"Baoyao Technology" Secures Tens of Millions in Angel Round Funding for Building AI Virtual Cell Foundation Model | Exclusive Report by 36Kr

36Kr learned that the AI virtual cell platform company "Baiyao Technology" recently completed an angel round of financing worth tens of millions of yuan. The round was led by Fengrui Capital, with Shunxi Capital participating, and Mingde Capital serving as the exclusive financial advisor.

As early as 2023, the founding team of Baiyao Technology released a single-cell pre-trained foundation model with knowledge enhancement, cross-species capabilities, and hundreds of millions of data points. Subsequently, they rapidly iterated and released the first single-cell pre-trained foundation model based on a graph structure, enabling AI to understand the laws of gene regulation and changes in cell states under internal and external environments, thereby simulating real cell behavior. The series of research results were published as cover articles in "Cell Research" and authoritative journals such as "Advanced Science" and "NSR".

Baiyao Technology told 36Kr, "The single-cell foundation model built by the team follows the pre-training + fine-tuning path, demonstrating excellent generalization ability and broad potential for downstream applications. This lays a technical foundation for building an AI virtual cell platform that can be industrialized."

During the training process, the model integrates prior biological knowledge, making the learning more targeted and the output more in line with biological logic. In addition, the model integrates over a hundred million single-cell gene expression data for the first time and conducts cross-species pre-training on both human and mouse species, thus learning common life laws and providing the possibility for effectively transforming animal experiment data and predicting human cell responses.

In the past two or three years, with the evolution of technology, "AI virtual cells" have become a highly sought-after hot track. The data acquisition technology represented by single-cell sequencing has matured, leading to an explosive growth of multi-omics data, which provides "fuel" for the training of AI virtual cell models.

Moreover, the emergence of the self-supervised pre-training strategy based on the self-attention mechanism has provided a powerful tool for processing high-dimensional and complex life science data, making it possible to finely characterize and dynamically simulate cells.

At the policy level, countries such as China and the United States are actively promoting "using basic AI models for life simulation". China's 14th Five-Year Plan emphasizes the accelerated development of biotechnology and the bioindustry and has accordingly planned major national scientific and technological infrastructure for AI life science. In April 2025, the US FDA also advocated using AI models and organ chips to replace traditional animal experiments in antibody drug R & D.

In the industrial aspect, overseas companies represented by Xaira Therapeutics and Asimov have taken the lead in exploring the application of AI virtual cell models in areas such as the discovery of new tumor targets, directed differentiation of stem cells, and engineering cell transformation.

The Baiyao team introduced, "As the smallest life unit outside of viruses, cells integrate key levels of biological signals. Using AI to simulate cells can answer a series of key questions in basic scientific research and industrial drug R & D. With continuous breakthroughs in global omics observation technology and artificial intelligence algorithms, cell models will gradually incorporate more dimensions of omics information, evolving from single-omics models to multi-modal models. Eventually, the entire dynamic life process of cells will be deduced in the virtual world. This will allow us to master the life intelligence of the smallest unit and explore more complex tissues and organs, disease models, and various life activities with new tool platforms and methodologies."

However, building AI virtual cells is a grand project with high complexity and multi-dimensional challenges. In addition to algorithmic difficulties, it has extremely high requirements for resources such as computing power and data. To create a virtual cell foundation model that covers multi-modal data and is close to being complete, the investment scale may reach the level of "billions of US dollars".

In response, the Baiyao Technology team pointed out that the current AI virtual cell technology is still in its early stage of development. The first batch of large virtual cell models launched by companies such as Baiyao have initially verified the theoretical path from 0 to 1. The technical characteristics of large models determine that there is a strong Matthew effect in this field. In the future, if a company can continuously lead in high-quality data accumulation and algorithm iteration, the life science R & D platform it builds is expected to achieve exponential optimization in efficiency and cost for both basic research and the biomedical industry.

Investors' Views

Xie Da, the vice president of Fengrui Capital, said that cells are the basic unit of life and the "main battlefield" for disease causation research and drug R & D. Virtual cells rely on biological omics data and the capabilities of large artificial intelligence models to build a "digital twin" of cells, which is expected to empower common scientific research and drug R & D scenarios such as cell development, drug sensitivity, and antibody production, thereby significantly reducing the cost and cycle of trial and error. Baiyao Technology has been deeply involved in the field of virtual cell research for many years and has taken the lead in completing the development of multiple foundation models and specialized models for specific tasks. It has also cooperated with domestic and foreign pharmaceutical companies. We are optimistic about the company's ability in multi-modal data collection and modeling and look forward to its continuous progress in areas such as cell drug development and biological manufacturing.

Hu Mingyi, a partner at Mingde Capital, said: Life science research is spanning multiple key scales - from macroscopic phenotypes and tissues to microscopic molecular pathways, and then to the cell function scale that bridges the gap. In the past, we lacked the ability to conduct dynamic and quantitative research at the integrated scale of cells and could only infer cell dynamics from cross-sectional data. Now, the integration of massive multi-omics data and the Transformer algorithm enables us to build high-fidelity virtual cells. This platform will completely break down the barriers between mechanism research and application development, bringing life science into a new era of simulation and programmability. The Baiyao Technology team has a combined advantage in AIVC key algorithms and high-quality data. We look forward to exploring this historical opportunity with them.