Eli Lilly has committed up to $3.4 billion across two milestone only partnerships for recombinase based gene editing technology, betting that AI can finally solve the engineering challenge that stymied researchers for decades. Recombinases offer cleaner, more precise gene…
Eli Lilly will pay nothing upfront. The $2.25 billion comes only if programs advance.
That milestone-only structure is the most honest thing about the deal Eli Lilly announced last week with Profluent AI, and it tells you more than the headline number. What Lilly is buying access to — a class of gene-editing enzymes called recombinases — has been the technology researchers wanted but could not fully control for decades. The milestone structure means Lilly is not paying for a product. It is paying for an option on something that may not exist yet.
Recombinases are enzymes that can slip entire functional genes into a genome at a precise location, essentially pasting new DNA rather than cutting and hoping the cell's repair machinery does the rest. CRISPR, the gene editor most people have heard of, takes the second approach: it cuts, and the cell fills the gap. Inserting is cleaner. It was also always harder to handcraft. Researchers could point CRISPR at a spot and mostly hit the right target. Recombinases offered better accuracy but resisted the engineering that makes a tool practical — small changes to their sequence often broke them entirely.
AI has changed that calculus. Profluent, a four-year-old startup founded in 2022 by CEO Ali Madani, built a system that designs novel proteins by learning from massive libraries of biological sequences, including natural recombinases that researchers could never get to work in practice. The company demonstrated the approach in mid-2025 with OpenCRISPR-1, an AI-designed CRISPR editor published in Nature that has since been used by thousands of labs worldwide, according to Profluent. The recombinase work has not been published yet. That gap is what the deal is meant to fill.
The Seamless connection makes the picture clearer. Lilly signed a separate recombinase partnership in January with Seamless Therapeutics, worth up to $1.12 billion and focused on hearing loss, regulatory filings show. Together the two agreements represent $3.37 billion in potential recombinase commitments in a single year, from a company that has said almost nothing publicly about its gene-editing strategy. No press release framed either deal as a platform shift. The commitments just appeared in regulatory filings, and most coverage treated them as routine licensing news.
That is the part worth sitting with. When a company the size of Lilly commits that kind of money across two separate partnerships for the same underlying capability, in the same year, without public announcement or investor-day framing, it has already decided something. The milestone-only structure is the clearest available signal of what it decided: the technology is real enough to bet on and unproven enough that Lilly will not pay for it upfront.
What makes the bet credible, and also risky, is that Profluent's recombinase results have not been peer reviewed. OpenCRISPR-1 established the company's credentials for CRISPR editing. The recombinase work is at the same stage that CRISPR editing was two years ago — interesting, plausible, unpublished. Whether Profluent can replicate with a different class of enzyme what it achieved with CRISPR is the question the $2.25 billion in milestones is priced around.
The broader implication is harder to miss. If recombinases become programmable — if AI can reliably design versions that land where researchers want them — the addressable range of gene therapy expands. Today's approved gene therapies target single-gene disorders where a broken copy can be replaced: certain blood diseases, a handful of inherited eye conditions, spinal muscular atrophy. Whole-gene insertion through recombinases could in principle reach more complex diseases, where the problem is not a missing gene but a regulatory fault or a multi-gene issue. That is speculative in the near term. It is also the reason the milestone-only structure makes sense for both sides.
The peer-reviewed data will matter more than the press release. Watch for it.