Funding & Investments

Vilnius enzyme designer Biomatter raises $18m Series A led by Astanor to build a lab and grow in the US

The Lithuanian company, which designs enzymes from the target reaction up, will spend the money on a 2,000 m² Vilnius facility opening in 2027, a larger sales team and its US business.

Author
John Pulisic
Published
Reading time
4 min read
Views
Loading views
Biomatter cover image

Biomatter, the Vilnius company that uses AI to design custom enzymes, has closed an $18m Series A led by Brussels-based Astanor. The round was oversubscribed, and existing backers UVC Partners, Inventure and Metaplanet all put in more money.

The size of the cheque matters less than where most of it is going. Instead of spending mainly on compute or headcount, Biomatter plans to open a 2,000 m² site in the Lithuanian capital in 2027, where enzymes will be both designed and tested. For a sector where many AI-for-biology start-ups still sell promises, the company is betting that physical lab capacity close to its models is what will set it apart.

The company did not disclose a valuation or how much it has raised in total.

Designing from the reaction, not from nature

Most industrial enzymes begin life as natural proteins that scientists then tune over many rounds of trial and error. Biomatter, founded in 2018 and previously known as Biomatter Designs, works the other way round. It starts with the chemical reaction a customer's product needs and designs a protein to perform it, using a platform it calls Intelligent Architecture that combines generative AI with physics-based modelling.

That approach depends on data. Biomatter says it has spent years producing its own measurements on enzymes that look nothing like those found in nature, and every customer programme adds more results to that proprietary set. Each design is checked in the lab before delivery. The founders, who include chief executive Laurynas Karpus alongside Vykintas Jauniškis and Irmantas Rokaitis, published early evidence in 2021 that generative models could produce working enzymes far removed from natural sequences.

The company points to one result it considers a proof of the method: a methylase for mRNA vaccine production that does not need S-adenosylmethionine, or SAM, a costly additive that natural methylases rely on. Biomatter describes this as a reaction that nature does not carry out at all.

“Our goal is to make a custom enzyme as easy to order as a catalog product, including for reactions that don’t have one today,” said Laurynas Karpus, CEO and co-founder of Biomatter.

Customers already buying

Biomatter's case to investors rests heavily on revenue-generating work. By its own count, it has delivered more than 50 custom enzymes to over 25 enterprise clients in Europe, the US and Japan. The names it lists include AstraZeneca, BASF, Sanofi, the Japanese group Kirin Holdings, Tereos, and Neogen, the US food safety firm it signed a product development arrangement with in August 2025. Some of those relationships, the company said, have grown from a single project into multi-year programmes covering several enzymes.

One of its designs is already on lab shelves. Biomatter worked with Thermo Fisher Scientific on the Invitrogen SuperScript V reverse transcriptase, an enzyme that converts RNA into DNA, improving its resistance to oxidation. That stability is part of what lets the reagent ship without cold storage and sit at room temperature for up to six months.

Earlier this year, in January, Biomatter also became the first Baltic company to join AstraZeneca's BioVentureHub, the drugmaker's start-up programme in Gothenburg.

The investors lean on this track record. Oliver Schoppe, a partner at UVC Partners, said the firm raised its stake because Biomatter's enzymes are already in products made by large manufacturers on both sides of the Atlantic. "Most companies working on AI for biology are still proving it works," he said.

Where the $18m goes

The Vilnius facility is the centrepiece. By moving more of the build-and-test cycle under one roof, Biomatter wants to shrink a design loop that it says already takes weeks, compared with the months or years typical of conventional enzyme programmes, down to days.

The rest of the money will fund a larger commercial team and a bigger presence in the US, where about half of Biomatter's customers already sit. It will also keep investing in its software platform and the enzyme dataset that feeds it.

The choice of lead investor says something about where Biomatter sees demand. Astanor, founded in 2017, has made around 50 investments and concentrates on agrifood and health. Harry Briggs, a partner at the firm, said the industry has long been limited to training models on what already exists in nature. "The constraint has been data," he said, adding that repeat business from large customers across several industries was what convinced Astanor.

A step up from the seed round

The Series A arrives a little over two years after Biomatter raised a €6.5m seed round in August 2024, which Inventure and UVC Partners led. Both firms returning in a round that was oversubscribed suggests insiders have seen enough in the intervening period to increase their exposure rather than leave the new round to outside money.

It also adds to a run of sizeable rounds for Lithuanian tech, and to a steady flow of European capital into engineered biology, from microbial chemistry to enzymes for recycling plastics.

The harder test is operational. Biomatter now has to build and staff a large lab in Vilnius, open it on schedule in 2027 and show that it can turn designs around in days, while serving a customer base that is half American from a base in the Baltics.

What did you think of this story?

Written by

John Pulisic

49 stories on European Startup Insider

Have a story like this? Pitch it to us.

We publish founder profiles, company news and practical guides from across Europe.

You’ve finished this story

Up next

Darmstadt's Daisytuner raises €6m seed to run AI models faster on chips beyond Nvidia

Elaia and LEA Ventures co-led a €6m seed round in Daisytuner, whose Daisyflow platform tunes AI models at inference time for whichever processors a company runs.