel Skovronsky said in May 2026 that AI will not shorten the years patients spend in clinical trials.
Snapshot
- Employees
- ~50,000
- End of 2025, up from about 47,000 a year earlier; about 12,000 in R&D
- Technology budget
- Not disclosed
- R&D expense was $13.3B in 2025, up 21% (2025 Form 10-K)
- AI/ML headcount
- Not disclosed
- Platform
- LillyPod and TuneLab
- On-premises NVIDIA DGX SuperPOD (live Feb 2026); federated TuneLab (Sep 2025)
- Models
- Proprietary, open-weight and partners'
- Open-weight LLMs incl. NVIDIA Nemotron 3 Ultra (Jun 2026); Chai Discovery, Insilico Medicine, Isomorphic Labs, OpenAI
- Reach
- Company-wide access
- LillyPod's language models available across Lilly (Jun 2026); every chemist uses AI (May 2026); no user count
- Where AI pays first
- Back office, then manufacturing
- R&D takes longest, CFO Lucas Montarce said (Mar 2026)
Key findings
Eli Lilly has reported no business return it credits to AI and has not said how many of its roughly 50,000 employees use AI; Lucas Montarce said in Sep 2025 that headcount outside manufacturing stayed relatively flat as revenue grew, crediting AI, automation and finance discipline together, while total headcount rose from about 47,000 to about 50,000 in 2025.
Eli Lilly built LillyPod, an NVIDIA DGX SuperPOD with 1,016 Blackwell Ultra GPUs that it owns and operates, and inaugurated it in Indianapolis on Feb 25, 2026; Chief AI Officer Thomas Fuchs explained in Aug 2026 why Lilly chose to own its compute instead of relying only on the cloud.
Thomas Fuchs said in Jun 2026 that LillyPod runs open-weight language models, including NVIDIA's Nemotron 3 Ultra, and makes them available across Eli Lilly with no token or budget limits; a podcast's summary of a Sep 2025 interview with Fuchs put Lilly's AI projects at more than 1,000, and Lilly has not said how many run in production.
Eli Lilly and NVIDIA said on Jan 12, 2026 that they will jointly invest up to $1 billion over five years in an AI co-innovation lab in the San Francisco Bay Area, where agentic wet labs feed computational dry labs in one learning loop.
Eli Lilly launched TuneLab in Sep 2025 to give biotechs federated access to discovery models trained on data Lilly says cost more than $1 billion to obtain; TuneLab passed 100 members in Jun 2026, added its first outside models, from Chai Discovery, and is offered inside Benchling and Schrödinger's LiveDesign (Jan 2026).
Daniel Skovronsky said on May 28, 2026 that Eli Lilly uses AI in just about every step of drug development and that every chemist uses AI as a tool, while Lucas Montarce said on Mar 2, 2026 that AI pays first in the back office, closing the books and forecasting.
Where Lilly sits
- Stage
- Experimenting
Eli Lilly has reported no business return it credits to AI, for any use case or for the company as a whole: Lucas Montarce said in Sep 2025 that headcount outside manufacturing stayed relatively flat as revenue grew, crediting AI and automation together with finance discipline and giving no figure, and Eli Lilly's 2025 Form 10-K says there can be no assurance that its significant investments in AI will benefit its efficiency or profitability. Eli Lilly has also not said how many of its roughly 50,000 employees use AI: Thomas Fuchs, Chief AI Officer since Oct 2024, said in Jun 2026 that open-weight language models on LillyPod are available across the company with no token or budget limits, and Daniel Skovronsky said in May 2026 that every chemist uses AI, but neither gave a user count. Eli Lilly therefore stays at Experimenting, short of Piloting, which needs one AI-credited return or an officially reported rollout with a user count and a named owner.
How the stages are defined - Ownership
- Chief AI Officer Thomas Fuchs, alongside EVP and CIDO Diogo Rau on the Executive Committee
- Platform
- Hybrid: LillyPod and TuneLab built in-house on NVIDIA hardware; partner models bought
- Models
- Proprietary discovery models, open-weight LLMs such as NVIDIA Nemotron, partner models
- Hosting
- LillyPod on-premises in Lilly facilities, inaugurated in Indianapolis; TuneLab on an unnamed third-party host; no cloud named
- Reach
- LLMs available company-wide (Jun 2026); no user count
How they deploy
Own the compute
Eli Lilly built LillyPod as an on-premises supercomputer, which Lilly calls the most powerful owned and operated by a pharmaceutical company, to run its proprietary drug discovery models, starting with chemistry, and by Jun 2026 to serve open-weight language models across the company. David Ricks said in a podcast recorded on the day of the Oct 2025 announcement that the build was constrained only by power. Thomas Fuchs cast the choice in Jun 2026 as “controlling our own destiny” and argued in an Aug 2026 WWT interview that owning compute at scale changes what is possible for discovery, development and delivery.
Proprietary data as the advantage
Daniel Skovronsky told investors on May 28, 2026 that models trained mostly on language and the internet perform poorly in chemistry and biology, and that the data to train better ones sits mostly inside a few large companies such as Eli Lilly. Skovronsky said Eli Lilly is growing that data quickly through its own experiments, through partners who trade their data for access to Lilly's models, and through data factories it is building with NVIDIA.
Share models with biotechs
Eli Lilly offers its discovery models to outside biotechs through TuneLab, part of the Catalyze360 program that also gives partners capital, lab space and development help. TuneLab uses federated learning, built on NVIDIA FLARE, so that neither Lilly's data nor a partner's is exposed, and partners contribute training data in return for access. Eli Lilly agreed in Jan 2026 to offer TuneLab inside Benchling and Schrödinger's LiveDesign, and in Jun 2026 Charles River joined to provide nonclinical testing and Axcelead DDP joined as the first contract research organization to contribute its own data. David Ricks said in Jan 2026 that 180 biotechs were in Catalyze360, and compared TuneLab to an App Store for scientists in a Jun 2026 Financial Times interview.
Partner for specialist models
Eli Lilly pairs its own models with AI biotechs and model builders: Isomorphic Labs (Jan 2024), OpenAI for antimicrobials (Jun 2024), Chai Discovery for biologics design (Jan 2026), Insilico Medicine for oral small molecules (Mar 2026) and Profluent for AI-designed recombinases (Apr 2026). Chai is building a model trained on Lilly data for Lilly's use alone.
Back office first, R&D last
Lucas Montarce said on Mar 2, 2026 that every line of Eli Lilly's income statement will benefit from AI over time, with selling, general and administrative costs first, starting in the back office, manufacturing in the middle term, and R&D taking the longest. Daniel Skovronsky said in May 2026 that AI will make clinical development smoother but cannot shorten the years needed to enroll and follow patients.
- Deployment model
- Eli Lilly splits AI ownership across a few senior roles. Diogo Rau, Executive Vice President and Chief Information and Digital Officer and an Executive Committee member, leads information technology, cybersecurity, digital health, and advanced analytics and data science, according to the 2025 Form 10-K. Thomas Fuchs, Senior Vice President and Lilly's first Chief AI Officer since Oct 2024, sets AI strategy across drug discovery, clinical trials, manufacturing, commercial work and internal functions and works alongside Rau; Lilly has not published Fuchs's reporting line. Discovery science sits in Lilly Research Laboratories under Daniel Skovronsky, and TuneLab sits in Catalyze360, the biotech partnering program led by Nisha Nanda, with Aliza Apple heading TuneLab.
- Model providers
- Eli Lilly trains its own foundation models on LillyPod from chemistry, biology, clinical and manufacturing data (Jun 2026), which NVIDIA said would include protein diffusion, small-molecule graph neural network and genomics models (Feb 2026), and publishes some discovery models, built on drug disposition, safety and preclinical data from hundreds of thousands of molecules, on TuneLab. Thomas Fuchs said in Jun 2026 that LillyPod also runs open-weight language models, including NVIDIA's Nemotron 3 Ultra of about 550 billion parameters, and makes them available across the company. TuneLab added Chai Discovery's miniprotein design suite in Jun 2026, its first outside models; Lilly said in Oct 2025 that TuneLab would add workflows using NVIDIA Clara models, and NVIDIA said in Feb 2026 that TuneLab plans to offer NVIDIA BioNeMo models. Chai is also building a custom model on Lilly data for Lilly alone (Jan 2026), and Lilly works with Insilico Medicine, Isomorphic Labs, Profluent and OpenAI on discovery. David Ricks said in a podcast published Nov 2025 that he uses Claude or xAI's model for science questions during meetings.
- Data and infrastructure
- Eli Lilly's LillyPod is the first NVIDIA DGX SuperPOD built with DGX B300 systems, with 1,016 Blackwell Ultra GPUs delivering more than 9,000 petaflops of AI performance; NVIDIA said it was assembled in four months and inaugurated in Indianapolis on Feb 25, 2026. LillyPod sits within existing Lilly facilities, is cooled with Lilly's existing chilled water system and is run with NVIDIA Mission Control software on NVIDIA Spectrum-X networking; Lilly said in Oct 2025 that LillyPod would run on 100% renewable electricity, and NVIDIA said in Feb 2026 that Lilly aims to reach that by 2030. TuneLab is hosted by a third party that Lilly has not named and runs its federated learning on NVIDIA FLARE. Eli Lilly's 2025 Form 10-K says it relies partly on third-party cloud and data hosting providers without naming them, and the co-innovation lab with NVIDIA is to start work in South San Francisco.
Platforms and internal tools
LillyPodAnnounced Oct 2025; live Feb 25, 2026
LillyPod is Eli Lilly's on-premises AI supercomputer, an NVIDIA DGX SuperPOD with DGX B300 systems and 1,016 Blackwell Ultra GPUs. NVIDIA said at launch that LillyPod would train protein diffusion, small-molecule and genomics foundation models, that Lilly's genomics team works with 700 terabytes of data on it, and that Lilly's internal AI platforms on it let employees build chatbots, agentic workflows and research lab agents (Feb 2026). Thomas Fuchs said in Jun 2026 that LillyPod serves open-weight language models across the company with no token or budget limits.
- Users
- Language models available across the company (Jun 2026)
- Models
- Proprietary Lilly models; open-weight LLMs including NVIDIA Nemotron 3 Ultra
Lilly TuneLabSep 2025
Lilly TuneLab is a federated AI and machine learning platform that gives selected biotechs access to discovery models trained on Eli Lilly's drug disposition, safety and preclinical data, in exchange for training data of their own. TuneLab is hosted by a third party, runs its federated learning on NVIDIA FLARE, belongs to the Catalyze360 program, is offered inside Benchling and Schrödinger's LiveDesign (Jan 2026), and added Chai Discovery's models in Jun 2026, its first from outside Lilly.
- Users
- 100+ biotech members (Jun 2026)
- Models
- Lilly models, Chai Discovery miniprotein design; NVIDIA Clara and BioNeMo models planned
NVIDIA co-innovation labAnnounced Jan 12, 2026
The co-innovation lab is a San Francisco Bay Area site where Eli Lilly scientists and NVIDIA AI engineers are to build biology and chemistry models on NVIDIA BioNeMo in a scientist-in-the-loop setup linking agentic wet labs to computational dry labs, with up to $1 billion of joint investment over five years. Lilly said work was expected to begin on Mar 31, 2026, and the Jan 2026 announcement placed the start in South San Francisco.
Vendors and partners
NVIDIAOct 2025
LillyPod hardware, co-innovation lab partner, Clara and BioNeMo models planned for TuneLab · NVIDIA supplied the DGX SuperPOD behind LillyPod (Oct 2025) and is co-investing in the Bay Area lab announced Jan 12, 2026, where NVIDIA and Lilly said they will also explore AI in clinical development, manufacturing and commercial operations. Thomas Fuchs said in Jun 2026 that Lilly was the first in its industry to run NVIDIA's Nemotron 3 Ultra model on-premises.
Chai DiscoveryJan 2026
Biologics design platform and a custom model trained on Lilly data; models offered on TuneLab · Chai's Jan 9, 2026 agreement followed a period in which Lilly evaluated Chai's model designs; a second agreement in Jun 2026 put Chai's miniprotein design suite on TuneLab.
Insilico MedicineMar 2026 (expanded)
Pharma.AI platform for oral drug discovery · Insilico Medicine announced the expanded deal on Mar 29, 2026; it is worth up to $2.75 billion, including $115 million upfront, and builds on a licensing collaboration that began in 2023 (Pharmaceutical Technology, Mar 30, 2026).
ProfluentApr 2026
AI-designed recombinases for kilobase-scale DNA editing in genetic medicine · Bloomberg reported on Apr 28, 2026 that the Profluent deal with Lilly is worth up to $2.25 billion.
Isomorphic LabsJan 2024
Multi-target AI drug discovery research · Isomorphic Labs received $45 million upfront from Lilly; its Lilly and Novartis deals together could reach nearly $3 billion.
OpenAIJun 2024
Generative AI to invent antimicrobials for drug-resistant bacteria
Genetic LeapSep 2024
AI engine for RNA-targeted medicines · Fierce Biotech reported in Sep 2024 that the Genetic Leap deal with Lilly could be worth up to $409 million.
Benchling and SchrödingerJan 2026
Offer TuneLab inside their R&D software, Benchling's cloud platform and Schrödinger's LiveDesign · Both agreements were announced in Jan 2026; BiopharmaTrend reported that full TuneLab access inside Benchling would roll out later in 2026.
Charles River and Axcelead DDPJun 2026
Contract research partners on TuneLab: nonclinical testing (Charles River) and discovery data and model use for clients (Axcelead DDP)
Yseop
Natural language generation for patient narratives in clinical study reports · Emerj reported Lilly's use of Yseop in Aug 2024, citing a Yseop case study; Lilly has not confirmed current use.
Leadership
AI, data and platform leaders
Executives
AI org map
Chief AI Officer beside a CIDO on the Executive Committee; discovery AI in the research labs
Eli Lilly named Thomas Fuchs its first Chief AI Officer in Oct 2024 with a remit across discovery, clinical trials, manufacturing, commercial work and internal functions, working alongside Diogo Rau, the Executive Vice President and Chief Information and Digital Officer who sits on the Executive Committee. Rau's organization covers information technology, cybersecurity, digital health, and advanced analytics and data science. Daniel Skovronsky's Lilly Research Laboratories, which reports to David Ricks, runs the discovery science that LillyPod and TuneLab serve.
Eli Lilly has not published how many AI and machine learning specialists it employs. Eli Lilly's AI work sits in three places: Diogo Rau's information and digital organization, which includes advanced analytics and data science; Thomas Fuchs's Chief AI Office; and Lilly Research Laboratories, which employs much of Lilly's roughly 12,000 R&D staff and where every chemist uses AI tools, according to Daniel Skovronsky (May 2026). A podcast's summary of a Sep 2025 interview with Thomas Fuchs said Eli Lilly runs AI certification programs for its technology employees. The NVIDIA co-innovation lab adds a Bay Area team of Lilly scientists and NVIDIA engineers.
Reporting lines
Chair and CEO
Information and Digital
Diogo Rau, EVP and CIDO
Executive Committee member; the line to the CEO is inferred from Executive Committee membership, not published
Advanced Analytics and Data Science
Leader not named publicly (2025 Form 10-K)
Cybersecurity
Chief Information Security Officer
The CISO reports directly to the CIDO (2025 Form 10-K)
Chief AI Office
Thomas Fuchs, SVP and Chief AI Officer
Reporting line not published; works alongside the CIDO
Lilly Catalyze360
Nisha Nanda
Reporting line not published
Lilly TuneLab
Aliza Apple
TuneLab is a Catalyze360 offering; the head's own reporting line is not published
Eli Lilly's 2025 Form 10-K lists Diogo Rau among its executive officers but not Thomas Fuchs, and Lilly has not said whether Fuchs reports to Rau, to Daniel Skovronsky or to David Ricks; coverage of his appointment said only that he would work alongside Rau. Tim Coleman, senior vice president and chief technology officer, who led the team that deployed LillyPod according to Kellogg Magazine, retired after 35 years at Lilly, Endpoints News reported on Apr 2, 2026; the reports named no successor.
Where the AI workforce sits
Research and development
About 12,000 of Eli Lilly's roughly 50,000 employees work in R&D (end of 2025); Daniel Skovronsky said every chemist uses AI tools (May 2026).
Not disclosedInformation, digital, analytics and data science
Diogo Rau's organization; Eli Lilly gives no headcount for it.
Not disclosedChief AI Office and the NVIDIA co-innovation lab
Thomas Fuchs's team and the Bay Area lab staffed by Lilly scientists and NVIDIA engineers.
Not disclosed
Named people
Use cases
Eli Lilly's published AI use cases center on drug discovery: AI design tools used by every chemist, proprietary foundation models trained on LillyPod, the TuneLab platform for biotechs, and partnerships with NVIDIA, Chai Discovery, Insilico Medicine and Profluent. Eli Lilly also makes open-weight language models on LillyPod available across the company, and uses AI in finance, manufacturing quality control, regulatory writing and marketing; a podcast's summary of a Sep 2025 interview with Thomas Fuchs said Lilly runs more than 1,000 AI projects, and Lilly has not said how many run in production. Eli Lilly has not reported a business return for any AI use case and has published no firm-level AI return; CFO Lucas Montarce credited AI and automation, along with finance discipline, for keeping headcount outside manufacturing roughly flat as revenue grew (Sep 2025).
| Use case | Area | Status | ROI |
|---|---|---|---|
| AI-assisted molecule and protein design Every Eli Lilly chemist uses AI as a tool that suggests which structures to make, and protein engineers use it to engineer antibodies and proteins, Daniel Skovronsky said in May 2026. | Drug discovery | In production | |
| Proprietary foundation models on LillyPod Eli Lilly trains its own foundation models on LillyPod from chemistry, biology, clinical and manufacturing data, and Thomas Fuchs said one such model helped co-design a small-molecule candidate whose new chemical fragment Lilly chemists now reuse on other targets (Jun 2026). | Drug discovery | In production | |
| Open-weight language models and agents on LillyPod Eli Lilly runs open-weight language models, including NVIDIA's Nemotron 3 Ultra, on LillyPod and makes them available across the company with no token or budget limits (Jun 2026); Lilly's internal AI platforms on LillyPod let employees build chatbots, agentic workflows and research lab agents (Feb 2026). | Enterprise | In production | |
| Lilly TuneLab for outside biotechs TuneLab gives more than 100 biotech members federated access to Eli Lilly's drug disposition, safety and preclinical models in exchange for their own training data. | Drug discovery | In production | |
| Biologics design with Chai Discovery Eli Lilly is deploying Chai Discovery's platform to design biologics for multiple targets, and Chai is building a model trained on Lilly data for Lilly's use alone. | Drug discovery | Pilot | |
| Generative chemistry with Insilico Medicine Eli Lilly expanded its work with Insilico Medicine in Mar 2026 to use Insilico's Pharma.AI platform for oral drug discovery across several disease areas. | Drug discovery | Announced | |
| AI-designed recombinases with Profluent Eli Lilly and Profluent are developing AI-designed recombinases, enzymes for kilobase-scale DNA editing, for genetic medicines across several diseases. | Genetic medicine | Announced | |
| Closed-loop discovery at the NVIDIA co-innovation lab Eli Lilly and NVIDIA plan a Bay Area lab whose first project links agentic wet labs with computational dry labs in a continuous learning system, with work expected to begin in South San Francisco early in 2026. | Drug discovery | Announced | |
| Antimicrobials with OpenAI Eli Lilly uses OpenAI's generative AI to invent new antimicrobials against drug-resistant pathogens under a collaboration announced in Jun 2024. | Drug discovery | Announced | |
| Patient narratives for regulatory submissions Eli Lilly drafts patient narratives for clinical study reports with Yseop's natural language generation software, Emerj reported in Aug 2024 from a Yseop case study. | Clinical development | In production | |
| Book closing, process automation and forecasting Eli Lilly's finance organization uses AI to close the books and records, automate processes and forecast, Lucas Montarce said in Mar 2026. | Finance | In production | |
| AI and automation on new production lines Eli Lilly uses automation and AI to make the lines at its new manufacturing sites as efficient as possible (Mar 2026), and David Ricks said in May 2026 that Lilly is looking at AI to automate unsafe or repetitive manufacturing steps. | Manufacturing | In production | |
| Computer vision for quality control Eli Lilly uses computer vision for quality control, according to a podcast's summary of a Sep 2025 interview with Thomas Fuchs, which lists it among more than 1,000 AI projects spanning discovery, development, manufacturing and commercial work. | Manufacturing | In production | |
| Digital twins and robotics Eli Lilly plans to use LillyPod with NVIDIA robotics technology to build digital twins of manufacturing processes and supply chains and to monitor production in real time. | Manufacturing | Announced | |
| The Memory Radio Station Eli Lilly's travelling Alzheimer's awareness campaign uses generative AI to turn visitors' memories into images they review and approve before the images join a public archive. | Medical and commercial (Europe) | In production | |
| AI-powered career coaching Eli Lilly offers employees and leaders AI-powered coaching alongside human coaches through its Explore Your Career program. | People | In production |
Adoption
Eli Lilly has not published how many of its roughly 50,000 employees use AI tools. Thomas Fuchs said in Jun 2026 that open-weight language models running on LillyPod are available across the company with no token or budget limits, and a podcast's summary of a Sep 2025 interview with Fuchs said Lilly runs more than 1,000 AI projects, certifies its technology employees in AI and offers AI education to everyone at Lilly; Lilly has not said how many of those projects run in production. Eli Lilly's executives describe use by function: Daniel Skovronsky said in May 2026 that every chemist uses AI and that Lilly uses AI in just about every step of drug development, Lucas Montarce said in Mar 2026 that finance uses AI to close the books and forecast, David Ricks said in a podcast published Nov 2025 that he runs one or two AI models in every meeting, and Lilly's workplace page lists AI-powered coaching for employees. Outside the company, TuneLab passed 100 biotech members in Jun 2026, and Catalyze360 had 180 biotechs in Jan 2026.
By the numbers
Rollout, first pilot to today
Jan 2024
Isomorphic Labs announced multi-target AI drug discovery collaborations with Eli Lilly and Novartis, with $45 million upfront from Lilly.
Jun 2024
Eli Lilly began a collaboration with OpenAI to invent antimicrobials for drug-resistant bacteria with generative AI.
Oct 2024
Eli Lilly named Thomas Fuchs its first Chief AI Officer, starting Oct 21, 2024, with a remit across discovery, trials, manufacturing, commercial work and internal functions.
Sep 2025
Eli Lilly launched TuneLab, giving biotechs federated access to models trained on data Lilly says cost more than $1 billion to obtain.
Sep 2025
A podcast's summary of an interview with Thomas Fuchs said Eli Lilly was running more than 1,000 AI projects across discovery, development, manufacturing and commercial work, with AI certification for its technology employees.
Oct 2025
Eli Lilly announced the pharmaceutical industry's most powerful supercomputer, an NVIDIA DGX SuperPOD with more than 1,000 B300 GPUs, to serve as an AI factory.
Jan 2026
Eli Lilly signed a biologics design collaboration with Chai Discovery and announced the NVIDIA co-innovation lab, backed by up to $1 billion over five years.
Feb 2026
Eli Lilly inaugurated LillyPod in Indianapolis, and NVIDIA said Lilly employees can use internal AI platforms on it to build chatbots, agentic workflows and research lab agents.
Mar 2026
Lucas Montarce said LillyPod was up and running in Eli Lilly's offices and that finance was using AI to close the books and forecast.
May 2026
Daniel Skovronsky said Eli Lilly uses AI in just about every step of drug development and every chemist uses AI as a tool.
Jun 2026
TuneLab passed 100 members and added its first models from outside Eli Lilly, Chai Discovery's miniprotein design suite.
Jun 2026
Thomas Fuchs said LillyPod was serving open-weight language models across Eli Lilly with no token or budget limits.
Economics & controls
Spend and ROI
Spending on AI
Eli Lilly does not disclose a technology budget or an AI budget, or what LillyPod cost. The co-innovation lab carries up to $1 billion of joint Lilly and NVIDIA investment over five years, and Lilly's AI partnership headlines include up to $2.75 billion with Insilico Medicine ($115 million upfront, Mar 2026) and up to $2.25 billion with Profluent (Apr 2026), mostly milestone payments. David Ricks listed supercomputers from NVIDIA among Lilly's growing R&D costs in a podcast published Nov 2025; R&D expense was $13.3 billion in 2025, up 21%, and $11.6 billion in the first nine months of 2026.
Reported ROI
Eli Lilly has not published a dollar return from AI for any use case or for the company. Eli Lilly's 2025 Form 10-K warns that there can be no assurance its use of AI or its significant investments in AI will improve efficiency or profitability. Figures that circulate in secondary sources, such as 1.4 million hours saved (Emerj, Aug 2024) and a vendor's claims for patient narrative automation, have no primary Lilly source.
Headcount
Eli Lilly employed about 50,000 people at the end of 2025, up from about 47,000 a year earlier, while revenue rose 45% to $65.2 billion. Lucas Montarce said in Sep 2025 that headcount outside manufacturing stayed relatively flat because of a discipline to use AI and automation and strong finance control, and David Ricks said in a podcast published Nov 2025 that his goal is to keep headcount flat outside manufacturing, with R&D headcount growing in single digits.
Where AI pays
Lucas Montarce told investors on Mar 2, 2026 that AI will benefit every line of Eli Lilly's income statement over time: selling, general and administrative costs first, starting with back-office work such as closing the books and forecasting, then manufacturing, then R&D, which will take the most time. Daniel Skovronsky said in May 2026 that AI may change the economics of clinical development and who gets paid for its tasks, but not how long trials take.
Governance and risk
Board oversight
Eli Lilly's board receives regular updates from management on AI strategy, risks, development, governance and ethics, and reviews innovation, including AI, as a strategic focus area, according to the 2026 proxy statement. The board's Science and Technology Committee monitors AI developments in pharmaceutical R&D, and the Audit Committee oversees cybersecurity and data protection.
Disclosed risks
Eli Lilly's 2025 Form 10-K warns that AI output may be inadequate or biased, that some AI model providers are young companies with less contractual and compliance maturity, that employees' unmanaged use of AI could put confidential information and intellectual property at risk, and that laws such as the EU AI Act may limit its use of AI. Eli Lilly also lists data leakage, healthcare fraud and abuse, and intellectual property infringement among possible unintended consequences.
Data boundary
Eli Lilly keeps its most valuable discovery data in-house: LillyPod runs on-premises in Lilly facilities, including the open-weight language models staff can use, and TuneLab uses federated learning on a third-party host so that partners use Lilly's models without seeing Lilly's data and Lilly does not see theirs. Chai Discovery's custom model trained on Lilly data is for Lilly's use alone. Thomas Fuchs argued in Jun 2026 that no private company should set the limits on the AI research drugmakers can pursue.
AI training
Eli Lilly runs AI certification programs for its technology employees and makes AI education resources available to everyone at the company, according to a podcast's summary of a Sep 2025 interview with Thomas Fuchs.
Security reporting
Eli Lilly's Chief Information Security Officer reports directly to Chief Information and Digital Officer Diogo Rau, who updates the Executive Committee on cybersecurity, according to the 2025 Form 10-K.
Consumer-facing generative AI
Eli Lilly's Memory Radio Station campaign has participants review and approve each AI-generated image before it is published, handles personal data under the GDPR, and states that the data is not used for clinical research, commercial profiling or product promotion (Jun 2026).
Recent signals
Aug 25, 2026
Chief AI Officer Thomas Fuchs explained in a WWT interview why Eli Lilly built LillyPod as an on-premises supercomputer instead of relying only on the cloud.
Jun 25, 2026
R&D World reported that Thomas Fuchs said LillyPod runs open-weight language models, including NVIDIA's Nemotron 3 Ultra, and makes them available across Eli Lilly with no token or budget limits.
Jun 18, 2026
Endpoints News reported that TuneLab passed 100 members and added Chai Discovery's models, the first outside models on Eli Lilly's platform.
May 28, 2026
Daniel Skovronsky said Eli Lilly uses AI in just about every step of drug development, that every chemist uses AI, and that Lilly's proprietary data is its advantage in training models.
May 12, 2026
David Ricks told employees at Eli Lilly's 150th anniversary that AI is speeding up drug design and research, and told reporters Lilly is looking at AI to automate unsafe or repetitive manufacturing steps.
Apr 28, 2026
Eli Lilly and Profluent agreed to develop AI-designed recombinases for genetic medicine, a deal Bloomberg valued at up to $2.25 billion.
Apr 2, 2026
Endpoints News reported that Tim Coleman had retired as Eli Lilly's chief technology officer after 35 years at the company, weeks after speaking for Lilly at LillyPod's launch.
Mar 29, 2026
Insilico Medicine announced an expanded deal with Eli Lilly, worth up to $2.75 billion with $115 million upfront, to use its Pharma.AI platform for oral drug discovery.
Mar 20, 2026
Eli Lilly's 2026 proxy statement said the board receives regular updates on AI strategy, risks and governance and credited David Ricks with driving the NVIDIA supercomputer partnership.
Mar 2, 2026
Lucas Montarce said LillyPod was up and running and that AI pays first in Eli Lilly's back office, then manufacturing, then R&D.
Feb 25, 2026
Eli Lilly inaugurated LillyPod, its NVIDIA DGX SuperPOD with 1,016 Blackwell Ultra GPUs and more than 9,000 petaflops of AI performance, at a ribbon-cutting in Indianapolis.
Jan 12, 2026
Eli Lilly and NVIDIA announced a co-innovation AI lab backed by up to $1 billion over five years, and David Ricks discussed it with NVIDIA's chief executive at the J.P. Morgan Healthcare Conference.
Jan 9, 2026
Schrödinger said it would offer Eli Lilly's TuneLab inside its LiveDesign software, after Benchling announced a similar TuneLab integration.
Jan 9, 2026
Chai Discovery announced that Eli Lilly will deploy its AI platform to design biologics and that Chai will build a model trained on Lilly data for Lilly alone.
Nov 11, 2025
David Ricks said on the Cheeky Pint podcast that LillyPod is on-premises to run Eli Lilly's proprietary discovery models and that he runs one or two AI models in every meeting.
Quick answers
- How does Eli Lilly deploy AI?
- Eli Lilly deploys AI on LillyPod, an NVIDIA supercomputer with 1,016 GPUs that it owns and runs in its own facilities, where it trains drug discovery models and serves open-weight language models that Chief AI Officer Thomas Fuchs said in Jun 2026 are available across the company. Eli Lilly shares discovery models with more than 100 biotechs through TuneLab. Eli Lilly has reported no business return it credits to AI and has not said how many employees use AI.
- What does Eli Lilly use AI for?
- Eli Lilly's published AI use cases center on drug discovery: AI design tools used by every chemist, proprietary foundation models trained on LillyPod, the TuneLab platform for biotechs, and partnerships with NVIDIA, Chai Discovery, Insilico Medicine and Profluent. Eli Lilly also makes open-weight language models on LillyPod available across the company, and uses AI in finance, manufacturing quality control, regulatory writing and marketing; a podcast's summary of a Sep 2025 interview with Thomas Fuchs said Lilly runs more than 1,000 AI projects, and Lilly has not said how many run in production. Eli Lilly has not reported a business return for any AI use case and has published no firm-level AI return; CFO Lucas Montarce credited AI and automation, along with finance discipline, for keeping headcount outside manufacturing roughly flat as revenue grew (Sep 2025).
- Who leads AI at Eli Lilly?
- AI at Eli Lilly is led by Diogo Rau (EVP and Chief Information and Digital Officer); Thomas Fuchs (SVP and Chief AI Officer); Nisha Nanda (Group Vice President and Head, Lilly Catalyze360). Ownership of AI on the org chart: Chief AI Officer Thomas Fuchs, alongside EVP and CIDO Diogo Rau on the Executive Committee.
- What AI platform does Eli Lilly use?
- LillyPod, an NVIDIA DGX SuperPOD with 1,016 Blackwell Ultra GPUs that Eli Lilly owns and operates, went live at a ribbon-cutting in Indianapolis on Feb 25, 2026, and TuneLab has shared Lilly's discovery models with outside biotechs since Sep 2025.
- Which AI models does Eli Lilly use?
- Eli Lilly trains its own discovery models on LillyPod, runs open-weight language models there, including NVIDIA's Nemotron 3 Ultra (Jun 2026), and adds partner models such as Chai Discovery's miniprotein design suite on TuneLab (Jun 2026).
- How many Lilly employees use AI?
- Thomas Fuchs said in Jun 2026 that LillyPod's language models are available across Eli Lilly with no token or budget limits, and Daniel Skovronsky said in May 2026 that every chemist uses AI; Eli Lilly has not said how many of its roughly 50,000 employees use AI.
- How far has AI spread at Eli Lilly?
- Eli Lilly has reported no business return it credits to AI, for any use case or for the company as a whole: Lucas Montarce said in Sep 2025 that headcount outside manufacturing stayed relatively flat as revenue grew, crediting AI and automation together with finance discipline and giving no figure, and Eli Lilly's 2025 Form 10-K says there can be no assurance that its significant investments in AI will benefit its efficiency or profitability. Eli Lilly has also not said how many of its roughly 50,000 employees use AI: Thomas Fuchs, Chief AI Officer since Oct 2024, said in Jun 2026 that open-weight language models on LillyPod are available across the company with no token or budget limits, and Daniel Skovronsky said in May 2026 that every chemist uses AI, but neither gave a user count. Eli Lilly therefore stays at Experimenting, short of Piloting, which needs one AI-credited return or an officially reported rollout with a user count and a named owner.
Open questions
- What business return, in cost, headcount or revenue, does Eli Lilly credit to AI in finance, manufacturing, regulatory writing or drug discovery, given that the company has published no figure for any use case or for the company?
- How many of Eli Lilly's roughly 50,000 employees use the language models served from LillyPod each week, through which interface, and which open-weight models besides NVIDIA's Nemotron 3 Ultra run there?
- Does Eli Lilly stand by the more than 1,000 AI projects in a podcast's summary of a Sep 2025 interview with Thomas Fuchs, and how many of them run in production and outside drug discovery?
- Which cloud providers run Eli Lilly's enterprise AI workloads, and which third party hosts TuneLab?
- What did LillyPod cost, and how is its capacity split among discovery, clinical development, manufacturing and commercial work?
- Did the NVIDIA co-innovation lab begin work in South San Francisco on Mar 31, 2026 as planned, and how many Lilly and NVIDIA staff work there?
- Has any molecule designed with LillyPod, TuneLab or a partner's AI model entered clinical trials?
- To whom does Chief AI Officer Thomas Fuchs report, how is AI work divided among his office, Diogo Rau's organization and Lilly Research Laboratories, and who succeeded Tim Coleman as chief technology officer?
- How many AI and machine learning specialists does Eli Lilly employ?
- Does Eli Lilly still use Yseop to draft patient narratives, and does the 1.4 million hours saved figure that circulates in secondary sources come from Lilly?
- What has the OpenAI antimicrobial collaboration produced since Jun 2024?
Sources
Eli Lilly
- Lilly partners with NVIDIA to build the industry's most powerful AI supercomputer (Oct 28, 2025)
- Lilly launches TuneLab platform (Sep 9, 2025)
- Lilly collaborates with OpenAI on drug-resistant bacteria (Jun 25, 2024)
- Two new Executive Committee members and expanded leadership roles (Nov 6, 2025)
- Lilly launches The Memory Radio Station across Europe (Jun 26, 2026)
- Advancing medicine discovery with AI: the co-innovation lab (Jan 2026)
- A new supercomputer is coming to change the way we make medicines: LillyPod (Feb 2026)
- Diogo Rau, EVP and Chief Information and Digital Officer, biography
- Fostering a thriving workplace (Aug 2026)
- 2025 year in review
Filings and investor materials
- Eli Lilly 2025 Form 10-K (filed Feb 12, 2026)
- Eli Lilly 2024 Form 10-K (filed Feb 19, 2025)
- Eli Lilly 2026 proxy statement (filed Mar 20, 2026)
- Q4 2025 earnings call transcript (Feb 4, 2026)
- J.P. Morgan Healthcare Conference transcript, David Ricks (Jan 13, 2026)
- TD Cowen Health Care Conference transcript, Lucas Montarce (Mar 2, 2026)
- Bernstein Strategic Decisions Conference transcript, Daniel Skovronsky (May 28, 2026)
- Bernstein Healthcare Forum transcript, Lucas Montarce (Sep 25, 2025)
- 2026 annual meeting transcript (May 4, 2026)
Partners
- NVIDIA: Lilly and NVIDIA leaders on the co-innovation lab at J.P. Morgan Healthcare (Jan 13, 2026)
- NVIDIA and Lilly announce co-innovation AI lab (press release via Drug Discovery Online, Jan 12, 2026)
- NVIDIA: Now live, Lilly's AI factory for pharmaceutical discovery and development (Feb 26, 2026)
- Chai Discovery announces collaboration with Eli Lilly (Business Wire, Jan 9, 2026)
- Schrödinger partners with Lilly to make TuneLab available in LiveDesign (Business Wire, Jan 9, 2026)
- Charles River joins Lilly TuneLab (Jun 18, 2026)
- Isomorphic Labs kicks off 2024 with two pharmaceutical collaborations (Jan 7, 2024)
- Profluent announces strategic partnership with Eli Lilly (Apr 28, 2026)
- Profluent: Lilly, Profluent ink deal on AI drugs worth up to $2.25 billion (Bloomberg, Apr 28, 2026)
- WWT: Eli Lilly's Chief AI Officer on why Lilly bet big on on-prem AI (Aug 25, 2026)
News and interviews
- R&D World: Six months in, Lilly says its supercomputer is starting to change the work (Jun 25, 2026)
- Podcast show notes: Thomas Fuchs on AI innovation across healthcare and pharma (Sep 23, 2025)
- Endpoints News: Lilly's TuneLab crosses 100-member mark, adds Chai models (Jun 18, 2026)
- Endpoints News: Lilly's chief technology officer retires after 35 years (Apr 2, 2026)
- Kellogg Magazine: former Lilly CTO Tim Coleman (Summer 2026)
- BioSpectrum Asia: Axcelead DDP joins Eli Lilly TuneLab (Jun 18, 2026)
- BiopharmaTrend: Eli Lilly makes its AI models available via Benchling (Jan 8, 2026)
- Pharmaceutical Technology: Eli Lilly and Insilico Medicine's $2.75bn AI R&D pact (Mar 30, 2026)
- Fierce Biotech: Eli Lilly jumps deeper into AI with $409M Genetic Leap deal (Sep 2024)
- TechCrunch: how Chai Discovery landed its Eli Lilly deal (Jan 16, 2026)
- Pharmaceutical Technology via Yahoo Finance: Eli Lilly and NVIDIA deepen ties with $1bn AI partnership (Jan 12, 2026)
- GuruFocus via Yahoo Finance: Lilly deepens AI drug bet with new Insilico pact (Mar 30, 2026)
- Pulse 2.0: Lilly expands AI drug discovery push (Jun 24, 2026)
- WRTV: Eli Lilly's CEO on AI and GLP-1 drugs at the 150th anniversary (May 12, 2026)
- Cheeky Pint podcast transcript: David Ricks, CEO of Eli Lilly (Nov 11, 2025)
- pharmaphorum: Lilly names Thomas Fuchs as its first chief AI officer (Oct 2024)
- CIO Dive: Eli Lilly appoints Thomas Fuchs as chief AI officer (Oct 9, 2024)
- European Pharmaceutical Review: Eli Lilly appoints Chief AI Officer (Oct 9, 2024)
- Emerj: Artificial intelligence at Lilly (Aug 28, 2024)
Data confidence
- Confirmed
- Stated in a primary or reputable public source.
- Estimated
- Derived or approximated from public reporting.
- Unconfirmed
- Reported, but not yet corroborated.
Eli Lilly profile · Enterprise AI diffusion report by Imagine · facts checked against their sources on Nov 1, 2026
