Barcelona's THEKER raised €73 million this week in the largest robotics Series A ever closed in Europe, and pulled CRV, Samsung, and LVMH onto the cap table to do it. The number matters less than where the robots already are. THEKER's machines are running inside live Inditex production facilities, which is a threshold very few robotics companies anywhere have crossed.
European robotics has spent years long on research and short on capital willing to back hardware at scale. That gap is the week's spine: the money is finally arriving, and it is arriving for companies with machines on real factory floors rather than demos.
Startups & Funding
THEKER closes €73M to put reconfigurable robots into live factories
THEKER raised €73 million ($85M) on June 11 in a round led by CRV, with Samsung, LVMH, Cathay Innovation, 20VC, Henkel Ventures, Korelya, and Bright Pixel Capital joining. The raise takes total funding to roughly $106M, a year after an €18 million seed.
Founded by JiaQiang Ye Zhu and Carla Gómez out of the robotics community at Barcelona's Polytechnic University of Catalonia, THEKER builds generalist industrial robots rather than humanoids. The hands, arms, and form reconfigure for the task, whether sorting packages, packing clothing, or handling bottles, without manual reprogramming for each new SKU.
The strategic backers are the signal. Samsung and LVMH are not financial tourists in a robotics round, and the company plans to scale to 100 staff by early 2027 as it deepens deployments with tier-one operators.
Undo raises $37M to give AI agents the runtime context to debug
Cambridge-based Undo closed a $37 million (£27.6M) round on June 16, led by Elsewhere Partners. The company records what software actually does at runtime, then feeds that execution history to AI coding agents so they can perform root-cause analysis across development, test, and production.
The bet sits on a real gap. Coding agents are fluent at writing code but weak at diagnosing why a complex system broke, because they lack the runtime context a human debugger reconstructs by hand.
Undo plans to scale product, support, and go-to-market across the US and Europe. The capital is a wager that automated debugging, not just code generation, is the next contested layer in AI software engineering.
Lithuania's PDKINEMATICS raises €2M for battlefield-proven drone guidance
Vilnius-based PDKINEMATICS closed a €2 million seed on June 17, co-led by Coinvest Capital and Iron Wolf Capital. Its Gannet system is a low-cost precision-guidance module that lets operators release drone munitions from up to 700 metres, roughly six times higher than unguided drops, without losing accuracy.
The problem is concrete. In Ukraine, around 300,000 drone-deployed munitions are dropped monthly, almost all of them unguided.
Gannet is jam-resistant, platform-agnostic, and already integrated into Ukrainian operations, with integration onto a new UAV type claimed in under four weeks. The systems were selected for Iron Wolf 2026, Lithuania's largest military exercises.
(Tech.eu / EU-Startups)
Products & Technology
VivaTech and Nvidia's Paris keynote test whether sovereign AI is real yet
Nvidia's Jensen Huang delivers the GTC keynote at VivaTech in Paris from June 17, with European AI infrastructure as the headline theme. The question underneath the staging is whether the sovereign-AI commitments made across Europe in 2025 are turning into running compute.
The clearest test case is Mistral's data centre near Paris, built around 13,800 Nvidia GB300 chips and expected online this year. Announcements at the keynote will show whether Europe is building its own AI capacity or mostly hosting someone else's.
Climate & Energy
A new diagnostic goes into JT-60SA as Europe's fusion programme moves to exploitation
EUROfusion reported on June 17 that a divertor spectroscopic imaging diagnostic is being installed in JT-60SA, the world's largest operating tokamak, a joint Europe-Japan machine. The instrument measures impurity radiation in the divertor, the region that handles the reactor's exhaust heat and is one of the hardest engineering problems in fusion.
This is unglamorous and important. Fusion's path to the grid runs through diagnostics like this, which turn a machine that can hold plasma into one that scientists can actually read and control.
Research & Deep Tech
Oxford builds a new family of Schrödinger's cat states from exotic quantum parts
Oxford physicists reported a new class of quantum superpositions assembled from components that are themselves strongly nonclassical, demonstrated in the motion of a single trapped ion. Earlier cat states were built from ordinary ingredients; these are built from exotic ones, which opens a richer space of states to work with.
The practical interest is error correction. Encoding information in oscillator-like states rather than simple two-level qubits is a leading route to fault-tolerant quantum computing, and a wider toolkit of such states is a usable resource rather than a curiosity.
(Phys.org / University of Oxford)
One to Watch
Reconfigurable robots versus humanoids, decided on real factory floors
The robotics money this year is split between two bets. Humanoids promise a general-purpose body for human environments, while THEKER's wager is that reconfigurable, task-shaped machines reach paying industrial work sooner.
THEKER's edge is that its robots are already inside Inditex facilities, not staged in a lab. Watch the deployment count with tier-one operators through end of 2026 and into 2027.
If reconfigurable systems rack up live industrial installations while humanoids stay in pilots, the strategic backing from Samsung and LVMH will look less like a bet and more like a read of where embodied AI actually lands first.