Deep Tech Brief #11

Welcome to this week's issue of Deep Tech Brief.

Every Saturday, we break down deep tech companies building genuinely hard things, in plain language. No jargon, no hype.

Three companies this week: gallium oxide wafers in Berlin, resident robots in the North Sea, and woven robot limbs in Budapest.

Let’s dive in.

The usual explanation for slow electric car charging is the battery. It is not, or at least not only. A large part of the limit is a component almost nobody discusses: the power semiconductor, the switch that converts and controls high-voltage electricity on its way into the car.

Silicon ran out of headroom for this job years ago, which is why the industry moved to silicon carbide. Silicon carbide works, but it is difficult and expensive to manufacture, took decades and enormous capital to industrialise, and the supply is concentrated in the United States, Japan and China. For a continent trying to electrify everything at once that is an uncomfortable dependency, which is roughly what the European Commission said when it unveiled Chips Act 2.0 in June.

NextGO Epi, spun out of Berlin's Leibniz Institute for Crystal Growth, is betting on the next material along. Gallium oxide belongs to a class of semiconductors that can hold off higher voltages in a thinner layer, and crucially it can be grown using melt-based methods closer to how ordinary silicon is made, which is the widely cited reason it should end up much cheaper than silicon carbide. The company's own numbers put it at ten times more power-efficient, up to six times the voltage density, and up to 75% cheaper to produce. Its stated headline application is cutting an electric car charge from an hour to ten minutes.

The company makes the epitaxial wafer: the thin, electrically active crystal layer grown on a gallium oxide substrate, which is what device manufacturers build on. It says it is the only producer in Europe making these at industrial quality, in diameters up to four inches.

Where they are: founded in 2025 by three doctorates out of the institute, Ta-Shun Chou, Andreas Popp and Andreas Fiedler, on more than a decade of research there. A €2 million pre-seed in July led by Vireo Ventures, with Ultratech Capital Partners and IBB Ventures, is its first institutional money. The company says it already has revenue, customers on three continents and more than twenty active collaborations, including Foxconn and Taiwan's ITRI.

Worth watching, because being the only European source of a strategic material is a rare position and the policy tailwind arrived the same month as the funding. The milestone ahead is diameter. Four inches is a research-scale wafer, and the volume power electronics industry runs on six inches and up, so the test is whether the growth process holds quality as the wafers get bigger.

Learn more: nextgoepi.com

Europe is covering its seabed in infrastructure. Offshore wind farms, export cables, interconnectors, pipelines and the data cables that carry almost all intercontinental internet traffic. All of it needs inspecting, and after the recent run of damaged cables in the Baltic, all of it needs watching.

Here is the economics of doing that today. You hire a survey vessel, which can cost up to $100,000 a day, and you staff it. By chief executive Jean Crosetti's account, vessels and crew make up 80 to 90% of what offshore inspection costs. The robot in the water is close to a rounding error.

That single fact shapes everything. Because each trip is so expensive, inspection is episodic: a campaign now and then, months of nothing. Subsea cables and offshore assets sit essentially unwatched in between, which is fine until it is not.

Bubble Robotics, founded last year and working out of Paris, Zurich and San Francisco, is attacking the ship rather than the sensor. It builds resident robots designed to live at sea permanently, with a reported six-month endurance, inspecting and monitoring continuously without a support vessel or a crew above them. The enabling shift is recent and unglamorous: edge computing good enough to make decisions underwater, and satellite links good enough to get data back.

They are not alone in seeing this. Norway's Kongsberg Ferrotech raised €12 million with backing from the NATO Innovation Fund for subsea inspection and repair robotics, which tells you the category has serious money and defence attention behind it.

Where they are: founded in 2025 by Jean Crosetti and Patricia Apostol, robotics engineers from NASA's Jet Propulsion Laboratory and ETH Zurich, out of the Entrepreneurs First programme. A $5 million pre-seed led by Episode 1 Ventures, with Asterion Ventures and Norrsken Evolve. The company reports signed letters of intent worth over $4 million, with first deployments planned in offshore wind, subsea infrastructure and maritime security.

One to watch. The cost argument is arithmetic rather than opinion, and letters of intent at pre-seed are a real signal in a conservative industry. The milestone ahead is endurance: turning those letters into deployments, and proving a machine can survive six months alone in the North Sea, which is the whole business in one number.

Learn more: bubble-robotics.com

Robot intelligence has moved extraordinarily fast. Robot bodies have not, and the reason is how they are made.

A dexterous robotic hand is assembled from hundreds of precision parts: bearings, screws, rigid linkages, motors, wiring, all fitted together largely by hand. That makes bodies slow to build, expensive, awkward to redesign, and full of the joints where things eventually fail. It also forces a permanent trade-off. Strength wants rigid parts, safety around people wants softness, and manufacturing has never let you have both.

Allonic's answer is to stop assembling and start weaving. Its process, which the company calls 3D Tissue Braiding, starts with a skeletal core and braids soft, load-bearing fibres directly around it, integrating tendons, actuators and wiring into the structure as it goes. A finger comes out of the machine as one continuous piece rather than a kit of parts. It is closer to making rope than machining metal, and the design software turns a model into production code much as a 3D printer does.

The pitch is speed as much as cost. Chief executive Benedek Tasi describes going from an idea to a physical robot in minutes instead of weeks, which changes not just the price of a robot arm but how many versions a team can try.

Where they are: founded in Budapest by Tasi, Dávid Pelyva and David Holló, with a team of around 15 and a growing US presence. The €6 million pre-seed in February, led by Visionaries Club with Day One Capital, Prototype, SDAC Ventures and TinyVC, is reported as the largest pre-seed round in Hungarian history. The angel list is the interesting part: individuals from OpenAI, Hugging Face, ETH Zurich and Northwestern.

One to watch. Betting on the manufacturing layer while everyone else builds robot brains is a genuinely contrarian position, and the people closest to the software side are the ones funding it. The milestone ahead is proof on the bench: showing a braided limb matches a conventionally assembled one on precision, strength and lifetime, in the hands of the industrial partners now lining up pilots.

Learn more: allonic.co

Know a deep tech company we should break down next? Hit reply and tell us. The best editions come from real tips.

See you next Saturday.