Organ-on-chip flow simulation for pharma drug testing
A pharma-tooling startup (introduced via "HMT") builds lab hardware that simulates fluid flow (e.g., blood flow patterns) around tissue samples so pharma companies can test how drug compounds behave in a more realistic, human-relevant environment — without the complexity of a full physiological setup, and without reaching for animal testing as early. Business model: a reusable device plus a consumable base, generating recurring per-experiment revenue. First application: testing how oral drugs cross the gut barrier, built as a platform they intend to extend to other applications from there. Currently running pilot studies with specialized pharma testing companies deliberately, to build real-world evidence before going to pharma directly. Plan: seed round in spring next year, market entry targeted 2028. - Q&A: ~80% of drugs today are described as "not currently testable" well with existing tools by their estimate, which is the scale of the addressable problem. IP: two components — a proprietary device (in "design freeze") and a broader platform layer.
Firefly Sensing — photonic sugar/refractive-index sensor for food and beverage
Founder Don van Elst. Problem: food and beverage manufacturers run extremely tight quality control (e.g., testing sugar content) and simply discard product that misses spec — wasteful and costly. Existing refractometers that measure this in-line are too bulky or expensive for most production lines. Firefly's solution: a sensing system built on photonics research out of Eindhoven University, achieving industry-standard resolution at an order of magnitude lower cost and a footprint small enough to integrate almost anywhere on a line, which larger competitors' hardware cannot match. - Traction: one real deployment already saved a customer 175,000 litres of sugar per year — from a single sensor in a single machine in a single factory. - Market: primary target is food and beverage, described as a $64B total market for this kind of measurement. Three industry pilots underway, two paid, with company sizes ranging 4,000-100,000 employees (undisclosed names). - Roadmap: incorporate pilot feedback, industrialize the design, start an outsourced production chain next year, scale locally with partners within three years, globally within five. - Funding: runway to mid-2028 already secured; currently looking for investors with strong food-and-beverage networks and high-tech manufacturing/scaling expertise rather than pure capital. - Q&A: target price ~€1K per unit eventually (current factory-grade equivalent costs ~€15K), with a recurring revenue component from periodic sensor-head replacement. Adjacent markets identified (chemical processing, bioprocessing, lab-on-chip/organ-on-chip) but deliberately deferred until food and beverage is proven, since those markets take longer to mature.
Memristor-based AI chip materials
An unnamed-on-tape startup pitching a new class of materials ("memristors," described as behaving like synapses) integrated directly into semiconductor chips on standard manufacturing lines — a claimed breakthrough achieved "over the summer." The pitch: because AI workloads are fundamentally different from classical zero/one digital logic (more like how brains compute), new materials are needed to keep improving efficiency, and their chips are pitched at roughly 100x the energy efficiency of comparable GPU-class compute. Funding ask: a large round (stated as billions of euros on tape, which given the context most plausibly reflects a translation/transcription issue — treat that figure with caution) aimed first at the low-power, low-latency edge-inference market. Differentiator claimed vs. the handful of other players in this space: the difficulty of actually integrating these materials into a real chip manufacturing flow, not just proving the material in a lab.
Rapid-prototyping semiconductor manufacturing (mature-node silicon in weeks)
A solo/small-team founder pitching a new model for custom chip ("ASIC") prototyping: silicon in about two weeks for roughly €25,000 per wafer, on a mature process node (so no need for cutting-edge lithography or the highest-performance chip processes). The core argument: unlike software or PCB iteration, committing to custom silicon today is treated as an irreversible, multi-year, multi-hundred-million-euro decision — so most hardware teams avoid it and instead accept bigger form factors, more power draw, and heavier firmware. The startup's thesis: make custom silicon cheap and fast enough to use from day one of a hardware roadmap, iterating with real customers, rather than as a late, one-shot bet. They're vertically integrating three normally-separate pieces — chip design software, manufacturing equipment, and the fab recipe itself — and are working with a major industry partner plus three universities (Eindhoven, and two others named on tape as "Lombard" and "Nijmegen," likely transcription artifacts for the actual partner names) on chemistry, characterization, and design tooling. Self-funded to date; now raising, with "first working silicon" framed as the near-term proof milestone.
Antenna and RF testing automation
An antenna-testing hardware/software company (speaker introduced as "Manu" on tape) targeting the wireless-device certification industry: every connected device (routers, phones, cars, etc.) must be extensively and repeatedly tested against RF emission/calibration standards before it can ship, and that testing today is slow and expensive — some customers run 500+ physical test chambers in parallel just to keep up. The company's system runs many of these RF tests far faster than manual chamber-based testing. - Traction: first system installed and running; doubled revenue year over year; customers named include HP, Ericsson, and Robin Radar; contracts with the US government. - Coverage: currently automates roughly 90-95% of the required test suite, targeting the remaining edge cases over the next two years; expanding from single-product testing into larger installations (e.g., full test rooms for larger radar/satcom terminals) and toward automated diagnostics as a software add-on. - Funding: raising to scale commercial operations, especially in the US, with early expansion into Asia. - Q&A: a reduction in customer run-cost of up to 60% is claimed depending on test-case volume.
Neutral-atom quantum computing
A quantum computing startup spun out of research at TU Eindhoven, building on a "neutral atoms" hardware platform. The pitch frames the opportunity in terms of past computing transitions (missile guidance → weather prediction → personal computing → AI) and argues quantum computing's realistic near-term value is in modeling problems classical computers handle poorly (better medications, better materials for wind turbines and solar panels). The stated problem with quantum computing today: machines cost upward of $100M, some over $1B, need enormous power (up to nuclear-plant-scale), and are inaccessible to nearly any real business case. Their claimed breakthrough: two new methods (credited to two named team members, "Max and Joly," heard on tape) that remove bottlenecks in how atoms are assembled and measured — steps that are normally roughly 1,000x slower than every other operation in the machine — enabling meaningfully faster error correction. Ask: a few million euros to build a demonstrator inside an HPC center at a price point 10-100x cheaper than current quantum computing offerings, with the explicit long-term goal of "thousands of machines, millions of users" rather than a handful of billion-dollar installations.
Parafix — perovskite X-ray detectors
Problem: X-ray-based screening (medical imaging, airport baggage scanning, industrial inspection such as battery or memory-chip testing) is being pushed to detect more (drugs, explosives, cancer) at higher resolution, but the detector market is split between "indirect" detectors (slow, low-resolution, not material-sensitive) and "direct" detectors (have the right specs but are expensive, hard to manufacture, and supply-chain constrained). Parafix's approach: a new detector built on a perovskite-based material combined with thin-film manufacturing, enabling custom-size, high-resolution, material-sensitive detectors that OEMs currently cannot buy at all. Team of three: a chemist, a physicist, and a strategist, including (per the pitch) the original inventor of the perovskite-based detector approach and 40 years of combined thin-film manufacturing experience. Plan: a market demo in 2027, seed round in progress. First target application flagged in Q&A as scientific/lab instrumentation rather than mass-market imaging, as the most credible early wedge.