DNAstack
The most valuable layer of the AI stack isn’t models or compute—it’s data. We need enduring infrastructure to govern and protect health data as personal, institutional, and national assets.
By Marc Fiume
The AI model you are using today will be obsolete in months. So will the chips it was trained on. The most enduring layer in the AI stack is not the model or the hardware. It’s the data.
Take your genome, the sequence of 3 billion letters that forms your molecular blueprint. It is the one that you were born with, and the one you will die with. It tells the history of your ancestry and predicts the future of your health. It is shared with your parents, siblings, children, and relatives.
Model weights will change. Chips will be replaced. Your DNA will not.
Last month, Canada launched AI for All, a national AI strategy built on a simple but important idea: prosperity in the AI era will depend not only on world-class models and compute, but also on sovereign infrastructure, trusted data, and the ability to translate AI into better outcomes for Canadians. When Prime Minister Mark Carney announced the strategy from the atrium of Toronto General Hospital, home to Canada’s largest hospital-based research program and the hospital consistently ranked among the world’s best, he was making a statement. Canada’s AI opportunity is not only to build better models, but to transform the future of health.
An AI strategy principled in sovereignty must design for it at every layer: models, chips, and most importantly, data. And nowhere is data sovereignty more important than in health. Canadian health data is longitudinal, universally covered, ethno-racially diverse, and tied to real-world outcomes. The AI models that train on it will reflect the full diversity of our people and therefore be more effective at home and more representative abroad. It is a strategic national asset that must be protected on the order of our natural resources.
Canada does not suffer from a lack of health data. It suffers from fragmentation. Health Canada reported that only 29% of primary care providers in Canada share patient information outside their practice. Fax machines still persist. Patients carry printed records between care settings. That is not just inefficiency. It is a massive missed opportunity.
The solution is a national health data network. One that simultaneously connects information across provinces while enabling privacy-preserving AI-powered insights to generate transformative health and economic outcomes for millions of Canadians. The system must expect genome sequencing instruments, AI models, and chips to change, and enforce best practices in privacy, security, governance, and interoperability over the data that will not. This is the system we need to own to prosper in the AI era.
My grandfather immigrated from Italy to Canada in the 1950s, one of millions our country welcomed after World War II. He helped construct the Canadian National Railway, a network of steel and timber that connected goods and people across the country, enabling cities to grow and powering economies of that era. Today, we need a different network, a digital one that connects our health data and turns it into actionable intelligence. In the era of AI, our country’s value will no longer travel over rails. It will travel over wires.
Digital networks do for institutions what the railway did for cities: they let them share value across vast distances while enabling them to keep autonomy, create partnerships, and thrive in new economies as a country instead of disconnected silos. So, whether you are a patient in St. John’s, Whitehorse, Kamloops, or Toronto, your care can be guided by the same expert models, trained on the collective experiences of the entire country. A national health data network makes better care available to everyone regardless of postal code. This is AI for All.
This architecture is not theoretical. Sovereign, AI-powered networks built by Canadian companies already enable some of the most ambitious precision health initiatives in the world. They connect datasets across institutions and borders without moving them, unlocking discoveries otherwise hidden in silos. The building blocks exist, and they are Canadian: Cohere in sovereign models, PhenoTips in clinical data capture, Smile Digital Health in health data interoperability, and DNAstack in network intelligence. Canada has spent decades producing world-class AI and health technologies. It is time to turn those innovations into infrastructure.
The risk of relying on foreign-controlled services is real. Weeks ago, the US government compelled Anthropic to shut down access to Fable, its most powerful AI model, to foreign nationals. Without the ability to discern users, the model was disabled globally. Users relying on the service lost access overnight, without notice — a stark reminder that foreign service providers are under the control of political powers we do not get a vote in. Sequencing manufacturers, model developers, and cloud providers offer vertical solutions that bundle the data layer with the products they sell, creating vendor lock-in that limits flexibility as these technologies mature. Good design separates enduring data assets from the modular tools around them, keeping the data sovereign while sequencers, models, and cloud platforms compete and commoditize.
Canada is already generating health data assets too valuable to cede. The Canadian Precision Health Initiative, led by Genome Canada, is sequencing more than 100,000 genomes to build the country’s first population-scale genomic resource. For assets like this to compound, Genome Canada and others must be able to plug into a national network without ceding governance or control. The next wave of breakthroughs will not come from any one institution, but from every institution connected, each contributing what it knows, none giving up what it holds.
The most valuable layer of the AI stack is not models or chips. It is data. For Canada, our universal healthcare system and ethnically diverse population represents a tremendous and timely opportunity to drive transformational health and economic benefits for Canadians. Establishing a sovereign national health data network is how we seize it. We have the technology. What we need now is the national conviction to treat it as essential.the technology. What we need now is the national conviction to treat it as essential.