Artificial intelligence has become increasingly important in medical research, but one major obstacle remains: access to high-quality biological data. Network Bio is attempting to address that problem by creating a connected network around patient tissue and blood samples held by academic medical institutions.

The Palo Alto-based startup has secured $50 million in funding and established collaborations with Mass General Brigham, the University of Pennsylvania and the University of Colorado Anschutz. Instead of building another standalone medical database, the company is developing infrastructure that allows researchers to access information generated from samples across participating institutions.

The underlying challenge is straightforward but difficult to solve. Hospitals and universities maintain biobanks containing biological samples donated for research. These collections can be valuable for understanding diseases, developing treatments and identifying differences between patients. However, the information and samples are often managed independently by individual institutions.

That fragmentation limits the scale of research. A single hospital may have an impressive collection, but it may not contain enough samples to answer complex questions about how diseases behave across different populations.

Network Bio's approach is designed around connecting these resources while maintaining safeguards around patient privacy. The data involved is de-identified, meaning researchers are not intended to receive information that directly identifies individual patients.

The timing is important. The cost of generating molecular data has fallen, while advances in AI have made it increasingly practical to analyze very large datasets. Together, these developments create an opportunity for researchers to ask questions that previously required enormous amounts of time and resources.

The company is particularly interested in diseases outside traditional oncology research. Cancer has benefited from extensive biobank development and precision-medicine programs, while conditions such as cardiovascular, autoimmune and inflammatory diseases have faced different data-access challenges.

Network Bio is also working with NVIDIA on a disease-spanning AI model and exploring relationships with pharmaceutical companies. The objective is not simply to create another dataset, but to establish a research infrastructure that can support multiple scientific applications.

For healthcare executives and investors, the development points toward a broader shift in the health-tech industry. The competitive advantage of AI may increasingly depend less on access to generic computing power and more on access to specialized, high-quality domain data.

That could make data infrastructure a strategic layer of healthcare innovation. If hospitals, researchers, technology companies and pharmaceutical organizations can collaborate around secure data networks, AI systems could gain access to information that individual institutions cannot provide alone.

Network Bio's model is therefore about more than connecting biobanks. It represents an attempt to solve one of the less visible challenges behind medical AI: creating the data foundation required to make sophisticated research systems useful at scale.