Hardware & Biology: Durable Moats Beyond Code

All through its last raise, a cultivated-meat start-up heard investors ask why it needed its own facility, and whether a low-capex model would do. "At some point, someone needs to put something in the ground which produces a product," the founder said. Most of the people at Max Lebeau and Isabel Zhang's Ripple were founders in bio-manufacturing and industrial biotech, and several said this kind of physical work is their defence now that AI makes software easy to copy. Whether investors will pay for it was less certain.

The same founder described the moment the company had no choice. The financial model worked on media costs, price points and operating costs, yet the business was not viable unless capital costs came down, because of depreciation and the sums it would have to raise. The founder had hoped someone else would build a cheap bioreactor. Nobody did, so the company designed its own. Outsourcing, the founder said, would only have trained a supplier and handed it IP the company should own.

Built in-house because nothing fits

Others told the same story. Everything for scale-up is built for pharma, one founder said, and no high-volume, low-margin industry will accept batch production with enzymes replaced by hand every day. So the company engineered and built its own process, and decided there was more value in it than in the product it started with. A founder scaling bacterial nanocellulose builds part of its system in-house because no pilot plant offers it, and keeps it as a trade secret.

A founder building photobioreactors for algae described where the defence sits. The patents cover the hardware, but the trade secrets lie where the biology, the hardware and the operating knowledge meet, and that knowledge will never be published. AI helps with discovery and engineering, the founder said, but biology does not always behave as the models predict. A founder in industrial biocatalysis added that AI is producing a flood of engineered enzymes, and none reaches industry without a hardware layer to deploy it.

Everyone knows Ginkgo; no one knows BioPhero.

Data from the hard part

The same work piles up data. A company that discovers and scales bacteria to break down PFAS in wastewater and soil has characterised its many strains genetically and by behaviour, and uses machine learning to predict the best combinations. The founder said it was less about clients asking for it than about positioning. Another participant separated two kinds of data. Data for designing products, such as new enzymes, is weak if everyone draws on the same public databases. Process data is different, because it shows investors that a company keeps de-risking and keeps getting closer to its cost targets.

One of the hosts wondered whether a library of pollutant-degrading bacteria could underpin a model licensed to companies outside the core market. The answers were cautious. Taking data forward one product at a time costs time and money, one participant said, and a founder building an AI platform for bio-based materials said clients guard their data hard. The worst case is a bacterial library reaching a competitor. A more modest trade already works. When one of the biggest sensor makers could not say whether its sensors would work in one start-up's fermentation, it supplied the expensive sensors in exchange for data.

The same old questions

One of the hosts asked whether investors now favour wet-lab and hardware work because software has lost its moat, or whether that was something the sector told itself. Isabel offered counter-examples from HCVC's own portfolio. Ark Biotech began with large bioreactors for cultivated meat, moved to mammalian cells for pharma and then to software, an operating system for bioprocesses, mainly after investor feedback. Automata, a lab-automation company, found its hardware hard to sell and built an operating system for lab experiments instead. A founder said the questions in a raise have not changed. Has next year's offtake been sold, and has the scale-up risk gone? Another said those pivots came from companies creating new markets, while a start-up selling into existing demand hears little pressure to pivot.

An early-stage investor blamed culture. Investors who once backed a website for finding a dog groomer have brought the same hunt for quick, cheap returns to hardware, and they will have to accept a different spread of outcomes. One of the hosts said funds that last seven or eight years plus two will not cut it, and that their own fund runs for 12 years plus three. Isabel said there have been no major IPOs outside pharma, and money to move from pilot to commercial plants is still scarce. The sector, Isabel said, has been having this conversation for five years. A participant named what is missing, a winner to point LPs at after Ginkgo, Zymergen and others faltered.

Max closed on the exits that do exist. BioPhero, which makes pheromones for forestry, sold for what Max thought was around 200 million, which is good if not great. K18 makes peptides for hair products. "Everyone knows Ginkgo; no one knows BioPhero," Max said. Such companies exist, but they take longer to build than most funds are designed to wait.

This Ripple was hosted by Max Lebeau (Positron Ventures) and Isabel Zhang (HCVC) at The Drop 2026 on 16 September.