RNA interference took Alnylam roughly fifteen years to walk from its first big-pharma partnership to its first approved drug. The Merck alliances came in 2003 and 2004; Onpattro, the world’s first RNAi therapeutic, did not clear the FDA until August 2018. In between sat a Novartis deal, a Takeda alliance, a Sanofi readout, patent litigation with Dicerna, and the slow grind of figuring out how to ferry a fragile double-stranded RNA molecule into a human liver cell without it being destroyed on the way. Hold that fifteen-year number next to the headline on Alnylam’s new deal.

On June 3 the Cambridge biotech announced a three-year collaboration with Inceptive Nucleics that STAT and Investing.com both pegged at “up to $2 billion.” The number that actually changed hands is $30 million, in a mix of cash and equity in the startup. The rest is a wishlist of milestone payments tied to preclinical hits, regulatory approvals, and commercial sales: the standard biotech-deal scaffolding that lets a press release shout a billion-dollar headline while the receipts read more like a small Series A. CEO Yvonne Greenstreet said the company is “thrilled to partner with Inceptive to push the boundaries of what is possible in the discovery of RNAi medicines.” That is a CEO doing a CEO’s job. It is not the news.

The news is what the AI is being hired to do, and what it is not. Inceptive’s models are not being asked to invent a new modality. They are being asked to do faster and better what Alnylam has already spent two decades industrializing.

That framing matters because it is the inverse of the pitch that has run the AI-drug-discovery sector for a decade. Inceptive was founded in 2021 by Jakob Uszkoreit, one of the eight Google Brain researchers who wrote “Attention Is All You Need,” the 2017 paper that gave the world the transformer architecture underneath ChatGPT and every other large language model on the market. His co-founder is Stanford biochemist Rhiju Das. The pitch is that the same family of models that learned the structure of human language can learn the structure of RNA molecules, and that the bottleneck in modern RNA drug design is not chemistry but combinatorics. In September 2023, Andreessen Horowitz and Nvidia anchored a $100 million round into the company on that thesis. The Alnylam deal is its first major pharma validation.

What Inceptive will actually do, per the announcement, is chew through more than twenty years of Alnylam’s proprietary siRNA design data and use it to predict which candidate molecules to push into preclinical work. The models will help design the small interfering RNA sequences themselves, model their target messenger RNAs, and explore the chemical modifications that decide whether an siRNA stays potent inside a real cell. That is the same chemistry platform Nature Biotechnology profiled in 2018 when Onpattro crossed the finish line, and the one Toxicologic Pathology documented the GalNAc-conjugate safety profile for in a 2018 review funded by the company. The AI is being grafted onto a working machine, not asked to conjure one.

The reason that distinction is load-bearing is the rest of the sector. AI drug discovery is now about a $1.4 billion annual market, and by pharmaphorum’s count it has produced zero approved, marketed drugs. Recursion, Exscientia, and BenevolentAI were the three best-funded pure plays and the first three to actually read out clinical data. All three came up negative. Recursion has since shelved three programs and merged with Exscientia in a cost-cutting move. BenevolentAI’s lead atopic dermatitis candidate hit its safety endpoint and failed to beat placebo on efficacy, and the company has been laying off staff and retreating into early-stage partnership work. The pattern is consistent enough to call it: AI is genuinely useful for narrowing the candidate pool at the front end, and it keeps colliding with the deeper biology that decides whether a candidate becomes a drug.

Alnylam is a different kind of buyer, and that is the whole angle. This is not a venture-stage company hoping AI will conjure a pipeline out of nothing. Alnylam reported $3.7 billion in total revenue in 2025, Amvuttra alone generated $2.31 billion (up 138 percent year over year), and 2026 guidance for combined product sales runs between $4.9 billion and $5.3 billion. Four drugs are directly producing that product revenue, with partnered and royalty-bearing products including the Novartis cholesterol injection Leqvio on top. The European Journal of Internal Medicine published an Alnylam-funded review of the ATTR amyloidosis franchise on June 1, two days before the deal. The point of the fact stack is not the money. It is that this buyer has the wet-lab infrastructure, the regulatory experience, and the cash flow to actually find out whether AI candidates work in patients, which is the test the rest of the sector has not been able to run.

The thing to watch is whether the milestone payments ever actually trigger. Preclinical is the easy bar, and it is likely the one the $2 billion headline is being silently anchored to. The regulatory and commercial-sales tiers are where the sector has, so far, no proof. Alnylam has been signing partnership announcements like this one since 2003. Most of them never produced a drug. The ones that did took the better part of two decades and a small mountain of conventional medicinal chemistry to deliver. The Inceptive deal will be measured against that arc, not against the press release.

Sources

  1. STAT News – Alnylam to partner with Inceptive Nucleics for AI foundation models for RNAi therapeutics (2026)
  2. Investing.com – Alnylam, Inceptive form AI drug discovery collaboration (2026)
  3. European Journal of Internal Medicine – Alnylam-funded review of the ATTR amyloidosis RNAi franchise (2026)
  4. pharmaphorum – Recursion, Exscientia, and AI drug discovery’s moment of truth (2024)
  5. STAT News – Is this the beginning of the AI-in-drug-discovery era, or the beginning of the end? (2024)