MobiDrop has been officially accepted into the NVIDIA Inception program. By leveraging microfluidics and single-cell multi-omics technologies, MobiDrop is integrating high-quality life science data generation with virtual cell R&D, while exploring applications for virtual patients and clinical trials.

NVIDIA Inception is NVIDIA's global acceleration program for innovative technology companies, offering members AI development tools, technical training, and industry and venture capital connections. Companies such as Perplexity, Hugging Face, Cohere, and Insilico Medicine have all been featured in NVIDIA's official member case studies. MobiDrop is now part of this global innovation network.

MobiDrop has recently formed partnerships with Insilico Medicine (https://mp.weixin.qq.com/s/mMQ0mimn2ceKk7BJXJxemg) and INFevo (https://mp.weixin.qq.com/s/wpkg55E6B6XV2jOQ4ON62g) to jointly advance AI-ready multi-omics data and virtual cell infrastructure. In parallel, MobiDrop is developing a virtual cell engine built around its proprietary MobiBrain dry-wet closed-loop intelligent system, creating a continuous feedback loop among data generation, model learning, and experimental validation.

In January 2026, at the J.P. Morgan Healthcare Conference, Jensen Huang spoke about the application of computational technology in biomedical research: "I can't imagine a field more worthy of applying computer science than this (life sciences)... I hope we can change the trajectory of history."

For this vision to become reality, a fundamental question must be answered: What kind of cell data does AI actually need in order to understand life?

01 For AI to Understand Life, Cell Data Is Essential

For AI to learn about the physical world, it needs data on scenes, objects, and interactions. As AI's exploration extends into living systems, it similarly requires learning materials drawn from the real world.

This time, however, the subject of study is the cell—along with the complex molecular relationships within it and its responses to various interventions.

What state is a cell in right now? How will it change after a gene is perturbed or a drug is applied? And can a model's predictions be validated through real experiments?

These questions tie together observation, prediction, and validation in life science research more tightly than ever, and they place new demands on data generation.

Data that captures cell states helps researchers understand living systems; data that carries perturbation information—such as gene or drug perturbations—further helps researchers explore how living systems respond to change. For life models and virtual cells, data must convey richer biological information and meet the quality and scale requirements needed to support sustained research.

At the 2024 Databricks Data + AI Summit, Jensen Huang said: "Every company's business data is its own gold mine." In the life sciences, that value must be built on experimental data that is real, traceable, and sustainably produced.

In MobiDrop's view, the large-scale production of high-quality cell data is one of the key bottlenecks in advancing life models and virtual cells.

Addressing this challenge requires going deep into the experimental process: how to measure complex living systems, how to manage data quality, and how to support continuous model research iteration in terms of both efficiency and cost.

The life science material center that MobiDrop is building is designed to provide sustained data production capacity for precisely these questions.

02 From Microfluidics Expertise to a Life Data Production System

Building life data production infrastructure requires solid experimental technology and engineering capabilities.

MobiDrop's technological foundation stems from its sustained expertise in microfluidics and single-cell technologies. Across microfluidic operations such as droplet generation, microinjection, fusion, and sorting—as well as processes including tissue dissociation, single-cell sorting, sequencing library preparation, and bioinformatics analysis—MobiDrop has developed a corresponding portfolio of instruments, reagents, and software.

These capabilities span multiple stages, from real sample processing to data analysis, laying the groundwork for further automating experimental workflows and scaling up data production.

Toward AI for Science, MobiDrop is combining this expertise with automated perturbation experiments, multimodal measurement, and the development of an industrial quality control system.

Automated perturbation experiments focus on how cells respond to change. By capturing cell response data around interventions such as gene and drug perturbations, they provide real experimental data for model research.

Multimodal measurement addresses different levels of living systems. By continuously expanding species, sample types, and omics dimensions, it provides richer information for studying cell states and molecular relationships.

The industrial quality control system ensures the stability and traceability of the data production process. Through process standardization, in-process monitoring, and result feedback, it bridges quality management and large-scale production.

Together, these efforts serve a single goal: to measure living systems accurately, comprehensively, and rapidly—and to continuously drive down measurement costs—so that data from real experiments can better serve AI.

03 Let Data Enter Models, Let Predictions Return to Experiments

The value of the life science material center is further demonstrated by the connection between data and the research process.

In MobiDrop's approach, experimental protocols must be translated into instrument-executable workflows, experimental quality must be monitored in process, and data and its provenance must be traceable. Only then can experimental results gradually accumulate into reusable research resources that continuously support model learning and validation.

At the other end of this system is feedback from models to real experiments.

Experiments generate data; data supports model learning; models make predictions; predictions undergo experimental validation; and new results feed back into the models. Through this continuous feedback loop, MobiDrop aims to advance our understanding of living systems through the combined power of computation and experimentation.

Around its proprietary MobiBrain dry-wet closed-loop intelligent system, MobiDrop is advancing the coordinated R&D of a virtual cell engine and experiments, integrating model training needs, experimental design, and validation feedback into a single R&D workflow.

Two recent partnerships are advancing this vision.

Insilico Medicine and MobiDrop have formed a strategic collaboration to jointly build AI-ready multi-omics data infrastructure.

MobiDrop × INFevo: Jointly building virtual cell infrastructure and closing the "data–model–experiment" loop.

From data production to model application and on to validation in real experiments, MobiDrop aims to leverage its expertise in experimental technology and data production to work with research and industry partners in advancing life model and virtual cell research.

Moving from virtual cells to virtual patients will further require integrating cellular-level drug responses with clinical characteristics, treatment records, and follow-up data. MobiDrop aims to apply its data production and experimental validation capabilities to explore related research with pharmaceutical companies and clinical institutions, supporting clinical trial design, patient stratification, and study population selection.

Building a Long-Term Data Foundation for AI to Explore Life

NVIDIA Inception is NVIDIA's global ecosystem program for startups, helping companies grow through technical tools and training, ecosystem exchange, and market support.

This acceptance provides MobiDrop with a new opportunity to further expand technical exchange and ecosystem collaboration in the AI field.

For MobiDrop, joining the global innovation ecosystem goes hand in hand with continuing to build its own experimental and data production capabilities. The interdisciplinary exploration of life sciences and AI requires collaboration across fields—models, experiments, engineering, and data—and requires translating that collaboration into concrete research questions and validation processes.

The potential of AI for Science will be unlocked gradually through validation in real living systems.

When experiments can continuously generate high-quality data, when data can be learned by models, and when model predictions can return to experiments for validation, life data production infrastructure can become a vital bridge connecting scientific questions, computational exploration, and experimental discovery.

Looking ahead, MobiDrop will continue to advance automated perturbation experiments, multimodal measurement, and industrial quality control system development—building an AI-oriented life science material center, providing high-quality data support for life models, virtual cells, and virtual patient research, and driving their application in drug development and clinical trials.

Starting from a single cell, making life computable.