There’s a lot of “AI in healthcare” floating around right now, and most of it looks alike: a chatbot bolted onto an existing product, trained on the same public text everyone else can reach. Some of these tools are genuinely useful. They summarize notes, answer questions, take the friction out of paperwork. But they share a common weakness. If an AI is built on data anyone can get, then anyone can build it. There’s no real differentiation, and no new biology being discovered along the way. One investor who helped build a multibillion-dollar healthcare-AI company put it plainly: the most valuable healthcare AI platforms will be built on proprietary biological data. This concept separates AI that talks about health from AI that actually measures it.
Two very different kinds of healthcare AI
It helps to split the field in two.
First, you have interface AI and software that makes existing information easier to handle. Chatbots, scribes, search. It runs on top of data that already exists and doesn’t produce any new biological insight of its own. The value is in the convenience.
Second, you have measurement AI. Models trained on biological data to reveal things we couldn’t see before. These don’t reorganize what we already know; they pull new signals directly out of the body. This is the category that can catch disease earlier, predict who’ll respond to a treatment, or put a number on how fast someone is aging. This is Infinite Epigenetics.
Both are worth having. But only one is hard to copy, and only one generates genuinely new medical knowledge.
Why biological data is the moat
In AI, the architecture is rarely the lasting advantage. The methods get published, and today’s frontier technique is next year’s open-source library. What’s hard to replicate is the data, and biological data most of all, because you can’t scrape it off the internet. It has to be generated one real human sample at a time, in a certified lab, usually over years.
That’s a different kind of asset altogether. A competitor can copy an algorithm in an afternoon. They can’t copy years of sample collection, the consent and data agreements behind it, the lab know-how to read the signal cleanly, or the depth that only comes from following the same biology over time.
It’s the difference between “we added AI” and “we built something only we can build.” The Infinite Epigenetics platform is anchored on one of the world’s largest DNA-methylation datasets: more than 120,000 samples, about a million signals read per sample, backed by 50-plus peer-reviewed publications, 80-plus research partnerships, and more than a thousand proprietary algorithms built on biology no one else has access to.
The flywheel that widens the gap
Proprietary data also compounds over time. Every test run on the platform reads a fresh set of signals from a new person, and every sample becomes new training data. More tests produce more data, more data trains a sharper model, a sharper model delivers deeper insights, deeper insights pull in more adoption, and more adoption produces still more tests.
That self-reinforcing loop, a data flywheel, means the distance between a data-native platform and a fast follower doesn’t hold steady. It grows. The platform gets more valuable, and more defensible, the more it’s used. A chatbot built on public data has nothing like it. It’s as good on day one as it’ll ever be, and so is everyone else’s.
The precedents already exist
The Infinite platform isn’t a hypothetical. Our epigenetic testing is already in the market today (120,000+ tests and counting), and three multi-billion dollar companies have validated each piece of the category thesis.
Exact Sciences, a molecular-diagnostics pioneer, showed that a business built around a single chronic-disease test could scale in the public markets. GRAIL, a multi-cancer detection company showed that methylation, the very signal at the heart of epigenetics, could power disease detection from blood. And Tempus, a precision-medicine leader, showed that proprietary clinical data paired with AI is itself a durable platform, not just a product.
Each proved one part of the model. Infinite Epigenetics, a biological-AI platform built on epigenetics, brings all three together: the same methylation signal, applied across multiple chronic diseases, powered by a proprietary biological dataset and the foundation model trained on it.
What it means for medicine
The implication extends beyond business strategy. AI built on proprietary biological data also produces new science, well beyond any competitive moat. A model trained on a large, deep biological dataset can surface patterns nobody had the resolution to see before: earlier disease signatures, better ways to match patients to treatments, biomarkers that didn’t exist as concepts last year.
For us, this is the real prize. The most valuable healthcare AI won’t be the one with the slickest chat window. It’ll be the one that reads the body most deeply because it’s built on biology that took years to gather and can’t be downloaded by anyone else.
Adding a chatbot is easy. Building a proprietary biological dataset and a foundation model trained on it is extraordinarily hard, and that difficulty is the whole point. Interface AI organizes what we already know. Measurement AI, built on biology, expands it. The future belongs to the platforms that own the data underneath, which is not where the hype is loudest, but is where the value and the science actually compound.
Infinite Epigenetics is building a biological AI platform on one of the world’s largest proprietary methylation datasets, with operating companies TruDiagnostic (clinical) and Tally Health (consumer). The companies referenced are mentioned as category precedents only; no valuation or side-by-side claims are implied. Written for general education; not investment advice.