The bottleneck in drug discovery is often a simple, physical one. A chemist sketches a promising molecule on a whiteboard, but getting a physical sample to test can take a contract research organization (CRO) six weeks or more. Onepot.ai, a San Francisco startup that emerged from stealth last November, is betting that an AI chemist and a robotic lab can shrink that wait to about five days [C&EN, November 2025]. It is a bet on turning the whiteboard sketch into a vial on the lab bench, and it has convinced a slate of high-caliber investors to back the physics of it.
The AI Chemist and the Pot
The company’s system is built on two components. The first is Phil, an AI model trained on in-house experimental data and literature to plan chemical reactions. The second is POT-1, the robotic lab that executes them. A researcher submits a target molecule from Onepot’s enumerated catalog of 3.4 billion compounds; Phil devises a synthesis path, and POT-1’s robotic arms get to work [Onepot.ai research, 2026]. The platform currently supports staple reactions like Suzuki-Miyaura coupling and amide bond formation, with more in beta testing [Onepot.ai onePot_core.pdf, 2026]. The output is shipped as a dry compound or in solution, aiming to slot directly into a partner’s high-throughput screening workflow.
The founding story is a classic one of frustration meeting capability. Co-founders Daniil Boiko and Andrei Tyrin met in a chemistry lab at M.V. Lomonosov Moscow State University and later built their first automated lab in March 2024 [C&EN, November 2025]. Boiko’s PhD work at Carnegie Mellon was central to Coscientist, an AI system that autonomously designs and executes complex experiments [CMU Engineering, December 2023]. That pedigree in autonomous science provided the technical validation that attracted a notable seed round.
Why Top-Tier Investors Wrote the Check
In November 2025, Onepot.ai announced a $13 million seed round led by Fifty Years. The investor list reads like a who’s who of deep tech conviction: Khosla Ventures, Speedinvest, NAVEC, Norrsken VC, and angels including Google’s Jeff Dean and OpenAI co-founder Wojciech Zaremba [C&EN, November 2025]. The table below outlines the key backers.
| Investor | Type | Notable For |
|---|---|---|
| Fifty Years | Lead VC | Early-stage deep tech & science bets |
| Khosla Ventures | VC | Frontier technology and hard science |
| Jeff Dean | Angel | Google AI leadership |
| Wojciech Zaremba | Angel | OpenAI co-founder |
| Speedinvest | VC | European early-stage focus |
| Norrsken VC | VC | Impact-focused deep tech |
This capital is earmarked for expanding the team and a second lab facility, aiming to turn the prototype into a reliable, scaled service [Yahoo Finance, November 2025]. For investors, the appeal is twofold: automating a critical, time-sucking step in the multi-billion dollar drug discovery pipeline, and rebuilding elements of small-molecule synthesis capacity in the United States.
The Incumbent Mountain to Climb
The ambition is clear, but the path runs directly through established giants. The primary competitive pressure comes from massive, global CROs like WuXi AppTec and Enamine. Their advantages are not trivial. They offer immense scale, decades of process expertise, and deeply entrenched relationships with every major pharma company. Onepot’s wedge is speed and, potentially, cost for early-stage discovery work. Yet, the risks for a young company are substantial.
- Process breadth. While Onepot supports several common reactions, drug discovery often requires exotic chemistries. The platform’s utility is limited to the reactions it can reliably automate, a catalog that must expand significantly.
- Commercial proof. The company has not yet named any commercial customers or disclosed traction metrics. Moving from pilot tests with partner biotechs to paid, recurring enterprise contracts is a different synthesis problem altogether.
- Operational scaling. Running one robotic lab for prototypes is one thing. Operating a fleet of them to meet commercial demand, with consistent yield and purity, is a formidable engineering and logistics challenge.
The bet, then, is that a 5-day turnaround is not just a nice-to-have but a fundamental accelerator for the iterative design-make-test-analyze (DMTA) cycles of modern drug discovery. If it works, it could compress early-stage discovery timelines by months.
Doing a back-of-the-envelope calculation: if a traditional CRO takes 45 days to deliver a compound and Onepot takes 5, that’s a 40-day saving per cycle. In a campaign requiring ten iterative cycles, that’s 400 days, or over a year, shaved off the critical path to a preclinical candidate. For a biotech startup burning $5 million a month, that time saved is not just speed, it is survival capital. To capture that value, Onepot must prove it can reliably beat the clockwork of a WuXi AppTec, not just in a demo, but on the loading dock of a frantic discovery lab every single week.
Sources
- [C&EN, November 2025] Molecule maker OnepotAI launches with $13M | https://cen.acs.org/business/start-ups/Molecule-maker-OnepotAI-launches-13/103/web/2025/11
- [Yahoo Finance, November 2025] Onepot AI raises $13M to help make chemical drug creation easier | https://finance.yahoo.com/news/onepot-ai-raises-13m-help-150000701.html
- [Onepot.ai research, 2026] onepot CORE, an enumerated chemical space | https://www.onepot.ai/research
- [Onepot.ai onePot_core.pdf, 2026] 1 onepot CORE, an enumerated chemical space | https://www.onepot.ai/onepot_core.pdf
- [CMU Engineering, December 2023] AI Coscientist automates scientific discovery | https://engineering.cmu.edu/news-events/news/2023/12/20-ai-coscientist.html