ROBOTICS & INDUSTRIAL TECH
European robotics investment in H1 2025 already exceeded all of 2024. The deals getting done are bigger, and the ones falling apart are falling apart on unit economics.
Robotics and industrial automation is where software-native VC meets hardware reality. A compelling technology demonstration is not the same as a scalable product. Bill of materials, failure rates, regulatory certification, and the cost to serve the installed base are the numbers that determine whether the investment thesis holds. We read them before you commit.
Why it’s different
Hardware changes the risk profile in ways that software-native due diligence doesn't reach.
Robotics and industrial tech companies are not software companies with a hardware component. They are hardware companies with software as the intelligence layer — and the hardware fundamentally changes the risk profile, the cost structure, the scaling dynamics, and the regulatory path. Software DD frameworks, applied to a robotics company, miss the things that are most likely to destroy or protect the investment.
01
Unit economics at prototype stage look nothing like unit economics at production scale
The bill of materials for a robotics prototype is irrelevant to the investment thesis. What matters is the BOM at 1,000 units — and whether the supply chain required to reach that cost point actually exists. Robotics startups are building with custom actuators, proprietary sensor combinations, and novel mechanical designs that have no off-the-shelf supply chain equivalent. A tariff shift, a sole-source component failure, or a contract manufacturing negotiation can blow unit economics overnight. We assess the BOM at target scale, not at current prototype costs.
02
IP provenance in robotics is structurally messy
Most robotics startups draw on academic research, open-source firmware, borrowed components from multiple vendors, and occasionally government-funded work. Without clean IP assignment from university partners, contractors, and previous employers, a capital-intensive product may have murky ownership of its core technical assets. Open-source licenses — particularly GPL variants — create compliance exposure in firmware that is especially difficult to manage in distributed hardware products.
03
Regulatory certification is not a milestone on the roadmap — it is a binary gate
A surgical robot requires MDR and potentially FDA approval. An autonomous warehouse robot operating alongside human workers must comply with EU Machinery Regulation. A drone deployed in European airspace faces EASA certification requirements. These are not compliance checkboxes — they are certification processes that can take 18–36 months and cost millions of euros. A company whose commercial timeline assumes regulatory approval on a target date that has no technical basis is carrying valuation risk that compounds with every quarter of delay.
Assessment Areas
Where we focus in Robotics & Industrial Tech engagements.
AI in Robotics & Industrial Tech
AI is what made robotics fundable again. It is also a new layer of technical risk.
The current wave of European robotics investment is explicitly AI-driven — computer vision, natural language interfaces for robot programming, reinforcement learning for adaptive manipulation. These AI capabilities unlock use cases that were previously unviable. They also introduce a layer of technical complexity that sits on top of an already complex hardware engineering challenge.
Opportunities we verify
AI-powered adaptability that removes the need for bespoke programming. The traditional barrier to robotics adoption in SME manufacturing was the cost and complexity of programming robots for specific tasks. AI systems that allow robots to learn new tasks through demonstration, natural language instruction, or limited training data change this economics fundamentally. We assess whether the adaptability is real — demonstrated on genuinely novel tasks with minimal human instruction.
Computer vision systems trained on proprietary industrial data. Computer vision for defect detection, part identification, and quality inspection in industrial settings requires training data that reflects specific components under specific conditions. Companies that have been collecting annotated industrial inspection data for multiple production environments have a dataset that enables accuracy that generic computer vision models cannot match.
Robotics-as-a-service models that shift the hardware risk. The most capital-efficient robotics business models are those where the company retains ownership of the hardware, charges on a utilisation or outcome basis, and absorbs the maintenance and upgrade risk. We assess whether the RaaS financial model is sustainable at the company's current cost of capital.
Risks we surface
Demonstration performance that doesn't generalise to production environments. The most common technical failure mode in robotics is a demo that works beautifully under controlled conditions and fails under real production conditions — different lighting, unexpected objects, unstructured human interactions, production floor vibration. We test generalisation explicitly, not just demo performance.
Custom hardware with no validated supply chain. Early-stage robotics companies frequently build with custom-designed components that have no alternative suppliers. If the sole manufacturer of a critical actuator or sensor goes out of business, changes their design, or imposes price increases, the company has no alternative. We map sole-source dependencies in the BOM as a standard risk assessment step.
AI inference on edge hardware without a validated update mechanism. Robots deployed in industrial settings need to receive AI model updates safely — without creating downtime, introducing regressions, or requiring field service visits. Many early-stage robotics companies have not built a validated OTA update pipeline for AI models on edge devices. This becomes a serious operational liability at scale.
Know what you’re backing before you commit.
X-Ray delivers a full product and tech verdict on any robotics or industrial tech target in one business day — covering the BOM economics, the IP provenance, the regulatory pathway, and the AI deployment architecture.
250+ European engagements · 100% partner repeat rate