Nvidia, Eli Lilly just say yes to making drugs together, using Vera Rubin GPUs
(2026/01/12)
- Reference: 1768246206
- News link: https://www.theregister.co.uk/2026/01/12/nvidia_eli_lilly_ai_drug_lab/
- Source link:
Nvidia has teamed up with pharmaceutical heavyweight Eli Lilly to plow up to $1 billion into a research lab over the next five years to advance the development of foundation models for AI-assisted drug discovery.
Announced at the JPMorgan Healthcare conference on Monday, the collaboration will span the infrastructure, talent, and compute necessary to develop these biology and chemistry models using Nvidia's BioNeMo software platform and Vera Rubin accelerators.
Introduced in fall 2022, just months before ChatGPT kicked off the AI arms race, BioNeMo is an open source framework for building and training deep learning models for use in drug discovery.
[1]
Located in the San Francisco Bay Area, the so-called co-innovation lab will bring together Eli Lilly's top biologists and chemists to work alongside Nvidia's software engineers and model devs, when it opens later this year.
[2]
[3]
"Combining our volumes of data and scientific knowledge with Nvidia's computational power and model-building expertise could reinvent drug discovery as we know it," Eli Lilly CEO David Ricks said in a canned [4]statement .
“The focus will be on creating a laboratory where AI software (the dry lab) and robotic hardware (the wet lab) talk to each other 24/7, keeping humans in the loop without the need for constant reprogramming the robot or manually completing every step of the experiments.. This amplifies the scientist’s productivity, elevating them to a strategic lead, while the system executes the iterative testing cycle at machine speed,” an Nvidia spokes person told El Reg .
[5]
From there, the two companies will harness Nvidia's newly unveiled Vera Rubin compute platform. Announced at CES last week, the system promises a fivefold increase in performance over Nvidia's prior-gen Blackwell GPUs. These chips will provide the computational grunt necessary to train new foundation models based on the lab's research.
This suggests the lab will be among the first to get its hands on the chips, which aren't expected to be available in any significant numbers until the second half of this year.
[6]AI is actually bad at math, ORCA shows
[7]'It looks sexy but it's wrong' – the problem with AI in biology and medicine
[8]Boffins warn that AI paper mills are swamping science with garbage studies
[9]Lab-grown human brain cells drive virtual butterfly in simulation
In the meantime, it's not like Eli Lilly is hurting for compute. At GTC DC last October, the pharmaceutical giant [10]revealed it had deployed a Blackwell Ultra-based SuperPOD complete with 1,016 B300 GPUs to support its exploration into computational biology and chemistry.
The lab's sole focus won't be limited to AI drug discovery, however. Researchers will also explore applications for AI in clinical development, manufacturing, and commercial operations.
For example, Eli Lilly is also investigating Nvidia's Omniverse Robotics platforms as a means to optimize its manufacturing plants and increase production of high-demand drugs. ®
[11]
Updated at 2038 to include a quote from Nvidia.
Get our [12]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[4] https://nvidianews.nvidia.com/news/nvidia-and-lilly-announce-co-innovation-lab-to-reinvent-drug-discovery-in-the-age-of-ai
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2025/11/17/ai_bad_math_orca/
[7] https://www.theregister.com/2025/07/27/biomedviz_ai_wrong_problems/
[8] https://www.theregister.com/2025/05/13/ai_junk_science_papers/
[9] https://www.theregister.com/2024/10/22/human_brain_tissue_butterfly_simulation/
[10] https://blogs.nvidia.com/blog/lilly-ai-factory-nvidia-blackwell-dgx-superpod/
[11] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[12] https://whitepapers.theregister.com/
Announced at the JPMorgan Healthcare conference on Monday, the collaboration will span the infrastructure, talent, and compute necessary to develop these biology and chemistry models using Nvidia's BioNeMo software platform and Vera Rubin accelerators.
Introduced in fall 2022, just months before ChatGPT kicked off the AI arms race, BioNeMo is an open source framework for building and training deep learning models for use in drug discovery.
[1]
Located in the San Francisco Bay Area, the so-called co-innovation lab will bring together Eli Lilly's top biologists and chemists to work alongside Nvidia's software engineers and model devs, when it opens later this year.
[2]
[3]
"Combining our volumes of data and scientific knowledge with Nvidia's computational power and model-building expertise could reinvent drug discovery as we know it," Eli Lilly CEO David Ricks said in a canned [4]statement .
“The focus will be on creating a laboratory where AI software (the dry lab) and robotic hardware (the wet lab) talk to each other 24/7, keeping humans in the loop without the need for constant reprogramming the robot or manually completing every step of the experiments.. This amplifies the scientist’s productivity, elevating them to a strategic lead, while the system executes the iterative testing cycle at machine speed,” an Nvidia spokes person told El Reg .
[5]
From there, the two companies will harness Nvidia's newly unveiled Vera Rubin compute platform. Announced at CES last week, the system promises a fivefold increase in performance over Nvidia's prior-gen Blackwell GPUs. These chips will provide the computational grunt necessary to train new foundation models based on the lab's research.
This suggests the lab will be among the first to get its hands on the chips, which aren't expected to be available in any significant numbers until the second half of this year.
[6]AI is actually bad at math, ORCA shows
[7]'It looks sexy but it's wrong' – the problem with AI in biology and medicine
[8]Boffins warn that AI paper mills are swamping science with garbage studies
[9]Lab-grown human brain cells drive virtual butterfly in simulation
In the meantime, it's not like Eli Lilly is hurting for compute. At GTC DC last October, the pharmaceutical giant [10]revealed it had deployed a Blackwell Ultra-based SuperPOD complete with 1,016 B300 GPUs to support its exploration into computational biology and chemistry.
The lab's sole focus won't be limited to AI drug discovery, however. Researchers will also explore applications for AI in clinical development, manufacturing, and commercial operations.
For example, Eli Lilly is also investigating Nvidia's Omniverse Robotics platforms as a means to optimize its manufacturing plants and increase production of high-demand drugs. ®
[11]
Updated at 2038 to include a quote from Nvidia.
Get our [12]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[4] https://nvidianews.nvidia.com/news/nvidia-and-lilly-announce-co-innovation-lab-to-reinvent-drug-discovery-in-the-age-of-ai
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[6] https://www.theregister.com/2025/11/17/ai_bad_math_orca/
[7] https://www.theregister.com/2025/07/27/biomedviz_ai_wrong_problems/
[8] https://www.theregister.com/2025/05/13/ai_junk_science_papers/
[9] https://www.theregister.com/2024/10/22/human_brain_tissue_butterfly_simulation/
[10] https://blogs.nvidia.com/blog/lilly-ai-factory-nvidia-blackwell-dgx-superpod/
[11] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aWV9Clep7AKPD7pP5gdIOgAAAAk&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[12] https://whitepapers.theregister.com/
Re: So what?
Sorry that handle is already taken.
So why bother? Hype, baby!
Also possibly desperation. Drug discovery is a difficult business and all the low hanging fruit was picked long ago.
"the system promises a fivefold increase in performance over Nvidia's prior-gen Blackwell GPUs"
Androgynous Cow Herd
Mkay - Fivefold increase in less than 2 years?
At what level precision?
Keeping Blackwells "Performance" level while improving precision would be WAY more valuable that increasing the pace it spits out crap data.
The sort of research they are talking about here requires 16 bit minimum precision to not be any enormous waste of time. Blackwell is 4 bit for he published numbers. When you require 16 bit - it is indistinguishable from Hopper.
So what?
This won't speed up the FDA though. LIkely will slow it down. So why bother?