Red River Insights - Industrial Autonomy top10
Software and robots that make factories run themselves
Red River Insights - September 2025
Dear friends,
Jensen Huang likes to say that every company that builds physical things will also run an AI twin alongside its real factory. He also argues that “everything that moves will be robotic someday, and it will be soon.”
In plain terms, the journey from design to production is being compressed. Car makers now plan entire plants as virtual factories before a single station is installed, so weeks of physical trials become virtual iterations and cleaner instructions flow to the assembly line. Volkswagen is rolling out an AI-enabled production stack with AWS, and Airbus has extended its partnership with Dassault Systèmes so more than 20,000 people use virtual twins across programs. These are not pilots. They are becoming the baseline for incumbents.
There are now more than 4.2 million industrial robots operating worldwide, and annual installations have stayed above half a million for three straight years, with Asia taking 70% of new deployments and Europe 17%. Germany ranks fourth globally with 429 robots per 10,000 employees. The prize will go to the software layer that works seamlessly across brands and sites.
All of this is why we dove deeper and shaped a theme we call Industry Autonomy: software and robots that make factories run themselves. By “Industry Autonomy” we mean the loop from engineering to execution to quality. We include tools that turn designs into ready-to-run instructions for machines, the control software that lifts uptime and speed while cutting defects, and inspection that feeds problems back to fix the process. In plain language, this is about making more good parts per hour with fewer stops and less retuning.
We have spoken to many operators, integrators and founders over the past months, and we are bullish on Europe’s potential to build category-defining companies here. This time we follow the value to where software and robots already move throughput, changeovers (the time to switch a line to a new product) and first-pass yield (the share of parts that pass inspection the first time). One piece of that puzzle is vision AI, where our portfolio company Robovision helps factories catch defects in real time and feed fixes back into the process.
Our latest RAMP ranking highlights this momentum. Scroll down to discover the ten companies shaping Industry Autonomy and the signals they send about the evolution from programmed cells to self-tuning factories.
RAMP's Industrial Autonomy top10
We hope this sparks interesting conversations. If you have any comments or would like to suggest a startup that should be included, feel free to reach out to us. Joseph, Chloé and Olivier will be delighted to discuss these trends and rankings.
(Ranking established on 21/09/2025)
We’ve highlighted 4 key trends illustrated by these companies:
The gap from design to the production line is closing
Neural Concept and PhysicsX compress engineering simulation and then output ready‑to‑run instructions for robots and machines, for example robot paths, process parameters and numerical control programs. In an automotive aerodynamics case, Neural Concept cut the number of CFD simulations by 97.5% while still meeting targets. NVIDIA shows AI‑powered CAE can move evaluations from hours to seconds, which is when digital twins stop being slides and start becoming inputs to production. PhysicsX raised 135 million dollars and says revenue quadrupled over two years. In short, when design tools generate instructions the line can use, changeovers and ramp‑ups shrink and engineers spend less time waiting for test runs. (CAE = computer‑aided engineering, CFD = computational fluid dynamics, NC = numerical control.)
Robot operations is becoming a category
RobCo, Inbolt and Sereact point to a common software layer for programming, scheduling and monitoring robot fleets, which makes any arm useful faster with less fixturing and fewer integrator hours. What’s new is that this layer is shifting from one‑off integrator scripts to off‑the‑shelf platform software you can roll out across production lines and sites.
At Stellantis, Inbolt cut robot‑cell downtime by 97% with a three‑month payback, at MS Direct, Sereact reports over 96% picking success after a few weeks. The practical takeaway is simple: a shared “robot ops” layer turns more tasks into software and makes rollouts easier to repeat across sites.
There isn’t one universal plug that makes every brand of robot speak the same language. Common “bridges” exist and are getting traction (for example, open‑source stacks used in industry, and standards that help machines and mobile robots exchange status and commands), but coverage isn’t universal.
Software‑defined factories are arriving
Daedalus and EthonAI show that quality, traceability and process control are becoming productized software stacks. At a semiconductor producer, EthonAI reports a 51% reduction in yield losses using its platform.
Daedalus is building AI‑run precision factories and the direction is clear: closed‑loop control across mixed equipment looks less like a one‑off project and more like a repeatable product.
Europe is building the AI factory layer
Industrial robot installations in Europe rose 9% in 2023 to a record 92,393. The EU’s robot density stands near 219 per 10,000 manufacturing workers. For context, South Korea is around 1,012, China about 470, and the United States about 295 per 10,000. The auto industry installed about 23,000 new robots in Europe in 2024, the second best result in five years. Funding is following, with European robotics financing in 2025 running ahead of 2024 so far. The setup is positive: a dense installed base and rising capital make Europe a good place to build software that can be deployed across brands and sites.
Many issues are still up for debate (reach out if you have an opinion or to suggest a company we might have missed!):
Humanoids or specialized systems, and on what timeline? Factories care about speed, accuracy and throughput. Today, specialist cells do some tasks in seconds, while general‑purpose robot demos can still take minutes per item. If that gap persists, specialized AI and tooling will win near term. Where do humanoids fit, and what about four‑legged inspection robots (like ANYbotics)?
Where does a shared robot software layer stay sticky? If robot makers bundle similar scheduling and programming into their stacks, what keeps independent “robot ops” platforms installed across brands and sites?
Do these tools run the production line or just show charts? We want products that send instructions to machines and change what happens on the floor. Who can prove they cut setup time and raise the share of good parts within one quarter?
What stops copy‑paste rollouts? Wins in one plant often fail in the next because brands, controllers and quality rules differ. Which vendors were built to work across sites and equipment from day one, with minimal rework?
More on RAMP's scoring method
The ranking of these startups is based on the estimated momentum of the company, but the algorithm does not assess the quality or reliability of the products/solutions developed by these companies!
Find out about the algorithm behind this ranking and the way scores are calculated here: Cheat sheet on RAMP
In case you missed them, our latest top 10s are here: Advanced Materials (July25), Quantum (June25), Spanish tech (May25) Sport Tech (April 25), HR Tech (March25), Open Source AI (Feb25), Agentic AI (Jan25), French Tech (Dec24)
All the previous Top 10 are here.
Other portfolio news:
New exit for Red River West! CMA CMG finalized its acquisition of Brut., strengthening the group’s digital media footprint while Brut. continues under its own brand. (More here).
Okeiro’s model was published in The Lancet, underscoring the exceptional uniqueness of the dataset Professor Loupy has built over the years (more here).
DeepOpinion was mentioned in IDC’s latest Market Glance on the AI-enabled workplace, underscoring its role in agentic automation for enterprise operations. (More here).
Worldia accelerated its Canadian expansion with new team appointments and showcased its platform at the ACTA Eastern Canada Travel Industry Summit (More here).
Ada Health met with pharma leaders at Fierce Pharma Week, presenting its DTC Clinical AI and Patient Finder for earlier diagnosis and engagement. (More here).
Jiko kicked off a September webinar series for corporate treasurers focused on quarter‑end liquidity and 2026 planning, aligning with insights from its Corporate Cash Confidence work. (More here)
Team news:
Team Red River West is back from our September offsite with tons of new ideas for supporting our founders, a fresh view on emerging trends, and a different theories on when AGI will arrive. After many debates we left with decisions and projects that set us up for a focused, high-energy finish to 2025.
(Also, crucial intel: we now know who actually excels at padel!)
Speaking of Padel, we were thrilled to host the very first Red River VC Padel Cup, thanks to all the funds who joined the game (16 funds and perfect parity!)
Where you can find us before year-end:
• September 24–26, 2025: Abel and Olivier at IPEM Paris
• November 19–20, 2025: Chloé, Joseph and Abel at Slush (Helsinki).
You received this newsletter because we met or worked together and we thought you might be interested in an update, if we got it wrong, feel free to unsubscribe with the link at the end!