Paris-based biotechnology startup Rivercell has launched with a $25 million seed round to build an AI model capable of predicting how human cells respond to drugs and genetic changes. Led by HV, the financing will support Rivercell’s proprietary biological data platform, Paris wet lab and AI Virtual Cell program as the company targets a major bottleneck in AI-driven drug discovery: generating enough high-quality data to model cellular behavior.
Rivercell is taking a different route into AI-powered drug discovery: instead of starting with molecules, it is trying to build a predictive model of the cells those molecules act on.
The Paris-based biotechnology and artificial intelligence company has emerged from stealth with a $25 million seed round led by HV, with participation from HCVC, Alven and Bpifrance Digital Venture. The funding will be used to scale Rivercell’s proprietary biological data-generation platform, expand its wet lab in Paris and launch its AI Virtual Cell program.
The company’s central thesis is that AI drug discovery has a data problem. Models can only learn cellular behavior if they have access to sufficiently broad and detailed measurements of what happens when cells encounter drugs, genetic changes and other perturbations.
Rivercell is building infrastructure specifically to generate that data.
Its planned AI Virtual Cell, or AIVC, is described as a world model designed to predict how human cells respond to treatments and genetic changes in silico. Rather than building a model around one disease indication, Rivercell says its approach is intended to learn more fundamental rules governing cellular responses.
That distinction could be important for AI drug discovery. Models trained primarily on disease- or molecule-specific datasets can be useful for individual tasks, but a sufficiently general model of cellular behavior could potentially be applied across therapeutic areas.
The challenge is obtaining the training data.
AlphaFold demonstrated what becomes possible when AI has access to large bodies of structured biological information. Protein structure prediction benefited from decades of accumulated experimental and publicly available data. Rivercell argues that comparable datasets do not yet exist for the dynamic behavior of human cells.
Cellular responses are also considerably more complex than static structures. Researchers may need to observe how cells change over time, under different treatments and across multiple biological measurements. Rivercell says its platform is designed to capture these multimodal and time-resolved responses at scale.
That makes Rivercell as much a machine-learning infrastructure company for biology as a drug-discovery startup.
The company says conventional laboratory instruments were not designed to generate the volume and structure of data required to pretrain a world model of the cell. It is therefore developing its own data-generation platform alongside an automated wet lab. CEO Yann Fleureau has described the system as analogous to a “GPU for the bio data center”—a piece of specialized infrastructure intended to make large-scale biological data generation possible.
The approach puts Rivercell into an increasingly crowded but rapidly expanding field of AI for drug discovery. Companies such as Owkin, Recursion and Insilico Medicine have pursued AI-based approaches to understanding disease biology, identifying targets and designing therapeutics. Other efforts are focused on biological foundation models, protein engineering and virtual experimentation.
The timing reflects growing investment in the category. McKinsey estimates that funding for AI-enabled drug discovery reached $8.4 billion in 2025, more than double the $4.1 billion recorded in 2023. The consultancy also argues that the industry’s biggest opportunity may increasingly lie beyond molecule design, where AI tools are already comparatively mature, toward identifying and validating disease mechanisms.
That is closely aligned with Rivercell’s strategy.
McKinsey estimates that generative AI could create $60 billion to $110 billion in annual economic value across pharmaceutical and medical-product industries, with research and early discovery accounting for an estimated $15 billion to $28 billion of that opportunity.
But a virtual cell is a much more ambitious proposition than applying an LLM to scientific literature or using machine learning to rank candidate molecules.
A useful model would need to predict biological consequences rather than simply recognize correlations. Rivercell’s data-generation platform is intended to create a closed loop in which experiments produce training data, the model generates predictions and those predictions help determine which experiments should be run next. Eric Durand, Rivercell’s chief scientific officer, says the company wants to use this loop to guide subsequent experiments and move more treatment discovery into virtual cells.
That could eventually reduce some physical experimentation, although Rivercell has not demonstrated that its system can replace laboratory or clinical validation. This distinction matters: predictions made in silico still need to be tested experimentally before they can support drug-development decisions.
Rivercell was founded in 2025 by Yann Fleureau, the entrepreneur behind AI diagnostics company Cardiologs, and joined in 2026 by Eric Durand. Fleureau previously co-founded Cardiologs, which was acquired by Philips in 2021. Durand has held data-science roles at Novartis and Owkin and was involved in the founding of biological foundation-model company Bioptimus.
The company is also assembling scientific expertise around its approach. Charlotte Bunne of EPFL and Fabian Theis of Helmholtz Munich and the Technical University of Munich are joining Rivercell’s Scientific Advisory Board, according to the company.
Europe is an important part of the strategy. Rivercell is building its data-generation and wet-lab operations in Paris while connecting to the broader European AI and pharmaceutical ecosystem, including the Basel life-sciences cluster.
The bigger question is whether Rivercell can turn biological data generation into a scalable AI infrastructure layer.
If successful, an AI virtual cell could become useful across target discovery, mechanism-of-action research, drug-response prediction and other stages of preclinical research. But the hardest part may not be building a larger model. It may be producing sufficiently diverse, reliable and experimentally grounded data to ensure that the model learns biology rather than artifacts in the measurement process.
That makes Rivercell’s $25 million round notable less for the size of the financing than for what it is trying to fund: the underlying data and laboratory infrastructure required to make a general-purpose AI model of human cellular behavior possible.
Market Landscape
AI drug discovery is moving beyond molecular design toward models that attempt to represent broader biological systems. McKinsey’s September 2026 analysis argues that the industry’s central bottleneck is increasingly the identification and validation of disease mechanisms, where novel causal biology remains scarce.
Rivercell’s strategy sits directly in this emerging category. Instead of treating AI as a tool layered onto conventional discovery, it is combining automated wet-lab experimentation, biological data infrastructure and machine-learning models into one system.
The opportunity is substantial, but so are the technical hurdles. A virtual cell must contend with biological heterogeneity, incomplete measurements, temporal dynamics and the difference between statistical prediction and causal biological understanding.
Competition is likely to increase as AI foundation models move deeper into life sciences. The strongest platforms may ultimately be distinguished not simply by model size, but by proprietary experimental data, closed-loop experimentation and the ability to validate predictions in the laboratory.
Top Insights
- $25M launch round: Rivercell is emerging from stealth with funding to build proprietary biological data infrastructure and an AI model of cellular responses.
- Data is the differentiator: The company’s strategy centers on generating multimodal, time-resolved experimental data that existing biological datasets generally do not provide at sufficient scale.
- Virtual-cell ambition: Rivercell wants its AI model to predict cellular responses to drugs and genetic changes, potentially reducing some physical experimentation.
- AI drug discovery shifts upstream: The company’s focus reflects a broader move toward modeling disease mechanisms and cellular biology rather than concentrating solely on molecule generation.
- Infrastructure matters: Rivercell is building laboratory and data-generation systems alongside its AI model, treating experimental data as core technology rather than an external input.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
