Deep Tech

Thirty years of software changed how we find and share things. Deep technology can change the physical world. That is where we want to invest.

The last 30 years of software and the internet have been remarkable. They have also been shallow. Most of what we built changed how we find things, not what things are.

Thirty years of shallow technology

Uber did not change how we travel. We still sit in a car on a road. Uber changed how we find the car. It joined buyers and sellers in a two-sided marketplace and removed some cost, maybe replacing the driver with a robot. There was an app for that. It turned out there were many apps for that.

Social media is similar. Instagram and Facebook changed how we share photographs. They made the smartphone more valuable, and gave us a better community notice board. They also learned to serve us advertising and to keep us in our own small corners.

That is useful. It is not fundamental. Very little of it changed the physical world, other than perhaps making us all more angry, with the resulting effects on the world around us.

The research nobody can read

The world now publishes about 3 million peer-reviewed journal articles a year, and the number grows by about 4% annually.¹ No researcher can read it all. No team can connect it all.

So people specialise. They stay in their own corner of the literature, and their own corner of thinking. Real change is rarely learning something new. It is first unlearning something old. That is hard to do from inside one corner, talking to the same people.

Today's computing power can read across all of it. It can connect findings that no single person would ever see together. It can propose new experiments to run.

There is a thesis here that the world has become so complicated that we have crossed the Rubicon, and are now unable to live without AI to help us connect all the dots. When I grew up we were all working on our cars: changing drive trains, cleaning and improving the performance of the combustion system, not just tyres and spark plugs. Anyone doing that on their EV?

From bits to atoms

The difference between deep tech and shallow tech is simple.

The evidence already exists. In July 2022, DeepMind, co-founded by New Zealander Shane Legg, from the land of Mana Tupu, and the European Molecular Biology Laboratory released more than 200 million predicted protein structures, covering almost every catalogued protein known to science.² Working out one structure used to take months or years in a laboratory.

In November 2023, DeepMind then reported 2.2 million new crystal structures, of which 380,000 were predicted to be stable. External laboratories had already made 736 of them. These are candidate materials for batteries, solar panels and chips.^3,4

STM Report (2018); EMBL (July 2022); Google DeepMind (November 2023).

These are tools, not finished products. A predicted protein is not a medicine. A predicted crystal is not a battery. But the starting point for discovery has moved.

The problems we have in front of us that matter most are physical. They don't exist in a better corner of the internet to deploy rage bait. They are clean water, clean air, a biodiverse planet, and abundant low-cost energy. Deep technology can work on them directly.

Take nuclear fuel. Natural uranium is only 0.7% uranium-235, the kind that splits easily, and conventional reactors run almost entirely on that small fraction. The other 99.3%, uranium-238, mostly ends up as waste.

Composition of natural uranium.

A travelling-wave reactor puts it to work. Its fast neutrons turn uranium-238 into uranium-239, which decays within days into plutonium-239. The plutonium then splits, releasing energy and the neutrons that breed more fuel and keep the chain reaction going. The reactor makes its own fuel from what we currently call waste. That's progress, and that's deep tech.

An honest look at my own industry

I say this as someone who works in financial services technology. At Moroku, we ask how to process a loan application faster, or how to help someone save a little more. That work is worthwhile. At a global level, it changes real problems very little.

If technology could change the incentives that keep people poor, at scale, that would be deep. Most fintech does not attempt it. But financial inclusion is a very deep problem, and deep tech can get us there.

Big names, small changes

Even the most famous technology companies mostly improve what exists. Electric vehicles are real progress. But a better battery factory and a cheaper car are manufacturing gains, not new science. Satellite internet connects more people to the internet. It does not change what they do there.

Reaching Mars would matter. It is a step to new materials, new knowledge and further exploration. We are not there yet, but we should get on our way with more impetus: capital, scientists and engineers, backed by a hunger for big ideas.

What this means for how we invest

We are custodians of our investors' capital. Returns matter, and every investment passes the same governance process: an Investment Committee with independent members, and a documented mandate.

But we do not want only incremental gains: another million here, another billion there. That has been done before. We want to back what has not been done before. Many of our investors have made money. What increasingly matters to them is impact and legacy: putting capital to great use on the big problems.

Venture Fund I invests at Series A and B in fintech, agritech and cleantech. Agritech and cleantech are where New Zealand already works with atoms: soil, water, energy and food. We will ask one extra question of every company: does this change something physical, or only how we find it?