Why Are Organoid and Organ-on-a-Chip Companies All Turning to AI Virtual Cells?

In 2026, Xellar Biosystems completed nearly RMB 400 million in Series A financing within six months, clearly earmarking the funds to accelerate the construction of virtual cells and biological world models, thereby advancing the closed-loop flywheel of “data generation–mechanism elucidation–model training–experimental validation.”

During the same period, DarkJade Sciences completed a Pre-A+ round of financing exceeding RMB 100 million, directed toward AI-integrated all-in-one equipment for organoids and AI-enabled organ-on-a-chip technology. CANCHEN TECH secured tens of millions in angel-stage financing, focusing on AI-driven virtual organs and full-scale digital twin simulation systems. As of August, more than six financing deals were disclosed in China’s organoid and organ-on-a-chip sector, with a cumulative amount exceeding RMB 1.1 billion.

Behind the Resurgence of Financing Fever, Organoid and organ-on-a-chip companies are unanimously doing the same thing—labeling themselves with “AI” and “AIVC,” and building coupled capabilities for AI + virtual experimentation.

Meanwhile, from a regulatory perspective, organoids and organ-on-a-chip technologies have moved beyond the research stage and are now integrated into the evidence chain for Investigational New Drug (IND) applications. The FDA’s ISTAND program has begun accepting virtual liver models for qualification as Drug Development Tools (DDTs). Furthermore, tumor organoid assays have been included in China’s guidelines for pricing items of medical laboratory tests, marking their entry into the pre-clinical testing system and standardized pathways.

As the sole reliance on animal testing in new drug development is broken, organoid and organ-on-a-chip companies are accelerating this replacement process by leveraging “AI + human-derived dynamic data.”

As the model layer achieves mastery of biological principles and scientists’ know-how, data and empirical insights must be grounded in the real physical world.

The essence of AIVC is to compress all observable features and activities of real cells into computable mathematical objects, enabling counterfactual computation and perturbation prediction—“how would the cell change if X were altered.” Training such models requires multi-scale, multi-modal biological data that include systematic perturbation responses, along with a structured, standardized, cross-scale, and traceable data architecture.

The crux of the issue lies precisely here. High-quality paired data required for AIVC are extremely scarce, batch effects and platform discrepancies are substantial, and data on systematic perturbations exhibit sparse distribution. Although public datasets already contain tens of millions of cellular profiles, most are “snapshot”-style static data, lacking key causal and dynamic data on drug/gene perturbations and multi-omics responses.

This is precisely the critical niche for organoid and organ-on-a-chip enterprises: providing authentic humanized microenvironments within highly biomimetic, miniaturized models that are controllable, observable, and amenable to intervention. By leveraging automated experimental systems, these platforms continuously generate high-dimensional data streams across multiple time points, doses, phenotypes, and omics layers, including imaging, transcriptomics, metabolomics, and functional readouts.

In terms of iterative logic, investing in AI+ and AIVC enables organoid and organ-on-a-chip companies to leverage their advantage in generating high-quality dynamic data as their ticket into the era of AI-driven productivity.

It cannot be ignored that engineering, standardization, low cost, and high stability are essential attributes for the establishment of the “data manufacturing” role. Leading organoid/organ-on-a-chip companies are transforming themselves into "Humanized, Perturbable, Multimodal Data Infrastructure Service Provider".

It is worth noting that AIVC is fundamentally reshaping the division of labor across the entire new drug R&D industry chain, shifting from a “fragmented toolchain structure” to an “integrated system of data + models + validation.”

This presents a rare opportunity for organoids and organ-on-a-chip technologies—originally positioned upstream as “water sellers”—to break through to higher value tiers. In the past, they were merely one link in the CRO chain, providing wet-lab tools; whereas in the AIVC era, they are Breaking away from the positioning as a single tool vendor and upgrade to become the "industrial-grade data infrastructure" of the industry chain.

The core logic of the breakthrough path lies in the fundamental transformation of roles: From "Data Endpoint" to "Data Entry Point". In the past, organoids and organ-on-a-chip systems represented the end point of business operations, with data generation marking the conclusion of services. However, under the AIVC architecture, these high-quality wet-lab experimental data will serve as fuel for AI models, functioning as the entry point for data infrastructure.

To seize this entry point, leading enterprises are building multi-dimensional data engines through cross-industry integration:

In July, iORGENtech partnered with Salus, a leading domestic platform enterprise in gene sequencing and spatial omics, to establish a collaborative facility for “organoids + genomics/spatial omics.” Xellar Biosystems collaborated with Nanomics, an AI-driven digital proteomics company, to explore the next-generation R&D paradigm of “organ-on-a-chip + proteomics.”

Further deconstructing the underlying logic, cross-sector collaboration effectively accomplishes two critical infrastructure tasks:

First, establish an automated and standardized data production workflow. The wet-lab capabilities and automation systems provided by organoid and organ-on-a-chip companies, along with “live data” encompassing temporal, functional, and mechanistic dimensions, are being deeply integrated with upstream omics tools. This collaboration transforms fragmented experimental data into standardized, industrial-grade data streams, aiming to address the supply challenge of large-scale, high-quality data required for training AIVC models.

Second, establish a closed-loop data flow between “phenotype” and “mechanism.” Modelable dry-lab analyses fill the gap in mechanistic elucidation left by wet-lab experiments: organoids and organ-on-a-chip systems record “what happens at the cellular/tissue level” (phenotype), while multi-omics data explain “what happens at the transcriptomic and proteomic levels” (mechanism). This multiscale integration faithfully recapitulates cellular regulatory logic, propelling AI models from “pattern recognition” toward deeper “mechanistic understanding.”

When organoid and organ-on-a-chip companies can provide industrial-grade data streams with complete "phenotype-mechanism" mapping, they will become indispensable partners for AI model iteration.

Within the pharmaceutical R&D industry chain, AIVC shares a similar commercialization pathway with organoids and organ-on-a-chip technologies. This includes key safety assessments such as traditional toxicity and efficacy prediction—covering drug-induced liver injury, cardiotoxicity, CNS penetration, and immunotoxicity—as well as personalized treatment simulation, combination drug screening, and disease progression prediction.

In the long term, the application scenarios of AIVC will further expand, such as virtual clinical trial design, cross-border virtual experiments, and even reshaping the way evidence is generated in the R&D chain. For organoid and organ-on-a-chip companies, this also means New Pathways for Commercial Expansion.

Xie Xin of Xellar Biosystems pointed out that, The future business model will not be solely about selling models, but rather a combination of "model + data infrastructure + decision evidence services." Its core value lies in replacing or compressing three types of costs: wet lab expenses in early-stage screening, non-translatable components of animal studies, and the high probability of substantial losses from failures after entering clinical trials.

Wan Fei of iORGENtech also affirmed this translation pathway. He believes that organoids have already realized a flywheel effect transitioning from “proving consistency” to “training specialized small models.” Specifically, vertical small models with “gray-box/semi-black-box” characteristics are trained using organoid data demonstrating high clinical consistency, and then applied to specific disease states and various drug evaluation scenarios. By continuously accumulating high-quality, standardized, multimodal life science data, a professional model system covering diverse application scenarios can be established. For organoid models validated through Investigational New Drug (IND) applications, the corpus data they generate can in turn feed back into the training of AIVC models.

Wan Fei believes that the continuous expansion of AI’s capabilities will inevitably drive organoids to evolve into the next generation of AIVC/AIVO (AI Virtual Organs). Physical organoids face commercialization ceilings due to constraints in production scale and cost, cold-chain logistics, and cross-border regulatory compliance. In contrast, within the AIVC/AIVO paradigm, the deliverable will shift from biological samples to data interfaces. At that time, The core costs of commercialization may be limited to electricity and internet fees.。

Compared with the world before the emergence of AIVC, organoid and organ-on-a-chip companies still occupy an equally, if not more, important position in the industry.

How to Position Your Role in This Industry Chain: The Underlying Logic of Thinking About the Future of AI4S.

The commonality lies in becoming a modular data infrastructure that fosters upstream and downstream collaboration. The lack of consensus, however, points to more diverse diverging paths.

Wan Fei believes that, in the era of AI democratization, the future scenario for drug development lies in pharmaceutical companies independently training AI agents or vertical models locally. These models, potentially based on general-purpose world models or expert models, can leverage proprietary clinical data or disease-specific datasets to train customized, proprietary Agents. In this landscape, organoid companies will become High-Quality Data Providers Behind Pharmaceutical Companies' Proprietary Model Training.

Under this logic, corpus generation requires not only foundational organoid technologies and research models, but also a stable, high-throughput, automated workstation/laboratory system, along with automated plugins for reading multimodal and multi-omics data, ultimately evolving into a fully automated data generation ecosystem that integrates the entire upstream and downstream value chain.

Wan Fei stated, “Both R&D and commercialization must return to the application itself. Differentiated scenarios determine the data input modules: some scenarios require only dynamic imaging, others necessitate the integration of multi-omics data, and still others demand comprehensive solutions.” This means that The commercialization of modular data infrastructure does not need to wait for highly rich and complete nodes.

In Xie Xin’s view, “An outstanding AI for Science (AI4S) company in the biopharmaceutical sector should Possesses capabilities in bioengineering, automated experimentation, AI modeling, and regulatory interface integration.” The narrative around AIVC and world models has shifted from ‘accelerating drug discovery’ to ‘building a computational system capable of simulating human biology.’ While establishing native data infrastructure, Xellar Biosystems is also developing model-layer capabilities, embodying a data-first foundation model strategy: first constructing a sustainable, human-derived data super-factory, and then training interpretable and regulatable virtual cell models on this foundation.

Further advancing from AIVC to virtual organs, Patient-Level Digital Twin System will be the true endgame for AIVC in the future. Xellar Biosystems’s internal strategy holds that simulating single cells or individual cellular states offers limited value; however, if it can simulate human systemic responses under disease conditions—particularly complex system behaviors such as PK/PD, barrier permeability, and immune responses—its commercial value will be amplified exponentially.

At its core, AIVC still faces a self-validating proposition similar to that of organoids and organ-on-a-chip systems: ensuring the consistency, validity, and high fidelity of model results with real human data. It is precisely for this reason that pioneers possess a sufficient window of opportunity to establish their irreplaceability in the AI era, leveraging existing high-quality data and automated systems.

Looking back at the history of innovative drug development, AI-driven Virtual Clinical Trials (AIVC) may relate to organoids and organ-on-a-chip technologies much as these latter models relate to animal testing. As the “steam engine” and “power loom” continue to iterate, new standards and application boundaries are constantly expanding. With the rise of AI as a new productive force, the narrative paradigm of new drug R&D will not be completely rewritten, but its storyline will extend, sketching out a blurred blueprint for the next phase.

The clear vision of the future remains to be written by those of today.