Michael Dempsey (Compound) on the 'first 40 months' framework for deep tech fundraising

Summary

Michael Dempsey, managing partner at Compound, argues that deep tech founders are thinking about fundraising the wrong way. Rather than sprinting between rounds with a narrative-first approach, they should map out what he calls the "first 40 months" — roughly the seed through Series B arc — as a hypothesis-building exercise that forces honest self-assessment on milestones, market fit, and momentum.

The framework is not a prediction tool. Dempsey is explicit that the point isn't to know exactly what gets a Series A or B done from day one. The value is having a baseline assumption to stress-test, especially now that venture funds have shifted toward momentum investing. In his reading, the 2021 playbook of building a narrative two to three months before a raise no longer works — founders now need 12 to 18 months of demonstrated progress to access the same capital that once moved on a story.

Narrative vs. earned progress

The market distortion Dempsey flags is real but nuanced. Foundation model companies and robotics plays are raising tens or hundreds of millions at nine-to-ten-figure valuations out of the gate, often on the strength of team pedigree and narrative rather than product milestones. Whether that's a distortion or rational pricing is, in his view, something the next five to fifteen years will sort out.

For everyone else, the 40-month lens is partly a survival tool. Seed-to-Series-A graduation rates sit around 50%, and the median company takes more than 12 months to make that crossing. Founders who have mentally modeled a 40-month runway have a fallback when the go-big bet doesn't land — a slower path that keeps the company alive and positioned for a second or third iteration.

Science project or venture bet

On the question of when research becomes fundable, Dempsey draws the line at commoditization curve clarity. The key question for Compound is whether a team has a credible hypothesis about how quickly the underlying technology gets cheap and where value capture concentrates as it does. A small technical breakthrough that can be scaled is the threshold; pure workflow digitization for scientists, by contrast, has been "largely a graveyard of startups" because getting consistent adoption inside academic or corporate labs is structurally hard. He points to the B2B SaaS-for-biotech thesis around Benchling as an example where the initial excitement hasn't translated into widespread commercial traction.

The more promising near-term biotech angle in Compound's view is tooling that helps human scientists structure experiments in ways that eventually allow robotic systems to run them — building data moats that position companies for the longer autonomous science transition rather than betting it arrives on a fixed schedule.

DeepMind and Isomorphic

On the Isomorphic Labs spin-out from DeepMind, Dempsey is measured. DeepMind and Google broadly are machines he says he can't dismiss — every month makes it harder to doubt they'll figure something out, and every month also makes it harder to see how that something scales into a venture-grade business. His standing concern with Google is the same one he applies broadly: everyone underestimates distribution and overestimates technical competency. Isomorphic's edge in biotech is its social capital — the credibility to move drugs through an industry where relationships matter as much as the science.

Crypto

Compound's crypto thesis splits into near- and long-term bets. Near-term, Dempsey is focused on real-world asset tokenization and using smart contracts to give mid-scale asset classes — he cites renewable energy as a portfolio example — access to capital flows and back-office efficiency. Longer-term, the portfolio includes Prime Intellect on decentralized compute and Molecule on decentralized science, where the thesis is that community-financed drug development can address indications pharmaceutical companies ignore. In the middle sits DePIN infrastructure, following the Helium model, as a mechanism for using token incentives to bootstrap novel networks.

The structural constraint Dempsey sees across all of crypto right now is talent density. Regulatory clarity is arriving, but the developer talent to build serious applications hasn't followed yet — partly, he suggests, because the legal risk of building in the space until recently was real enough to keep capable people away.