The clock is the moat

Why we invest where intelligence meets matter

Ask any founder one question: if a competitor with ten times your money started tomorrow, how long would it take them to catch up?

Not to copy your product. To deliver the same result to your customers, by whatever route is fastest: hiring your engineers, licensing a technology, buying a supplier, or going around you entirely.

For most of the last twenty years, a good software company could answer “years”. Building the product took a large team, distribution took longer, and that head start was part of the moat.

AI is collapsing that answer. A feature that took a team a quarter can be rebuilt in days. A prototype takes a weekend. Distribution, trusted workflows and proprietary data still matter. But the difficulty of writing software, on its own, no longer protects anyone.

In some industries the answer is still measured in years: the ones where intelligence meets matter, whether that matter is a cell, a molecule, a wafer, a robot arm or a plasma. That is where we invest at ZAKA, and the reason is time.

Some clocks can’t be sped up

AI speeds up thinking: designing, predicting, simulating, searching. It doesn’t speed up the physical world. Every hard-tech industry runs on clocks that no budget can compress.

The biology clock. If a trial has to show whether patients are alive two years after treatment, it takes two years. In 2024 David Baker shared the Nobel Prize in Chemistry for designing proteins with computers. A new protein can now be designed in days. Proving that it helps patients still takes years.

The physics clock. A leading-edge chip spends around three months in a fab, passing through roughly a thousand process steps, so every yield improvement takes months to prove. A new material may look promising in simulation, but producing it consistently at industrial scale requires experiments, equipment and repeated measurements. You can simulate more candidates and run more experiments in parallel. You still have to demonstrate that the result works in reality.

The institutional clock. A fab won’t put a new tool into production because a startup shows a good benchmark; the tool has to prove its reliability and yield on the line. A carmaker qualifies a new supplier over years. A regulator takes about a year just to review a new drug application, after years of trials.

The build clock. EUV lithography took more than two decades of research before ASML’s machines entered mass production in 2019. In 2012 Intel, TSMC and Samsung, three of the richest companies in the industry, co-funded ASML’s development. Even they couldn’t buy their way past the physics. They had to wait.

Here is the part most people miss. AI compresses everything around these clocks: discovery, design, simulation, analysis. As the rest gets faster, the clocks become a larger share of the time and money it takes to bring a technology to market. And a company that has already run the clock becomes harder to catch.

In software, AI is eroding the moat. Where intelligence meets matter, it is making the moat relatively deeper.

Some clocks will shrink. Simulation will replace some physical tests, and regulators are starting to accept human-relevant lab models in place of some animal studies. That isn’t a problem for the thesis. It’s where a second kind of company makes its money.

Time spent is not a moat. Time earned can be.

A biotech can spend five years in the lab and come out with inconclusive data. A chip startup can spend three years on a component nobody wants to qualify. A robotics company can collect thousands of hours of data that turns out to be useless for training. None of them has a moat just because time passed.

Time becomes an advantage when it produces something a competitor can’t buy, skip or reproduce: validated evidence, manufacturing know-how, protected IP, customer trust, or learning that compounds with every deployment. A tool that has been qualified on a production line has proved it won’t fail in a place where downtime is extraordinarily expensive. A rival with a better spec sheet still has to earn that trust.

The moat isn’t the time that passed. It’s what the company can do because it passed.

Two kinds of companies we back

The first kind earns an advantage a competitor would have to earn again.

The second kind sells time. It shortens a clock that a whole industry is forced to run.

These companies face a second test: can they capture a meaningful share of the time they save? A product that saves a manufacturer six months is useful. Becoming the trusted way an entire industry avoids those six months is a business.

The two clocks of venture

The obvious objection: if these industries are this slow, are they too slow for venture capital?

Sometimes they are. Plenty of companies have built remarkable technology while burning so much capital that their early investors got almost nothing. That is how cleantech hurt investors fifteen years ago.

So we watch two clocks.

The first is the competitor’s clock: how long a well-funded rival would need to reach the same customer outcome. We want it long.

The second is the company’s clock to its next real proof: how quickly the team can show something that removes a technical, commercial or regulatory risk that someone will pay for. We want it short, and cheap.

Long to replicate. Short to prove.

Not every milestone counts. A first patient dosed is progress, not evidence that a drug works. A working prototype is not a qualified product. A milestone matters when it resolves an uncertainty that customers, partners or the next investors actually care about. We think of it as proof per dollar: what can this team prove with the money it is raising today, and why will that proof make the company worth more?

One warning: the clock can also work against you. Patents age while clinical evidence accumulates. A component can be qualified for an architecture the industry then abandons. Data collected at great cost can lose value when a new generation of models needs less of it. A long clock is only an advantage if what you earn along the way stays relevant.

What this means for software

None of this means software is dead, and we are not orthodox about it. Software can own powerful advantages: distribution, network effects, regulated workflows, exclusive data. Some of the most interesting software companies will sit directly on top of physical industries, learning from lab experiments or factory data that nobody else can collect.

The line we draw isn’t between software and hardware. It’s between what gets cheaper to reproduce as intelligence improves, and what stays hard to replicate even when intelligence is abundant.

The question

So when I meet a founder, I want two answers.

If a competitor with ten times your capital started tomorrow, how long would it take them to reach the same customer outcome? And what can you prove with the money you are raising today?

A long answer to the first and a short, credible answer to the second is where the conversation gets interesting. The best founders in these fields aren’t building things that simply take years. They are building advantages that take years to replicate, and finding ways to prove them much sooner.

If you’re building something a richer competitor still couldn’t replicate quickly, I’d like to hear from you.