After AI computing power transformed the generative content industry, NVIDIA has set its sights on an even more impactful field—life sciences—announcing a significant expansion of its BioNeMo platform, aiming to bring the "Transformer moment" to the biological community. Most notably, this includes a five-year, in-depth collaboration with pharmaceutical giant Eli Lilly, with a planned investment of up to $10 billion to build its first "AI Co-Innovation Lab." NVIDIA is also partnering with instrumentation giant Thermo Fisher Scientific to transform traditional laboratories into automated data factories.
NVIDIA x Eli Lilly: A $10 billion bet on the integration of wet and dry laboratories
This collaboration is seen as a landmark partnership between the pharmaceutical and technology industries. NVIDIA and Eli Lilly will establish a joint lab in the San Francisco Bay Area, with both companies planning to invest up to $10 billion over the next five years, covering talent, infrastructure, and computing resources.
The core objective of this laboratory is to break down the traditional barriers between "wet labs" (where actual chemical/biological experiments are conducted) and "dry labs" (where computer simulations and computations are performed).
By incorporating the NVIDIA BioNeMo platform and future Vera Rubin architecture supercomputers, Eli Lilly plans to establish a "scientist-in-the-loop" continuous learning system. This means that AI will design molecules, which will then be automatically synthesized and tested by robots, with the data fed back to AI to refine the model, followed by further design by AI. This closed loop will operate 24/7, significantly shortening the drug development cycle.
In addition to research and development, the two parties will also use NVIDIA Omniverse to create a digital twin of the manufacturing line, optimizing the supply chain and production processes in the virtual world first.
BioNeMo platform undergoes major upgrade: More than just design, it also ensures "it can be made."
NVIDIA also announced a major update to the BioNeMo platform, making it more aligned with real-world drug development needs.
The new features include:
• RNAPro model:Specifically designed for predicting RNA structure, this is crucial for developing next-generation vaccines and gene therapies.
• ReaSyn v2 model:In the past, AI-designed molecules were often too far-fetched to be realistically synthesized. ReaSyn v2 is specifically designed to predict the "synthesibility" of molecules, ensuring that AI-designed drugs are something chemists can actually make.
• BioNeMo Recipes:It provides a standardized set of recipes, enabling research institutions to train and deploy customized biological models more quickly.
Thermo Fisher: Making Lab Instruments Smarter
To achieve full automation in its laboratories, NVIDIA partnered with Thermo Fisher, a leading scientific instrument manufacturer. The collaboration focuses on integrating NVIDIA DGX Spark edge computing and NVIDIA NeMo agent-based AI into experimental equipment.
This will transform experimental instruments from passive data producers into intelligent nodes with "edge thinking" capabilities. For example, AI agents can automatically monitor the quality of experimental data, adjust parameters in real time if anomalies are detected, and even automatically generate subsequent experimental procedures, turning the laboratory into a highly efficient "data factory."
Analysis of viewpoints
NVIDIA CEO Jensen Huang has repeatedly mentioned that "digital biology" is the next tipping point for AI. This collaboration with Eli Lilly and Thermo Fisher shows that NVIDIA has evolved from simply "selling shovels" (selling GPUs) to "teaching people how to mine" (providing the BioNeMo platform and lab architecture).
The collaboration with Eli Lilly is particularly significant. In the past, pharmaceutical companies have mostly adopted AI in a piecemeal manner, but this time it is a "whole plant output" concept, which fully integrates AI from the underlying computing power (Vera Rubin), the middleware platform (BioNeMo), to the application layer robots and digital twins (Isaac/Omniverse).
This also reflects a paradigm shift in drug discovery: from the past "high-throughput screening" to "generative design." If BioNeMo can generate usable drug molecules as precisely as ChatGPT generates articles, then the speed at which humanity fights cancer and rare diseases will increase exponentially.
Of course, the complexity of biology is far greater than that of language models, and the entire industry is watching to see if this $10 billion investment will bear fruit within five years.
