BREAKING  Benchling AI ships: agents that read your data, write your notebook, clean your CRO files 300,000+ scientists run experiments on one platform 20 of the 50 largest biopharma firms are customers $6.1B valuation at Series F - reportedly a 122x revenue multiple 2025  AlphaFold, Chai-1 & Boltz-2 now live inside the lab workflow NVIDIA & Baseten partner on biotech AI inference BREAKING  Benchling AI ships: agents that read your data, write your notebook, clean your CRO files 300,000+ scientists run experiments on one platform 20 of the 50 largest biopharma firms are customers $6.1B valuation at Series F - reportedly a 122x revenue multiple 2025  AlphaFold, Chai-1 & Boltz-2 now live inside the lab workflow NVIDIA & Baseten partner on biotech AI inference
YesPress Dossier · Company

Benchling.

The cloud where biology gets its work done. One system of record for designing experiments, tracking samples, and turning lab chaos into data scientists can actually trust.

San Francisco · Founded 2012 · B2B SaaS for life science R&D

Benchling logo

The logo for a company that decided the 400-year-old paper lab notebook had run its course.

Who they are now

The lab runs on it

Somewhere right now, a scientist at one of the world's largest drug companies is registering an antibody, pulling up a protein-structure prediction, and writing it all into the same window. No file shuffling. No emailing a spreadsheet to a colleague three time zones away. That window is Benchling.

More than 300,000 scientists open it. Over a thousand R&D organizations depend on it, including 20 of the 50 largest biopharma companies on earth. Gilead, Regeneron, Sanofi, Eli Lilly, Corteva - they keep their science in the same place a two-person startup in a shared incubator does. Benchling is, for a meaningful slice of modern biology, the operating system. It is the boring, load-bearing infrastructure that science quietly assumed it always had. It did not.

For a field obsessed with the cutting edge, biology spent a remarkably long time running on paper and Excel.
The problem they saw

A 400-year-old notebook

Here is the uncomfortable thing about breakthrough science: much of it was recorded the way Galileo recorded his - in a bound paper book, by hand. Sequences lived in one tool, samples in another, results in a fourth, and the connective tissue between them lived in someone's memory. When that someone left, so did the context.

The cost of that mess is not abstract. Experiments get repeated because nobody could find the first result. Regulatory submissions stall because the data trail has gaps. Discoveries slow down. Biology, the discipline most capable of changing human life, was being held back by software that had never been built for it - because, mostly, it had never been built at all.

Every other technical field got purpose-built software. Biology got a spreadsheet and a wish.

The deeper trap was specificity. Generic note-taking tools do not understand a plasmid. A document app does not know that a sample has a parent, a lineage, a freezer location. Biology needed software that spoke biology - and almost nobody wanted to build something that narrow and that hard.

The founders' bet

The GitHub of biotech

In 2012, Sajith Wickramasekara was a computer science student at MIT who had wandered into a wet lab and come away faintly horrified. He was a software engineer by training, and the tools the scientists around him used looked, to his eyes, like a crime scene. So he built something to help them edit DNA sequences. Then his co-founder Ashu Singhal joined, and the something started to grow.

Their pitch was almost arrogant in its simplicity: be the GitHub of biotech. Version control, collaboration, and a single source of truth - but for living things. In 2012 that sounded like a category error. Biology is wet and messy; software is clean and abstract. Investors could be forgiven for blinking.

They did not set out to build features. They set out to fix the tools they had personally found broken.

The bet was that biology would inevitably become a data discipline, and that whoever owned the system of record would own something enormous. It was a patient bet. Benchling spent years going deep on the unglamorous parts - sample lineage, molecular biology primitives, the schema of an experiment - precisely the work most software companies avoid because it does not demo well at a party.

The product

Seven apps, one source of truth

What started as a DNA-editing utility became a connected R&D Cloud. The trick was never any single app - it was that they share one biological backbone. A sample registered in Inventory is the same object referenced in the Notebook and analyzed in Insights. The data does not get copied, exported, or lost in translation. It just connects.

Notebook

A cloud electronic lab notebook with full data traceability - the paper book, retired.

Molecular Biology

Design and annotate DNA, RNA, and protein sequences; clone and model in-browser.

Registry

Links biological entities and physical samples to the experiments that produced them.

Inventory

Sample and reagent tracking with full lineage, down to the freezer shelf.

Workflows & Requests

Routes tasks, requests, and instrument data through the R&D lifecycle.

Insights

Interactive dashboards over harmonized data - reporting that does not require a data team.

Then, in October 2025, the obvious next move: Benchling AI. Instead of asking scientists to export their data into some disconnected chatbot, Benchling brought the models to the data. Three agents arrived - a Deep Research agent that reads internal experiments and public literature to answer hard questions; a Compose agent that turns scattered notes into clean notebook entries; and a Data Entry agent that converts messy CRO files into structured data. Alongside them came frontier scientific models - AlphaFold, Chai-1, Boltz-2 - living in the same window where the molecule is designed.

Most companies bolt AI onto the side. Benchling put it where the science already happens.

Above: the rare enterprise software whose hardest feature is understanding what a plasmid actually is.

How a side project became infrastructure

// Benchling, the short version

2012
Founded at MIT. Sajith Wickramasekara builds a DNA-editing tool; Ashu Singhal joins. The pitch: the GitHub of biotech.
2015
Goes enterprise. The product pivots from utility to platform and starts winning serious biotech customers.
2017
Registry & Inventory ship. Physical samples get linked to experimental records - the R&D Cloud takes shape.
2020
$50M Series D. ARR doubles for the fourth straight year, even mid-pandemic.
2021
$200M Series E, then $100M Series F. Valuation hits $6.1B - reportedly a 122x revenue multiple.
2025
Benchling AI & Biologics launch. Agents and frontier models go live; NVIDIA and Stanford spinout GXL sign on.
2026
Benchling Inference. A Baseten partnership gives customers on-demand GPU power for scientific models.
The proof

The numbers stopped being a pitch

Skepticism is fair. Plenty of companies claim to be infrastructure. Benchling's claim is unusually hard to wave away, because the adoption is concentrated exactly where it is hardest to win: the largest, most regulated, most risk-averse drug companies in the world. You do not casually migrate a pharma giant's R&D records. They did.

300K+
scientists on the platform
1,000+
R&D organizations
$6.1B
valuation (2021)
$412M
total raised

Annual recurring revenue, doubling on repeat

// illustrative trajectory - ARR roughly doubled four years running, per company statements

1x
Y1
2x
Y2
4x
Y3
8x
Y4
~16x
Y5

Relative scale, not absolute dollars. The shape is the point: compounding adoption is what convinced investors a biology software company was worth a 122x multiple.

The 2021 financing tells the same story in capital. Franklin Templeton and Altimeter co-led the $100M Series F; Tiger Global and Lone Pine piled in. A 122x trailing revenue multiple is the kind of number that either ages into a legend or a cautionary tale. Either way, it was a vote that the system of record for biology would be very large.

You do not migrate a pharma giant's entire research record on a whim. Twenty of the fifty biggest did.

The company it keeps

Partnerships now read like a who's who of the AI stack. NVIDIA is piping its NIM microservices - including the OpenFold2 protein model - directly into Benchling AI. Baseten supplies the GPU muscle behind Benchling Inference. GXL, a Stanford spinout, is building expert agents that run natively inside the platform. SciBite aligns captured data to enterprise ontologies at the moment of entry. Benchling is no longer just a notebook; it is becoming the place the rest of biotech's tooling plugs into.

The mission

Speed, but for medicine

Strip away the funding rounds and the mission is plainly stated: unlock the power of biotechnology and accelerate the pace of life science R&D. That is not a slogan you frame in a lobby. When a cell therapy or a vaccine moves through development faster because its data never got lost, the abstraction becomes a patient.

Notably, Benchling kept a free tier for academics and individual researchers even as it signed eight-figure enterprise contracts. The scientist-first posture is not charity; it is strategy. Today's grad student is tomorrow's principal investigator, and the tool you learned biology on is the tool you bring to the company you join.

Accelerating R&D sounds like a metric. Eventually it is a treatment that arrives a year sooner.
Why it matters tomorrow

The data was the moat all along

AI is only as good as the data it stands on, and biology's data has historically been a swamp - unstructured, scattered, untrustworthy. Benchling spent a decade quietly draining that swamp, one schema at a time. That groundwork is now the most valuable thing it owns. The models everyone is excited about are useless without clean, connected, biological data to point them at. Benchling has been collecting exactly that since 2012.

Which brings us back to that scientist, still at her desk. A decade ago she would have had a paper notebook, a folder of CSVs, and a colleague she had to interrupt to find last month's result. Today she registers the antibody, the structure prediction renders beside it, an agent drafts the writeup, and every byte of it stays connected and findable forever. The window did not just get prettier. The work underneath it got faster, and a little less likely to be lost.

Benchling's real product was never the notebook. It was making sure nothing a scientist learns ever disappears again.

Below: links, social profiles, and a few videos for the curious and the still-unconvinced.