Deep Tech Brief #07

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: a Swedish team rebuilding the fuse box, a German team putting light inside medical tubes, and a European team shrinking the data center to a shipping container.

Let’s dive in.

Every building has a fuse box. Inside sit circuit breakers, the switches that cut power when something goes wrong.

The design is more than a century old and purely mechanical: two pieces of metal that snap apart when too much current flows. It works, but it is slow, it does only one thing, and the contacts can throw a spark, a well known cause of electrical fires.

The world is now plugging in far more, and the equipment itself has become what everyone waits for. Wood Mackenzie data puts the wait for large circuit breakers at two and a half to three years, roughly double pre-pandemic. Projects are not stuck on money or permits. They are waiting for boxes.

Trued Holmquist, co-founder and CEO of Sweden's BL!XT, wants to throw out the moving parts. His product, BLIXT Zero, switches with semiconductors, the same chips that are in your phone. Nothing physically moves, so it cuts power in billionths of a second rather than thousandths, with no spark. And being electronic, it can measure what flows through it, report back, and be reprogrammed in software.

Atom Power in North Carolina won US safety approval in 2019, backed by Siemens, ABB and Eaton. BL!XT's claim is narrower: in 2021 BLIXT Zero became the first small breaker of its type anywhere to pass IEC certification, the international approval needed before it can be sold. The rules for this gear are written by the giants who dominate the market, a quiet way of keeping newcomers out, so BL!XT put a co-founder on the inside, helping write the standard for the next generation.

Where they are: founded 2018 outside Stockholm, around 20 people, EUR 5 million raised in early 2024 led by Union Square Ventures and Energy Revolution Ventures, with partners including Infineon and India's Havells. By Holmquist's account, the high volume product ships early next year, the first run is close to sold out before manufacturing has started, and a funding round is open with terms agreed.

Worth watching, because being first through a gate the incumbents control is hard to copy. The risk is price, which has held this back for a decade: chips cost more than two bits of metal. And BL!XT is a 20-person company competing with firms whose order books run into the tens of billions.

Learn more: blixt.tech

A catheter is a thin tube doctors put into the body to drain fluid or deliver medicine. Almost everyone reading this has had one. Essential, and also a doorway straight through the body's defences.

Infections from tubes and similar devices are between half and 60% of everything people catch in hospital, and in European intensive care units roughly 98% of urinary tract infections come from a catheter. Germany's Robert Koch Institute estimates 10,000 to 20,000 deaths a year in Germany alone.

The usual defences are blunt: change the tube often, give antibiotics when something starts growing. Both are losing ground as bacteria become resistant.

Puray, based in Munich, builds a light source into the tube itself, so the inner surface is disinfected continuously while it sits inside the patient, using a colour of light chosen to destroy bacteria without harming human tissue. Co-founder Martin Duffner's team has run it against E. coli and staph, two of the most common hospital bugs, and against human cells from the urinary tract to check nothing gets damaged.

The obvious version of the idea, coating the tube in something that kills bacteria, was tested properly and failed. The UK's CATHETER trial gave more than 6,000 patients across 24 hospitals either a silver coated tube or a standard one. The infection rate was 12.5% with silver, 12.6% without. Killing bacteria in a lab dish is easy. Stopping infections a doctor would actually diagnose is a different problem, and the one Puray will have to answer in a trial of its own.

Where they are: still years from patients. The technology works in the lab; the team is now in the long testing phase that comes before regulators approve anything. Funding so far is grants and prizes rather than a venture round, including EUR 300,000 from a German federal programme with Munich University of Applied Sciences. Duffner expects market entry around 2028, and is honest that raising money is the hardest part, because the proof investors want cannot exist until the trials are done.

One to watch, carefully. The science is sound and the problem enormous, but the road is long, the testing expensive, and this field is littered with ideas that shone in the lab and did nothing on the ward.

Learn more: puray.de

Most of the AI conversation is about chips. The harder problem is finding somewhere to plug them in.

A large data center needs an enormous amount of electricity, and getting connected to the grid means joining a queue that runs five to seven years in major markets, eight to ten in London and Amsterdam. The building goes up in under two years, so the wait for power, not the construction, sets the pace.

AI agents make it worse. A chatbot answers once and stops. An agent keeps going: it plans, calls the model again and again, checks its own work. Goldman Sachs cites research putting agents at roughly 60 to 130 times the energy of one chatbot question. The 130 gets quoted alone, but it is the top of a range.

GeoMind's answer is to stop dragging power to the data center and put the data center where the power already is. Computing, storage and networking get packed into units the size of a shipping container, dropped at sites with spare electricity. Co-founder Florian Fournier says a unit can be running in about four months, against 20 to 24 for a conventional build.

Data centers in boxes are not new; Vertiv, Schneider and Dell have sold them for years. GeoMind's bet is the software: a system designed to install, run and repair itself with nobody on site, plus encrypted storage and a marketplace where a host can rent out spare capacity. Timing helps: Gartner expects spending on cloud that keeps data inside national borders to grow about 36% this year.

Where they are: just starting to sell, signing deals with large industrial partners and, by Fournier's account, having recently closed a funding round. No customer names, site count or revenue are public.

One to watch, carefully. The argument about power, control and putting computing closer to where it is needed is a sensible read on where AI is going. What is unproven is delivery: this corner of the industry has a history of compelling visions that never won real customers, and until GeoMind can point to running sites and paying customers, this is a thesis, not a track record.

Learn more: geomind.io

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.