Robotics Startups funded by Y Combinator (YC) 2026

Micora builds a solar-farm-in-a-box that can be shipped anywhere and built without specialized labor. We have designed a new type of solar farm micorgrid that is palletized, lightweight, and deployable. Our system autonomously builds 100kWp of capacity per day and is easily scalable to MWs of capacity which is a size currently only serviced by slow moving and at capacity EPCs. As demand from data centers, advanced manufacturing, and electrification strains existing power infrastructure, the ability to deploy substantial generation quickly is becoming increasingly valuable and unavailable. We are the only company that services utility-scale power needs in weeks.

Deploy's core belief is that robots should be useful. Getting a generalist robot to do real work is fundamentally the same problem across industries. We build everything between the model and the task to bring generalist robots to the businesses that actually need them.

Shepherd Robotics builds general purpose robots for skilled physical work across AI infrastructure and advanced manufacturing. Our robots are powered by robot foundation models and use modular, flexible hardware designed to adapt to different environments and tasks. We train our systems on a combination of real world and simulation data so they can learn complex manipulation skills and generalize across workflows. We are starting with tasks such as data center construction and operations, cable and wire manufacturing, wire harness assembly, kitting, and factory packaging, where traditional automation struggles with variability, dexterity, and changing environments.

Every instance of intelligence occurring in nature is embodied. Scientific intelligence is no different. We are building general-purpose robotic experimentalists: systems that unite scientific reasoning and physical action to iterate toward discovery by observing, hypothesizing and experimenting.

Dreamscale Labs is building the fastest cloud inference platform for robotics. As complex physical AI models become difficult to fit on expensive on-robot GPUs, we let users offload inference to the cloud and serve them with real-time latency. So far, we've hit 40-50% lower latency than existing neoclouds.

Edgerun builds military exoskeletons. The first product is a 10-pound hip system that fits over existing kit and doubles a soldier’s range. Since April 2026, Edgerun built over a dozen iterations of their product through continuous field tests with Ukrainian fighters and secured $150M+ in LOIs. Founder David Venegas previously co-founded Durin (autonomous mining robotics) and worked at Stoke Space (at 19) and Hadrian. He is a high-school dropout, state-championship runner, adventure athlete, and pilot. Edgerun is a Y Combinator company backed by General H.R. McMaster, Cyan Banister, Joshua Steinman and Jason Miller. edgerun.com

Mass Magnetics turns recycled rare-earth scrap into high-performance permanent magnets, made in America. The materials we supply are the backbone of electric motors, sensors, speakers, hard drives, and electronics. Our process enables DFARS-compliant magnets for defense, aerospace, and industrial manufacturers who demand domestic supply without compromising on quality. Operations in Greater Boston, MA (Greentown Labs) and Columbus, IN.

Physical AI — robots and world models — is bottlenecked on one thing: diverse, real-world data at massive scale showing how humans actually do work in the physical world. It was never written down, and you can't buy it, because no one has ever captured it. Only real local experts, the people who do this work every day, can produce data this authentic and valuable. DeepReach builds the network that reaches them. We design and manufacture wearable stereo capture devices, then place them with local data partners who run collection as their own business — in warehouses, workshops, farms, kitchens and repair shops. They own the local relationships; we supply the hardware, the software platform, the quality pipeline and the payments. Anyone can run a data business. That structure is what produces diversity. Instead of thousands of hours inside one environment, we deliver hours spread across thousands of environments — which is what actually makes robot policies generalize. 475 devices in the field and 100+ data partners are live across 7 countries, adding new partners every week. They've collected nearly 150,000 clips in the past three months, with volume roughly doubling month over month. Our data is in production with multiple frontier model and robotics companies, and we're scaling to 10,000+ devices across 50 countries.

We deploy robots that learn new tasks from a single human demonstration in the factory. We're starting in contract manufacturing, where lines change every week and both traditional automation and today's VLAs fall over. Our founding team has 20+ years of experience in robotics. Navid was part of the core team that productized Spot at Boston Dynamics and led manipulation projects there. Before that, he was at Rethink Robotics. Mojtaba led whole-body control and manipulation efforts at Amazon Robotics and was a Research Assistant Professor at Northwestern University.

Our first model, Atom 1, pre-trained on the internets video, and fine tuned for physics. It has been action conditioned, so you can simulate robot actions within it. This lets Robot Labs train their robots using GPUs instead of relying on human collected manual data that is slow and expensive. In unseen and hard to replicate environments, this is the only path to getting ANY training data.

twentyfour26 builds autonomous CNC machining cells for rapid prototyping, so hardware teams can iterate with software speed.

Enact creates the physical environments to benchmark and hill climb robotics models on high-value tasks. We find where robots fail, and fix them with our in-house training loop.

We mass-produce modular data centers that deploy on stalled infrastructure projects losing billions per week. Delayed infrastructure projects hemorrhage $1B a week in the US alone - Pacific mass-produces modular data centers that deploy on those sites in days and turn their unused power into revenue. Data centers are stuck for years waiting for power, while delayed construction sites across the world sit on unused interconnects, losing billions of dollars per week in interest on trillions of dollars of debt. Our pods deploy on those sites and generate revenue for them in days. They use diamond wafers in the cooling system so pods run anywhere on earth, even a desert. We'll use the same mechanisms we're building today to deploy compute across the galaxy.

Hebbian Robotics builds APIs for robotics data teams to verify data quality for model training. Our customers use us to build quality control pipelines in just a few lines of code, with custom models 10x faster and 8x cheaper than general foundation models. https://hebbianrobotics.com/ https://github.com/Hebbian-Robotics/hflow

We are building robots that can work autonomously in existing wet labs. Unlike traditional lab automation, which requires dedicated equipment and fixed workflows, our robot moves between workstations, manipulates existing scientific equipment, transports lab materials, and executes experiments using the same lab infrastructure as a human technician. We make the robots easy to manage through natural interfaces like Slack, email, and phone. Our goal is to make deploying robots in wet labs as easy as onboarding a new employee, and to accelerate scientific research!

The technologies that define this century, including data centers, rockets, satellites, and robots, are built from complex metal parts. As these systems advance, the parts get harder to make. Past a certain complexity, casting and machining stop working, and the advanced methods that remain are slow, expensive, and hazardous. Ultrasonium has invented a new manufacturing method to make these critical parts 15x faster and 75% cheaper than the methods they replace, in the U.S. and at scale. Solid metal feedstock goes in and a finished part safely comes out, controlled by a physical AI layer and novel control processes. The systems that matter most, in aerospace, energy, and compute, are demanding metal geometries that can't be machined or cast and are too expensive and slow to make with incumbent methods, at a moment when domestic manufacturing capacity is a national priority. Our founding team has built and innovated some of the most complex engineering systems known to man: quantum computers, superconductors, nuclear reactors, and novel aerospace alloy processing methods.

Lambda Robotics builds and deploys robots for the infrastructure powering frontier AI.

OpenVector transforms any existing camera into an automated 24/7 workforce without requiring new hardware. We've built the world's fastest vision-language model inference engine that allows users to simply describe a task in plain English and automatically build a custom AI vision model to monitor their feed. When it detects an important physical event—like a stalled production cycle, a safety violation, or a waiting customer—it operates a computer just like a human to take immediate action. It integrates with legacy systems, makes calls, sends emails, updates a POS/CRM, or triggers webhooks in real time. OpenVector enables any camera to do real work.

We want to push machines to where humans cant go. Presidential early career award winning Duke Semiconductor ECE Faculty Tania Roy and Sanjeev Chauhan are building a frontier semiconductor research lab. Their focus is on solving physical AI challenges from the semiconductor device level and building products around the moat of impossible to copy innovation. Their first work is on a new vision sensor that can run neural network filters right in the camera pixels and only collect data that machines care about. As physical AI develops, machines needs to understand the world around them if they aim to interact with them.

Agency Tool Company ships software to robots in the field 20x faster than a Docker pull with ATC Deploy, sending only the bytes that changed and resuming when the network drops. Deployment tooling is the connection between the lab & the field that continually bottlenecks robotics scaling and lacks a strong robotics-native solution. Deploy is live today with 3 launch partners across ag, construction, and logistics robotics. Our second product, ATC Build, offers embedded-compute as a CI service so devs can run tests and validate their software on the actual target hardware. Build supports the Nvidia Jetson lineup, Raspberry Pi, and a number of other embedded platform targets. Jack Morrison and Davis Foster founded and sold Scythe Robotics after deploying hundreds of autonomous mowers across the country; Agency Tool Company is building the robotics deployment platform across hardware and software they wish they had. Learn more at https://agencytool.com

Proprio Robotics is building state-of-the-art physical intelligence to automate the entire data center lifecycle—from buildout to maintenance to decommissioning. There are currently 400,000 unstaffed data center technicians positions in the United States alone, which means the massive compute needs of AI are being bottlenecked by labor constraints. Proprio is solving this by building robots that do it instead, and is deploying data center maintenance robots starting Q1 2027.

The US Warehouse industry spends $75bn on labor each year. Yet, only 10% of US warehouses have any form of automation, because existing robotics are too expensive and hard to integrate. Manifold deploys autonomous robots that handle picking, loading / unloading, and palletizing — priced by pick. Integration in hours. No upfront costs. No retrofitting. No need to rebuild the operation around the robot. Our mission is to make it easy for any warehouse to adopt robotics.

Operate any robot, anywhere in the world.

In an era of cheap intelligence, the most important application of AI will be to improving human health, globally. Osseus provides off-the-shelf or custom medical and biological datasets and RL environments to frontier labs. We work with hospitals and clinics to provide our EHR and deploy AI models within clinical workflows, with models from both our frontier lab partners and our own research.

We’re building collaborative robotic crews for large-scale infrastructure projects, including solar farms and data centers. 002C: “002A, 5.3 mm left.” 002B: “002A, you’re running low. Next pallet is on the way.” 002C: “Am I clear to come in?” 002A: “Aligned. Panel released. You’re clear.” 002C: “Fastening.” 002B: “Next load arriving now.” This isn’t a human crew talking. It’s three robots building a utility-scale solar farm together. Our AI control system enables specialized robots to collaborate, learn reusable construction skills, like alignment, fastening, and manipulation, and transfer them across workflows, from solar farms to data centers. 🚧 🦺

Robocurve builds open-source tools and independent benchmarks to measure how well robots can do real-world jobs. Instead of relying on unverified demo videos from frontier labs, we score their models on reproducible benchmarks that anyone can trust. Today there are no well-run, standardized robotics benchmarks. Labs evaluate in-house, and no independent group has stepped in to run continuous benchmarking as a service. The result is that no one actually knows how good anyone else is, or where the real frontier sits. Good benchmarks require operating and maintaining physical hardware and real-world setups. Simulation only goes so far, since models that look strong in sim can show large performance gaps once deployed in the real world. Building real-world benchmarks means coordinating job-domain experts, evals engineering, and hands-on robotics all at once. Frontier labs are targeting general-purpose robotics by 2028, yet the field of robotics evals barely exists. Whoever builds the trusted measure of robot capability becomes the reference everyone relies on. We combine backgrounds in AI evals and robotics to build and grow this field as fast as possible. We already shipped v1 of our open-source framework, Inspect Robots, and ran our first pilot scoring a frontier model on a real robot.

Our mission is to build a city on Mars. Cosmic-1, the first equipment in our fleet, is an autonomous robot capable of heavy-lift manipulation. Its been deployed on some of the largest solar farms in the US, more than doubling labor productivity across tens of thousands of panel installs. We're scaling in solar, expanding in data centers, and building lunar construction robotics under contract with NASA.

We manufacture Nori: A sub-$2000 bimanual mobile manipulator that can clean rooms, restock shelves, and fold clothes. Nori can handle basic tasks out of the box, and its software lets anyone teach Nori to automate work specific to their home or business. Nori aims to solve the robotics data bottleneck. By deploying affordable robots to customers around the world, we can gather the diverse data required to train the next generation of generalist robotic policies. Every user of Nori makes Nori smarter. Since launching 6 weeks ago we have already deployed our first robot and have over $440k+ in sales. We can bring general-purpose robotics to homes and businesses that could never afford it before. There are 132.7 million households and 8.4 million employer business locations in the U.S. At $1,688 per robot, selling just one Nori to 0.5% of them represents a $1.2B market opportunity in the U.S. alone.

Our software turns any robot we deploy into an everyday independent worker at your facility. They are able to walk sites, handle instruments, overcome obstacles, collect measurements, and generate auditable compliance records, replacing hours of mandatory manual surveys and inspections. Built by Los Alamos robotics veterans and deterministic-manipulation researchers, we started in nuclear and are already expanding to other hazardous industries.

We build LEGION, a drone swarm that identifies, geolocates, and tracks every emitter on the battlefield using the enemy’s own radio signals. LEGION enables our warfighters to precisely track previously obscured targets like jammers, command posts, and drone operators so they can proactively respond to threats instead of reacting to them.

We're a robotics research lab pushing what's possible with physical AI. We're MIT computer scientists & friends since middle school, with robotics research backgrounds and industry experience at NASA JPL, SpaceX, and AWS.

We build modular humanoid robots and the software platform that allows anyone to train, deploy, and scale robots for real-world tasks. Today, teaching a robot a new skill requires robotics engineers, machine learning engineers, training infrastructure, and significant compute resources. We reduce that process to: Record → Upload → Train → Deploy. Our robots are modular and configurable, with interchangeable end effectors, configurable sensors, and multiple deployment configurations. We believe current humanoid embodiments are often overfit for specific use cases, so we're building a platform that adapts to different environments and applications. Long term, we want robotics to be as accessible as software development. Our goal is to become the default platform for training and deploying autonomous robots.

America does not just need more welders. It needs welding knowledge that can scale. The American Welding Society projects 320,500 new welding professionals will be needed by 2029, with about 80,000 jobs to be filled annually from 2025 to 2029. More than 157,000 current welding professionals are approaching retirement. Welders are only a small share of the skilled-trades workforce, but welding accounts for an outsized share of projected capability demand. The market calls it a labor shortage. AMR sees a learning bottleneck. We are building American-made robotic welding cells that combine computer vision to provide real-time seam tracking and post-weld inspection to turn welding from a scarce manual process into a closed-loop manufacturing system that learns from every weld as it happens.

Ornadyne builds flapping-wing drones for reconnaissance that look, fly, and sound like birds. Today’s drones are easy to detect acoustically, visually, and electronically. As counter-UAS systems improve, traditional quadcopters and fixed-wing UAS systems are becoming increasingly ineffective for close-range surveillance. Flapping-wing aircraft have existed for years, but have lacked the performance, endurance, and control required for real-world deployment. We’re changing that by engineering a high-performance, stealth-optimized system built from first principles for efficiency, autonomy, and reliable field use. The final frontier of surveillance won’t look like machines; it will look like wildlife already in the sky. Maybe birds aren't real?

Hub is a multimodal lab providing real world data to frontier labs and robotic companies.

PerfectBit builds learned policies for robots that do real work. We train manipulation and control policies that hold up outside the lab, in the unstructured environments where real tasks actually happen. Founded by two alumni of Meta's Superintelligence org: Péter Vajda led Media GenAI foundation-model R&D (Emu, Movie Gen, image and video editing), and Seiji Yamamoto led teams in the Core Llama group across LLM pre-training and post-training, inference optimization, and computer vision. We're early, deliberately quiet about the details, and hiring the first researchers who will shape what this becomes. San Francisco, on-site.

Building robots and training models for automating manipulation in warehouse, starting with e-commerce 3PLs

We deploy robots in warehouses to automate the most demanding job in logistics: unloading loose-loaded containers. Our customers include 3PLs (CEVA, Stord), distributors (DHL, PostNL) and many more logistics companies.

Fern provides RL environments for robotics companies powered by a foundational world model. Companies can plug and play their robot policies completely in software with varying starting scenes of their production environment, and we'll produce RL heads on top of your existing policies to improve speed and success rate. Our customers get a continual learning flywheel where their production robots improve every week.

Datoric develops custom datasets for voice models, robotics, and world models, treating research, collection, verification, and production as one continuous process. We work closely with frontier model teams to turn emerging limitations and model failures into testable data hypotheses, while running our own experiments ahead of customer demand. This allows us to operationalize validated methods into repeatable collection systems at scale. Data is collected through private invite-only applications separated by modality, customer, and trust level. Every submission remains linked to the contributor, device, task, session, consent, rights, and processing history that produced it, giving our internal QA and fraud models the context to detect problems that may appear legitimate in the finished file. Each collection reveals new failure cases and quality signals that improve the systems behind the next dataset. Once a collection method is validated, we scale it up as a solution to the model failure and it also becomes a reusable data recipe for future custom projects or independently collected, rights-cleared data products.

We provide defense companies with simulation infrastructure to train and validate autonomous systems.

Robotic factories that build robotics in the US. Currently operating a 12,000 sqft factory in the Bay Area and actively manufacturing customer orders for some of the largest robotics companies in the US and the International Space Station. We’re a group of Berkeley graduate robotics researchers who previously won a full scale autonomous IndyCar competition at 160mph.

Forge Robotics is changing the way that metal parts are fabricated. We are enabling a future where our intelligent robotic system completes all steps needed to take raw metal stock and transform it into fully welded parts. Beginning with welding, we are building a robot-mounted vision system and AI feature detection that creates a real-time 3D map and tells the arm exactly where to go at every instant.

Darwin is building general-purpose humanoid robots for real-world work. Our first robot, Darwin One, will learn from real-world experience and continuously improve as AI moves from the screen into the physical world. First deployments begin in early 2027.

Tornyol builds micro-drones that kill mosquitoes. We use smartphone microphones, car park assist sensors and some clever DSP and control to transform 40-gram toy drones into mosquito killers. Because these drones are so cheap and fast, we will lower the cost of mosquito control by 100x which will enable the eradication of mosquitoes from cities, saving more than a million people each year from Malaria, West Nile, Dengue, etc.