(Re)Share #54 - If you give a mouse a genome
EV charging | Genetic engineering | Robotic VLAs | Neuro biomarkers | Liquid cooling
I’ve been writing this issue for a couple weeks now and hit some big delays due to 3 live deals that we’re running with 2 maybe new investments to announce in a few months. It’s a fantastic time to be a deep tech, Europe-focused investor.
Lots and lots of interesting stuff that’s come out since we last spoke so let’s get to it.
Stuff Worth Sharing
Driving range anxiety - The infrastructure economics of EV charging remain a significant hurdle to mass adoption. This article details the safety engineering built into charging stations—and how those good intentions are stifling proliferation. I encourage a read, but the TL;DR is that we need to eliminate galvanic isolation links: the physical and electrical separation between the two circuits that make up a charging station. The argument for this safety requirement is that it’s overkill and contributes as much as 50% of a charging station’s total cost. The counterargument is… death. Obviously, safety is paramount, and even perceived risk would almost certainly curtail mass adoption. Gasoline‑pump fatalities are effectively non‑existent—unless you’re a male model.
If you give a mouse a genome - Being on the endangered species list just got marginally less awful. A Dallas‑based company, Colossal Biosciences, took the science world by storm when it genetically engineered mice to exhibit traits of the extinct woolly mammoth. It doesn’t hurt that these woolly mice are adorable. Using CRISPR gene‑editing technology, scientists introduced mammoth DNA sequences into mouse embryos, resulting in mice with long, wavy woolly hair and enhanced fat metabolism—adaptations their “ancestor” developed for life in cold environments. This experiment serves as proof of concept for the complex genetic engineering required to potentially revive a creature that’s been extinct for about 4,000 years. Colossal hopes to produce a mammoth‑elephant hybrid by 2028 and usher in a new era of de‑extinction science.
Out of body experience - DeepMind hit the robotics world in a big way with the release of two new Gemini Robotics models. The first is an advanced vision‑language‑action (VLA) model designed to enhance robotic interaction with the physical world. There’s a branch of AGI research that argues true intelligence requires physical awareness—after all, biological sentient life exists in a physical space governed by Newton’s laws. VLAs are the in‑silico answer, integrating physical actions as a new output modality. The second Gemini model focuses on embodied reasoning, enhancing spatial understanding and enabling robots to perform complex tasks like determining optimal grasp points and safe object trajectories. The article goes on to describe the three fundamental axes of robotic intelligence—generality, interactivity, and dexterity. These attributes track remarkably closely with Fly’s two most recent robotics investments (both unannounced), so watch this space.
Don’t blow your cool - One of the investment deep dives we undertook recently was on colocation data centers and the energy‑capacity crunch they’re facing. I’ve outlined physics‑proof compute as one of my Hard Problems for 2025, and if you’re interested too, this is the report for you. In what I can only describe as extremely comprehensive, the team at Semianalysis walks through everything you need to know about data‑center design, liquid cooling, PUE optimization, and the trade‑offs each hyperscaler has made. This might be a multi‑part read (it was for me), but it’s well worth the time.
What’s cooler than being cool? - A completely different approach to the data‑center energy crisis? Put them in space. Data centers currently consume roughly 1–2% of global electricity—a figure expected to double by 2030. Space‑based data centers could leverage abundant solar energy and the vacuum of space for ultra‑efficient cooling. Perhaps the most compelling case is the security benefits of a space or lunar deployment, which make sense—at least for military applications. I’ve seen a few startups pursuing this, and in my view it’s a matter of when, not if. My bet on “when,” however, comfortably exceeds a typical 10‑year fund cycle.
Teaching old dogs new tricks - Loyal, a clinical‑stage animal health company, received FDA clearance for their LOY‑002 drug, designed to extend the healthy lifespan of senior dogs. LOY‑002 acts as a caloric‑restriction mimetic, delivering the benefits of caloric restriction through a beef‑flavored tablet—without causing weight loss or appetite suppression. To further evaluate LOY‑002’s efficacy and safety, Loyal is conducting the STAY study, a large‑scale clinical trial involving over 1,000 dogs across approximately 70 veterinary clinics in the US. Long‑time readers know I’m the proud parent of the world’s goodest boy, Moose. A frequent argument I have with my wife is my 100% commitment to cloning Moose so we can truly be BFFs. She thinks I’m ridiculous and refuses to give the idea any credence. Personally, I feel it’ll be easier to find a second adequate‑enough spouse.
Bro-parenting - A research team out of China unveiled a groundbreaking study in which they corrected imprinting abnormalities, resulting in the development of viable bi‑paternal mice. Fievel now has two dads. Imprinting abnormalities have long been a barrier to unisexual mammalian reproduction. By editing 20 imprinted loci, scientists produced bi‑paternal embryonic stem cells capable of full development. It may all sound like weird science fiction—and it sort of is—but the study holds therapeutic promise for reversing disease phenotypes associated with imprinted gene dysregulation.
Sharing is caring - While others on Fly dove into energy stuff, I was learning as much as humanly possible about Parkinson’s disease (PD). This condition affects approximately 10 million people globally—a figure that has doubled over the past 20 years and is expected to double again by 2040. Despite its prevalence, like all neurodegenerative diseases, PD has no cure and offers few treatment options due to the inherent difficulty of testing and diagnosing brain‑related conditions. It’s widely accepted that neurology lags behind other high‑impact fields like cardiology and oncology, largely because capturing high‑quality longitudinal patient data is so challenging. One sign of promise is our growing ability to leverage AI and federated learning. A team from Oxford recently published an insightful thought piece on early research, predicted benefits, and suggested tactics for building a cross‑organization data corpus while maintaining patient privacy. The authors argue that this is the only near‑term way to meaningfully accelerate biomarker identification and patient stratification—both essential for viable therapeutic innovation. As many readers know, I have a special place in my heart for neurodegenerative research, so watch this space.
You cell a hard bargain - A Bay Area–based university collab, Arc Institute, unveiled the Arc Virtual Cell Atlas—a comprehensive, open‑source repository of single‑cell measurements encompassing data from over 300 million cells. This gift to the bio‑world aims to advance AI‑driven research by providing computation‑ready datasets. The Atlas features two primary datasets: Tahoe‑100M, which maps 60,000 drug‑cell interactions across 50 cancer cell lines, and scBaseCamp, which includes observational data from over 200 million cells spanning 21 species. Scaling laws have unlocked AI, but scientific disciplines are still stunted by the availability of high‑quality data. Based purely on the reaction from Fly’s techbio founders, I think this is going to lead to very big things.
UnAmerican Idle - As we’ve discussed in past issues, Europe is having a moment—not exactly the moment it wanted, but it’s the moment it’s been given, because the US seems intent on abandoning a global system of trade and shared security that dates back to the Marshall Plan and underpins the most peaceful era in human history. But hey—let’s bring coal jobs back… ANYWAY, defense spending has already reached decade highs, which will require a massive repurposing of idle factories. The European Commission’s “ReArm Europe” initiative seeks to mobilize approximately €800 billion for military expenditure, with shuttered auto factories likely the biggest near‑term winners. While the stakes are certainly concerning, this is a real opportunity for the EU to leapfrog into a modern arsenal.
Portfolio Flex
Transcelestial is back in the news with the announcement that they’ve been chosen as the space-to-ground connectivity provider of Europe’s 6GStarLab, a 6G low-orbit research facility.
Orbital Materials got a nod from The Economist in this overview of AI transforming material science.
Elea came out of stealth and announced their €4 million seed round, which Fly proudly co-led.