Pomegra
Toronto's Biossil has raised $70M USD from Founders Fund and OpenAI across two rounds to acquire and repurpose discarded clinical-stage drug molecules using large language models.
Key Takeaways:
- Founders Fund led a $22M seed in 2024, then co-led a $43M follow-on in 2025 with OpenAI; total raised is $70M USD at a $100M+ valuation.
- Biossil screens public trial data, filings, and research with LLMs to identify failed molecules worth acquiring - bypassing preclinical work that can take seven years.
- The company's ten-candidate pipeline includes two drugs in advanced clinical trials and three approaching regulatory submission, targeting Alzheimer's and sickle cell disease.
Lead
Biossil, a Toronto biotech founded in March 2023, disclosed $70 million USD in equity funding on April 23, 2026, ending three years of stealth. Founders Fund led a $22 million round in 2024, then co-led a $43 million follow-on in 2025 alongside the OpenAI Startup Fund. The company is now valued at more than $100 million. With that capital, it has bought or licensed 10 drug candidates that major pharmaceutical companies had abandoned - and already moved two into advanced clinical trials.
Co-founders Anthony Mouchantaf, a former head of Royal Bank of Canada's venture capital arm, and Dr. Alexander Mosa, a trained internal medicine specialist, built the company around a straightforward observation: roughly 90 percent of drugs entering human trials ultimately fail, and nearly all of those molecules are simply shelved rather than retested.
What Does Biossil's AI Platform Actually Do?
The platform reads publicly available data about clinical-stage compounds - trial results, press releases, securities filings, academic papers - and passes that text through large language models to generate structured descriptions of each molecule's properties. Those descriptions are converted into numerical vectors using an embedding model, which maps proximity between compounds. The closer two molecules sit in that vector space, the more they share in biological function or mechanism.
From there, Biossil identifies candidates that failed for reasons unrelated to their underlying chemistry: a trial designed for the wrong patient population, a developer that ran out of capital, a strategic pivot that had nothing to do with efficacy. The company then approaches the original owner to acquire or license the asset, inheriting whatever clinical data already exists.
The economic logic is compelling on paper. A molecule that cleared Phase 1 safety testing has already survived the stage with the highest attrition rate. Biossil skips four to seven years of preclinical and early human work, along with the costs that accompany it - often hundreds of millions of dollars in a conventional program. What it accepts in return is reputational baggage: every asset carries a prior failure, which shapes partner negotiations, investor conversations, and eventually regulatory submissions.
How Big Is the Market for Discarded Molecules?
Large pharmaceutical companies discontinue clinical programs constantly, often for reasons that have no bearing on a molecule's biology. A patent set to expire, a pipeline restructuring, a commercial forecast that no longer justifies the spend - each of these produces an abandoned asset with usable data attached. Drug repurposing has existed as a field for decades, most prominently in rare diseases where small patient populations drove original developers away. The systematic application of LLMs to screen this pool at scale, however, is a recent development.
Competition is forming around adjacent theses. Several well-funded AI drug discovery companies generate entirely new molecules from scratch rather than recycling existing ones. Biossil's model differs structurally - it pays acquisition costs upfront but avoids synthesis and early-human-testing costs downstream. The supply of high-quality failed assets is not unlimited, though, and prices for premium candidates are likely to rise as more buyers enter the market.
Pipeline and Recent Acquisitions
The ten-candidate portfolio spans multiple therapeutic areas. Two drugs are in advanced clinical trials. Three others are approaching submissions for conditional market approval. Disclosed targets include sickle cell disease and Alzheimer's disease, both of which have histories of expensive late-stage failures from major developers - and both of which represent large addressable patient populations.
The company's acquisition activity has already extended beyond its own capital base. Summit Therapeutics sold a failed antibiotic to Biossil in a transaction valued at $105 million, indicating that established public biotechs are willing to treat Biossil as a serious buyer rather than an academic repurposing shop.
The OpenAI Startup Fund's role as co-lead - rather than a small strategic check - reflects both the depth of the technical relationship and the fund's confidence in applied LLM use cases outside software. Biossil built its screening platform directly on OpenAI's models, making the investment a bet on a significant customer as much as a standalone biotech thesis.
Outlook
Biossil now carries a credible investor roster, a pipeline that has advanced faster than most three-year-old biotechs manage, and a thesis legible enough to explain in two sentences. The clinical stage is where the argument gets tested. Conditional approvals for three candidates in quick succession would validate the LLM-screening approach more convincingly than any funding round. Failures in that cohort - especially if concentrated among the company's highest-profile indications - would force harder questions about whether the platform identifies genuine second chances or simply packages expensive retreads. Investors who committed across two pre-public rounds have already made their bet. The market will now keep score.
