When Coexin first started making deep tech investments in 2017, our LP conversations were different from what they are today. We spent a lot of time explaining why the standard VC thesis — seed to Series A to exit in 7-8 years — doesn't describe how hardware companies actually develop. Not as a problem to apologize for. As a feature that explains why the returns, when they come, are outsized and defensible in ways that software exits rarely are.
That conversation has gotten easier. Three exits from our portfolio over the past four years — at multiples that software-only funds envied — have done more to clarify the thesis than any LP deck ever could. But the underlying structural point is still widely misunderstood, even by investors who think they understand deep tech. Let me be precise about what I mean.
The Standard Model's Assumptions
A conventional venture fund operates on a 10-year fund life, typically structured as a 5-year investment period and a 5-year harvest period. The investment model assumes that by year 4-5, you know which companies are winners, and by year 7-9, the winners have reached liquidity events. The portfolio construction is designed around this: you make enough investments to have statistical confidence that 1-2 will return the fund, and you reserve capital for follow-on into the winners.
This model was designed for software businesses because software businesses were where VC was invented. A software company with product-market fit can scale revenue 3-5x per year on incremental capital. Distribution is primarily digital. The main bottleneck is sales and marketing execution, not manufacturing, regulatory clearance, or physical infrastructure buildout. The compounding works quickly enough to produce the DPI expectations that institutional LPs require within the fund's life.
Deep tech doesn't work this way. Not because deep tech companies are less scalable — they can be more so — but because the time to first commercial revenue is structurally longer. A semiconductor materials company developing a new gate dielectric process needs 2-3 years of materials research, 1-2 years of process qualification at a customer fab, and then 2-3 years of production ramp before meaningful revenue recognition. That's 5-8 years from inception to commercial volume. A quantum computing company addressing real enterprise workloads is probably 7-12 years from the research stage where we find them to the commercialization stage where financial returns materialize.
What This Means for Portfolio Construction
Our answer has been to raise deep tech-specific funds with 12-year lives and appropriate LP expectations from the start. We don't pretend that deep tech companies will exit on software timelines. We structure our fund documents accordingly, and we select LPs who understand and genuinely accept the different time horizon — which, in practice, means a narrower LP base than a general-purpose software fund, weighted toward endowments, family offices, and strategic corporate LPs with multi-decade investment programs.
Portfolio construction also differs. We run a more concentrated portfolio than a typical multi-stage software fund — 8-12 companies rather than 30-40. This reflects two realities. First, our diligence process is genuinely intensive. We review IP filings, reproduce experimental claims where we can, interview the technical community around each company, and do end-market analysis that goes beyond the usual TAM slide. This takes time and scales differently than software diligence. Second, the companies we back need more support than capital — introductions to procurement officers at defense primes and semiconductor IDMs, recruiting for technical roles that require very specific domain expertise, and help navigating government grant programs that can extend runway without dilution.
The Return Profile Is Different — and Better
Deep tech companies that succeed tend to exit at higher revenue multiples than software companies, for reasons that are structural rather than incidental. A company with a proprietary materials process embedded in a fab's production line has switching costs that are enormous. Requalifying an alternative supplier at a leading-edge fab takes 18-24 months and carries yield risk that fabs are extremely reluctant to accept. The resulting pricing power and customer stickiness translate directly into margin and multiple expansion.
Our three exits to date have had enterprise value multiples in the 15-25x revenue range at exit, versus the 8-12x range that's typical for mid-tier software acquisitions. The absolute dollar returns were driven by that combination: longer hold, but higher multiple on a larger revenue base than software companies at comparable stages typically achieve. The internal rate of return on those three positions, while lower than the IRR on a theoretical 3-year software exit, produced better cash-on-cash multiple than virtually anything in our LP's software portfolios from comparable vintage years.
We've heard the argument that deep tech investing is structurally incompatible with venture fund economics. We've also heard it from people whose portfolios haven't produced the returns we've produced. These two data points are related.
The Geopolitical Tailwind Is Real
I want to be careful not to over-rely on policy tailwinds in our investment thesis — subsidies and government programs can change, and companies built primarily on government revenue are fragile. But the structural shift in how Western governments think about semiconductor and quantum technology supply chains has created a market environment that substantially de-risks deep tech investing compared to five years ago.
CHIPS Act funding, DARPA program investments in quantum computing and advanced materials, and allied nations' equivalents represent a first-revenue opportunity for early-stage deep tech companies that didn't exist in the same form in 2015-2018. A portfolio company that wins a DARPA contract in year 2-3 has 18-24 months of funded technical development that validates the science and creates customer relationships — all without dilutive equity financing. That changes the funding math significantly for companies in the early years before commercial revenue.
We've become deliberately better at helping portfolio companies navigate government funding processes over the past three years. It's not glamorous work, but it's real value added that software-focused VCs can't provide. When a portfolio company can close a $4M SBIR Phase II contract that validates their technology against a DoD specification, that's a significant de-risking event that shows up in the next private round's valuation.
The Patience Premium
The deep tech investing thesis ultimately rests on a simple proposition: if you have the patience to sit with companies through the technology development curve, and the technical expertise to know which companies will emerge from that curve with durable competitive positions, you can earn returns that aren't available to investors optimizing for shorter time horizons. That premium for patience exists because most institutional capital is structurally unable to wait. Pension funds, endowments, and fund-of-funds have allocation cycles and mark-to-market requirements that push toward liquidity. Capital that can genuinely commit to a 12-year horizon is scarce. Scarcity creates return.
We've built Coexin around that proposition. It requires a specific kind of LP, a specific kind of portfolio construction, and a specific kind of operational involvement with portfolio companies. None of that is replicable by a generalist fund trying to add "deep tech exposure." But for LPs and founders who understand the model, the alignment is unusually tight.
If this investment model resonates and you're building in deep tech, we'd like to understand your work. Contact Coexin.
