Claire Delaunay

The golden age of AI and robotics in industry - and agtech - is fast approaching.   Nvidia CEO Jensen Huang, speaking during a recent visit to China, reflected on the shifting tides in technology: “Over the next decade, as we develop a new generation of advanced plants and factories, they’ll be highly robotic — helping us address the global labor shortages we’re facing.”   And this transformation doesn’t stop at factories. It’s coming to our fields, our greenhouses, and even the mushroom caves.   Although the agrifood tech sector has been slower to embrace AI compared to other industries, its role is becoming increasingly vital, particularly in processing the vast and growing volumes of data generated by IoT sensors and computer vision systems across fields, greenhouses, and food production facilities. AI is now turning complex data—structured, textual, and visual—into actionable insights that drive automation and enable more intelligent decision-making. In agriculture, these data streams span GPS coordinates, humidity, radiation, and temperature measurements, as well as key input data related to water, nutrients, and pesticides. Unlike many industrial sectors, this data is highly fragmented and not readily compatible with generic AI platforms, opening the door for specialized, domain-specific tools such as AI-powered farm management systems, vision-based robotics, and predictive analytics.   At the same time, robotics is advancing rapidly alongside AI. Modern robots are becoming increasingly adept at navigating and operating in complex, unstructured environments—significantly broadening their potential applications. This progress is powered by advancements in sensing technologies and computing power, enabling real-time, deterministic operations that are essential for safety in dynamic settings. Among the most promising technologies we’re tracking are autonomous farm robots capable of working independently or in coordination with human operators, AI-powered drones with autonomous navigation, and swarm robotics systems that coordinate decentralized, multi-agent operations. The same autonomy and perception technologies are now being applied across crops from lettuce to orchards to row crops, driving one of the most important transitions in agriculture.   Including the latest investment that we announced yesterday as lead of 4AG Robotics $40m Series B,  together with Cibus Capital, Astanor has invested in five AI and robotics companies, each with the potential to become a category leader in fast growing multi billion markets.   4AG’s breakthrough in automating mushroom harvesting shows how full stack AI-powered robotics can solve not only labor shortages but also boost yields, standardize quality, and reduce emissions. https://lnkd.in/e-YeTzDf  

67

5 Comments

We just wrapped up a huge week at Figure - here’s what we announced: 1/ Announced >$1B in Series C. This gives us the strongest balance sheet in humanoid robotics - critical to scaling Helix (our AI) and BotQ (robot manufacturing) 2/ Launched a partnership with Brookfield, who manages $1T in assets globally. Together we’re planning to scale AI infrastructure for Helix, already collecting real-world data for Helix pretraining, and working commercially to deploy robots 3/ Announced Project Go-Big - our push to build the world’s largest humanoid pretraining dataset. We are scaling out human video collection and our F.02 humanoid can now learn directly from human video. This project is accelerated by our partnership with Brookfield, who owns over 100,000 residential units As a founder, timing is everything. Too early or too late is fatal Right now, we’re in the sweet spot for humanoids: advanced hardware is maturing that can walk and do human-like tasks, production manufacturing is spinning up, and most importantly - neural networks are delivering. I believe we’re at the starting line of scaling synthetic humans to billions of units

1,902

78 Comments

“How many takes did that demo require?” That single question exposes the gap between robotics content and robotics reality. Oliver Hsu, from Andreessen Horowitz, nailed the core issue in "The Physical AI Deployment Gap": research is sprinting ahead with VLA models and sim-to-real, while most factory robots still live in a world of fixed routines, tight constraints, and zero tolerance for surprises. The real gap is not intelligence. It’s deployment. Factories don’t care about a perfect 30-second demo. They care about: - running 24/7 - handling variance, drift, and imperfect parts - hitting cycle time - staying stable for months, not minutes At Inbolt, we build for that reality. Not for demos. We are live in production today on hundreds of robots across top manufacturers. Real lines. Real uptime constraints. Our focus is simple: turn CAD-level intent into live robot execution, continuously, fast enough for the robot to adapt while the line keeps moving. If your system needs perfect conditions, it’s not Physical AI. It’s theater. #Robotics #Automation #AI #Manufacturing #Inbolt

37

1 Comment

Y Combinator’s “Techno-Industrialist” call couldn’t be more aligned with what’s happening in the field right now. The next wave isn’t just humanoid robots—it’s practical, non-humanoid robotics and physical AI systems in industries that have long needed reinvention. As someone building in this space, I see firsthand how fast things are changing. At TACTUN, we’re helping machine builders cut their R&D cycle from months to days—by giving them a platform that delivers both the controller and no-code end-user software out of the box. Machines like hydraulic test systems that used to take years (and millions) to develop can now be launched in weeks. 📽️ Great quick intro from Diana Hu: https://lnkd.in/enqT_fzk 📌 YC’s piece on the “Techno-Industrialist”: https://lnkd.in/eeckNv7b Also inspiring to see companies like Gecko Robotics, Bear Flag Robotics and others showing what’s possible when the digital and physical worlds finally merge. If you’re working on the ‘real world’ side of AI—would love to connect and share notes. #brain4machinery

14

2 Comments

I’ve been working on the theory of why small, shippable factories are the perfect inversion to the centralized prefab factory model. It all collapsed into one line of math (credit to Gilles Retsin): Factory Efficiency = Usage × (Throughput × Product Value ) / ( CapEx + Deployment OpEx) The idea is that the closer you can drive Usage to 100 % while keeping the denominator tiny, the faster the flywheel spins.  A shippable microfactory tackles this by primarily focusing on a high Usage factor. Because the microfactory can be packed up afterwards and reused, its CapEx gets amortized over multiple projects, effectively raising its overall Usage across the year. If one project alone doesn’t fully occupy the factory, it can simply roll over to another job to maintain high Usage. Its CapEx is often an order of magnitude lower than a centralized factory (< $1M vs. $10-50M min), so the breakeven throughput is more achievable on a small pipeline of work. The real breakthrough here is portability. There is near-zero stranded capital and almost no idle time. To flesh this theory out, I wrote 3000 words on the topic for Brad Hargreaves and Thesis Driven. The article breaks down: - why “fixed factories, shipped goods” is being inverted to shipped factories, fixed goods - what a sub‑$1 M robotic cell does to CapEx per home versus a £45 M off‑site plant - early production/cost data on a shippable microfactory from Mollie Claypool, Gilles Retsin, Sam Baker and the Automated Architecture (AUAR) team - the hybrid reality of robots tackling the heavy structural components and human crews assembling faster and cheaper - technical constraints to pull this model off and future considerations on the optimal business model (own vs. rent) Major shoutout to Gilles Retsin for helping me pull this together. https://lnkd.in/eRTnbZPa

54

2 Comments

🎙️ In this episode, I talk with ⁠Nikolaus West⁠, Co-Founder & CEO of Rerun; their team is building the data stack for Physical AI: We get into the early days of Rerun: how an open-source visualization tool for multimodal data became widely adopted across robotics, spatial computing, and even inside companies like Apple and Meta. But that was just the start. Now, they’re building a full-stack platform for logging, querying, and managing robotic-scale data, from raw logs to model training. Niko shares his personal journey from business school in London to engineering in Sweden, to startups in retail, Kenya, and AR. Along the way, he learned the hard truth: physical AI teams are still flying blind when it comes to data. That pain turned into obsession, and obsession turned into Rerun. Building Deep Tech #69 "You can really build things with a small team." With Nikolaus West, Co-Founder of Rerun. 🎥 YouTube: https://lnkd.in/dpMmB2c5 🎧 Spotify: https://lnkd.in/dEyjeREW 🎧 Apple: https://lnkd.in/d3EGj4MZ

59

4 Comments

Chatting with ZEDEDA CEO, Said Ouissal, and he has an elegant way of explaining the evolution of industrial Edge. Zededa provides OS for orchestrating Edge AI. 2010s were about IoT/Edge Connectivity. 2020s were about Edge computing/processing. 2030s will be all about Edge AI.

36

4 Comments

Forbes published my latest article on what many investors may be getting wrong about robotics. Humanoid robots look like the future. But investing in humanoid hardware may not be. Manufacturing asymmetry is real. Competing with Asian supply chains is not a typical venture risk profile. Outside of companies like Tesla or Amazon, most hardware startups, even unicorns, will face extreme pressure. The long term U.S. advantage is far more likely to sit in embodied AI software and the intelligence layer, not in assembling robot bodies. In robotics, the control point will not be the factory. It will be the mind. https://lnkd.in/gmcVtCa5

90

15 Comments

Despite all the big funding rounds and flashy demos in US robotics, K-Scale’s inability to raise more money should worry us. We're at risk of replaying the LLM story all over again in robotics: - Chinese companies are going open-source and collaborating across the value chain (from EV suppliers to downstream integrators) - most western teams are going full-stack proprietary, closed-source, all in-house Guess which robots the next wave of research labs and startups will actually be able to build on when they want to invent new algorithms or tackle unseen real-world use cases?

273

33 Comments

🔥 Uncomfortable truth about humanoid robots. After listening to Nikita Rudin from Flexion Robotics [a Moonfire portfolio company] on the TWIML podcast with Sam Charrington, one point stood out clearly. https://lnkd.in/emAtmPWM There is no single humanoid robot today that truly generates end-to-end value. It might be doing components right, but not doing exactly what it is supposed to do. Many demos look impressive, but behind the scenes, there is often teleoperation, human intervention, or cleanup work that outweighs what the robot actually does. Most companies are honest about this. These systems are still beta. The interesting part is not the criticism. It is the timeline. The prediction shared on the episode was concrete. By the end of 2026 or early 2027, we will start seeing robots doing real, useful work. Not a flashy consumer moment, but real jobs in industrial settings. From there, scale accelerates through 2027 and 2028. Why has it taken so long? Because we keep calling problems solved too early. Locomotion is a great example. Walking on flat ground looks successful, but real locomotion involves handling uneven terrain, clutter, noise, and disturbances. Adding vision improves behaviour but massively increases complexity. Perspective, camera noise, lighting, and motion blur make training harder. The payoff is that robots can plan ahead rather than just react. Another key takeaway is pragmatism. Rather than pushing everything end-to-end, it often makes more sense to split the problem. Train locomotion to work anywhere. Train a planner to decide where to go. Add semantics so robots understand what they should avoid, not just what they can physically traverse. The sim-to-real gap is still one of the hardest problems. Closing it requires a deep understanding of the system, from motor currents and delays to sensing and actuation, not just better models. Industrial environments will come first. Clear tasks. Controlled settings. Engineers on site. Homes are a much harder problem and will follow later. The takeaway for me is this. The hype is early. The progress is real. And the window where robots start quietly doing real work is closer than most people think. Then the scale will follow. Scale of tasks and scale of the number of robots out there. From hundreds to thousands to tens of millions.

52

1 Comment

Manufacturing is a $20T industry constrained not by demand but by labor. $6–8T is spent annually on manufacturing labor, while robotics captures <0.3% of that. Reshoring is accelerating. Labor is thinning. The old model of rigid automation doesn’t scale. Proud of the foundational work in robotic intelligence we’ve done at CynLr under the stewardship of my partner Gokul N A over the last 6 years. We call it 𝑶𝒃𝒋𝒆𝒄𝒕 𝑰𝒏𝒕𝒆𝒍𝒍𝒊𝒈𝒆𝒏𝒄𝒆 - enabling robots to understand, adapt, and operate in real-world variability. https://lnkd.in/gjNZE6gW

99

Reshoring manufacturing to the U.S. hinges entirely on advances in automation and robotics—and those advances, in turn, rely on next-generation tools like integrated development environments, foundation models, smart robotics skills, and flexible automation platforms. This isn’t merely an incremental shift; it’s the next industrial revolution. I estimate it will unlock up to $30 trillion in GDP over the next decade, as countries compete fiercely for leadership. Eclipse is our time!

266

18 Comments

There’s a massive opportunity in automating industrial manufacturing processes. That’s why we’re backing RobCo in their series C as they scale their physical AI driven robotics platform and expand their U.S. presence. As global labor supply continues to tighten, automation in manufacturing is shifting from “nice to have” to essential. RobCo's Robotics-as-a-Service model removes traditional friction, enabling rapid deployment and real production output in weeks – faster than most B2B software implementation timelines. In Roman Hölzl we immediately saw that rare leader – one who pairs a bold, long-term vision for transformative AI with a relentless focus on real-world productivity today. We couldn’t be more excited to back RobCo as they execute on an ambitious roadmap! Press release: https://lnkd.in/euCx6WX2 Roman Hölzl, Dennis Hauser, Arjan van Staveren, Stefan F. Heidenreich Lightspeed, Sequoia Capital, Kindred Capital VC, Greenfield Partners, Leitmotif, The Friedkin Group, Alexander Schmitt, Luciana Lixandru, Leila Rastegar Zegna, Marc Sieberger, Raz Mangel #Lingotto #LingottoInnovation #PhysicalAI cc Clemente Dadoo Lomeli

154

8 Comments