Novo Nordisk is moving a Parkinson’s cell therapy out of its own labs and into Cellular Intelligence, turning a shelved pharma program into a venture-backed test of AI-enabled biotech.
Novo Nordisk is not simply handing off an experimental Parkinson’s treatment. It is showing how a large drugmaker can keep a foot in a risky field without carrying the whole cost and complexity on its own balance sheet.
The Danish pharma company has licensed STEM-PD, a clinical-stage Parkinson’s cell therapy, to Cellular Intelligence, the Boston startup formerly known as Somite AI. The treatment uses donor stem cells that are transformed into new neurons and transplanted into the brain, with the goal of replacing the cells lost as Parkinson’s progresses. Cellular Intelligence will now lead development, including a planned phase 2 study by the end of 2026, while Novo receives an equity stake and remains eligible for milestones and royalties if the therapy advances.
That structure matters. According to Bloomberg, the move puts a Novo asset in the hands of a Zuckerberg-backed startup at a moment when pharma companies are under pressure to make pipelines more productive and less capital intensive. It is not a traditional clean exit, where a company walks away from a program entirely. It looks more like a controlled relocation of risk, with Novo preserving possible upside while a smaller company takes on execution.
The background is important because Novo did not stumble into this deal. Last year, the company pulled back from cell therapy work as part of a wider restructuring under Chief Executive Mike Doustdar, who has been pushing faster decisions after Novo lost momentum in obesity drugs to Eli Lilly. The Parkinson’s asset needed a new home if it was going to keep moving, and Cellular Intelligence had a clear pitch: its platform is built around using large datasets and AI models to understand how stem cells become specific cell types.
For a therapy like STEM-PD, the challenge is not only proving that the cells can work in patients. It is also making them consistently, at scale, and with enough quality control to satisfy regulators. Cell therapies have often looked powerful in theory and brutal in practice because manufacturing can be slow, expensive, and highly variable. That is exactly the kind of problem Cellular Intelligence says its technology is designed to attack.
The startup says its capsule-based system can generate cell-signaling data far more efficiently than older methods, giving its models more material to learn from. In plain terms, it is trying to replace some of the trial-and-error work in regenerative medicine with a more predictable engineering process. That does not make the clinical risk disappear. Parkinson’s remains a difficult disease, and brain cell transplantation is about as high-stakes as biotech gets. But it does explain why Novo would rather see the program tested inside a focused startup than buried inside a reorganized pharma division.
Why pharma keeps using startup vehicles
This is becoming a familiar pattern. Large drug companies have deep scientific libraries, half-developed assets, and research programs that may no longer fit the corporate priority list. Startups have narrower missions, venture capital, and a tolerance for focused risk that public pharma companies often struggle to maintain. When those two sides meet, a dormant asset can become a new company story.
Novo has already used a similar playbook. Aspect Biosystems took on some of Novo’s diabetes cell therapy work earlier this year, with Novo also investing in that company. The message is consistent: cell therapy may still be strategically interesting, but not every program needs to sit inside Novo’s own operating structure. For shareholders, that can be attractive because it reduces internal spending while preserving optionality. For startups, it can be transformative because a clinical-stage asset gives them a faster route to relevance than building everything from scratch.
The Zuckerberg connection adds another layer. Cellular Intelligence’s earlier Series A financing included the Chan Zuckerberg Initiative, alongside Khosla Ventures and other investors. That places the company inside a broader push to apply AI, large biological datasets, and foundation-model thinking to medicine. Mark Zuckerberg and Priscilla Chan have spent years funding work around cell science, disease biology, and virtual cell models, and the venture logic is now becoming clearer: if biology can be made more computable, the companies that own the best data engines may capture value far beyond a single drug candidate.
Still, the market should be careful not to confuse an elegant funding model with a guaranteed medical breakthrough. STEM-PD is still experimental. Cellular Intelligence must show that its AI platform can improve real manufacturing and clinical outcomes, not only produce better language around them. Novo, meanwhile, gets a relatively efficient way to stay connected to a field it has stepped back from, without asking investors to fund the full journey inside the parent company.
The bigger story is that pharma IP is becoming more mobile. Programs that once might have been written off can now be placed into startups, paired with specialist data platforms, and financed by investors willing to wait for a bigger scientific payoff. If STEM-PD progresses, Novo will have shown that a non-core asset can become a startup catalyst. If it fails, the loss is more contained. Either way, the deal points to a future where drug pipelines are less like sealed corporate vaults and more like networks of capital, data, and clinical bets moving to wherever they have the best chance to survive.
Also read: Oregon is making data centers pay more for the grid they need • Cowboy Space shows how far AI infrastructure bets have stretched • A viral textbook claim shows why AI content needs receipts