“Designing new things isn't really the struggle. Coming up with a new idea, a new molecule, a new product, mining that idea space isn't the challenge that needs to be solved. What actually is the barrier is getting things into market. … In chemicals and materials, it's 10 years and it's 100 to 300, 400 million dollars. … The bottleneck to that progress is not the discovery piece, it's the scale-up piece.”
Grow Everything Biotech Podcast · Alternative Proteins · October 2026
Context
The hosts asked where AI speeds up bioprocessing and where it runs into real-world physics. Kufner, who runs a company turning waste CO2 into chemicals, argues the hard part is moving from the lab to large-scale production, not coming up with ideas.
Transcript
Very cool. Definitely.
…those designs don't take the physical economic manufacturing aspect into account. And it makes it very difficult if you don't have real world data. So we want to know from you, where does AI generally accelerate bioprocessing and where does it completely fall down when hit with real world physics?
…And all that stuff takes decades and costs hundreds of millions of dollars. I mean, the stat that I kind of used earlier, in pharma, they say that a new drug takes 10 years and costs a billion dollars. In chemicals and materials, it's really not that far away of the same thing. It's 10 years and it's 100 to 2, 300, 400 million dollars. And so the rate of progress has actually been slowing. And the bottleneck to that progress is not the discovery piece, it's the scale-up piece. And so that's why we've been very intentional about the way that we've built our business to unlock the benefits of that, which is we build assets ourselves, we license our technology into assets, we use those assets to generate a huge amount of data. That data then feeds our models, and that allows us to build accurate scale-down models. And this is actually. Paraphrasing from like co-founder…
Yes, it was. Oh, you said all the things I was hoping you'd say.
Yeah, I mean, let me just make sure I got it right. It just seems like you've got enough data and you've gotten this data from actually the work that you've done. And that in turn will make it easier to predict future deployments and their success before they're even deployed. Is that about right?
Yeah, exactly. Going back to the whole, what are the AI labs doing is what the AI labs doing here today. They've scoured all the publicly available data, right? I mean, this is not to talk negatively about the founder of DeepMind. He has the Nobel Prize and I have zero. Between us, we have one.…
More from Grow Everything Biotech Podcast
