Manufacturing and Robotics Startups funded by Y Combinator (YC) 2026

Operate any robot, anywhere in the world.

ProvenMetal manufactures the fastest circuit boards in the United States. Domestic circuit board manufacturing is slow and expensive, and while reshoring is happening, the supply chain isn't there yet. We automate the bottlenecks - part procurement, DFM review, and assembly - delivering circuit boards in days instead of weeks.

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.

We manufacture humanoid robots in San Francisco.

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 manual surveys and inspections. Built by Los Alamos robotics veterans and deterministic-manipulation researchers, we're starting in nuclear and expanding to other hazardous industries.

Robotics companies send Shotwell their video data and we detect failures. We do this for both their training data and for their deployments. We've developed models that can automatically produce these annotations with near-instant turnaround time.

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.

We built the fastest software to control robots from anywhere in the world. We make it possible for robotics companies to keep their robots at 100% uptime, and intervene instantly when a failure occurs.

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.

Aseon Labs builds robotic pitstops for self-driving cars. Autonomous vehicles can drive themselves, but they still rely on humans and centralized depots for charging, cleaning and servicing multiple times per day. As fleets scale, the operational layer makes up 70% of the cost of the service and is of the largest barrier to uptime, profitability and market expansion. Aseon Pods are robotic micro-depots that can be deployed directly within operating zones, bringing fleet servicing to where demand exists. By eliminating unnecessary depot trips and automating routine operations, Aseon helps autonomous fleet operators increase uptime, reduce reset cost by 50% and unlock scale. Founded by the team behind Pushme, which deployed 5,000 mobility infrastructure locations across 40 cities before its acquisition by TIER-Dott, Aseon is funded and building product in Redwood City for pilot deployment with major autonomous vehicle operators.

Strawberry farming is a $3B industry in California alone with 60% labor cost using 40-year old software. Synphony provides strawberry picking robots and software services including: bed-level analytics, data pipeline integration, and agentic solutions. Synphony's "side hustle" is selling data, evals, and viral demos to robot foundation labs while building capacity for general robot deployments. We've started with strawberry farming for great margins and outdoor data moat, but also received requests for robots in semi-conductors and autonomous space labs. Synphony's team has won over 15 hackathons together (taking home 4 Jensen-signed GPUs!). Sean did ML research at NVIDIA and led AI products used by IRS and Citibank. Lucas' research spanned robotics, IoT devices, and neuromorphic computing. Alex architected an app with 400k users.

Robotics is the new paradigm. With hardware converging, Intelligence is the biggest bottleneck. Our world was built for humans and we're giving robots the ability to reason and act like us. The way we learn any task - by seeing, doing, and feeling - is how we train robots. We're building general intelligence that works on any robot and fits seamlessly within our world.

We're building robots that just work and use them to deliver autonomous services at scale. Our first use case is semi-humanoid robots for manufacturing and logistics. Unlike traditional robotics companies, we don't sell robots, we provide physical agents for hire where companies pay for the robots based on usage, decreasing the barrier to entry and expanding the robotics market as a whole.

Nine Fives produces modern radio frequency test equipment and solutions to disrupt the $130B test equipment industry. By combining modern electronics, AI ready test automation, and strategic supply chains, we’re equipping our customers to cut their time to market in half.

We are building AI-native fulfillment powered by robotics. Warehouses without human labor that get better, faster, and cheaper with every order. Most robotics companies wait until their system is perfect before deploying. We don't. We ship imperfect policies into real warehouses, detect failures in real time, recover automatically when we can, and bring in a remote human operator when we can't. Every failure becomes training data. Every intervention makes the system better. The result is a fulfillment system that continuously learns from production.

Prototyping.io turns CAD designs into real mechanical parts through an AI-driven manufacturing system. We analyze designs for manufacturability and automate production workflows to deliver high-quality parts faster and at lower cost.

One Robot builds simulation environments that are realistic to see and realistic to interact with, so robotics teams can train and evaluate robot policies without being bottlenecked by robot time. Today, improving a VLA often means more real-world hours: setting up the scene, running trials, resetting, and repeating. This loop is slow, expensive, and hard to scale. For example, material handling and manufacturing assembly tasks, models need far more training and evaluation data than teams can collect in the real world. We use task-specific data to build world model-based simulation environments for hard manipulation tasks (for example, textiles and box folding). These environments help teams run more training and evals, find failure modes faster, and accelerate iteration on action policies with less dependence on real-world data collection and robot availability.

Asimov collects real-world human movement data from households and businesses to train humanoid robots. Unlike factory datasets that capture the same tasks in the same environments, we deliver the full diversity of real human environments, thousands of hours a day to leading labs.

Kyten Technologies is building a factory to manufacture custom aerospace-grade battery packs. 100,000+ drones, submarines, and satellites are being built in the US, but the legacy supply chain can't keep up. During our six years at Starlink, we put 5,000+ battery packs into space. Now, we're bringing rapid development and volume production of aerospace-grade battery packs to the entire industry.

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.

We’re archiving the physical world for embodied intelligence by collecting and labeling aligned multimodal data. To build dexterous and perceptive robots that generalize robustly, we need massive amounts of real-world data across multiple modalities and environments. We have thought deeply about the fine line between biomimicry and its application to humanoid systems. Based on this research, we design and deploy custom hardware across residential and manufacturing settings. We then post-process the resulting data through internal QA, anonymization, and annotation pipelines to deliver diverse, high-fidelity datasets at scale to frontier labs developing robotics foundation models and general-purpose robotics companies. We believe we are at a historic inflection point, with a unique opportunity to leave a dent on humanity and reshape physical labor markets forever. That's why our team dropped out of Stanford and Berkeley and moved to Asia to collect the world’s largest annotated multimodal dataset.

Tensr is building fully autonomous robotic factories that make scaling hardware as effortless as scaling on AWS. 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.

We build brains for robots. We work with manufacturers like Unitree to make robots listen, speak, and act. You can program your robot out of the box by literally talking to it, no coding involved. Robots running Lightberry are emotionally intelligent, always on, and fully autonomous. Just like in Star Wars!

Allus builds next-gen vision foundation models that bring real intelligence to manufacturing. Enabling factories to see, understand, and improve production in real time.

Cortex AI builds the world’s most diverse and large-scale real-world workplace robot & egocentric dataset — where the physical world becomes the next training and evaluation set for embodied AI. We power frontier labs developing robotics foundation models and general-purpose robots by providing the data they need: 1️⃣ Egocentric Data — real-workplace human video with hand/body pose, depth, and subtask labels. 2️⃣ Robot Data — trajectories collected from manipulators and humanoids in real industry settings. 3️⃣ Human-in-the-Loop Rollouts & Evals — real-world deployments with remote operators who recover robots when they fail, capturing data that feeds back into training and continuously improves models. Additionally, through the Cortex Marketplace, workplaces get paid to host data-collection and evaluation sessions, while labs access the in-the-wild data that truly matters. This draws on Lucas’s previous experience as co-founder of Carousell, a C2C marketplace that scaled to a $1B+ valuation.

China and Russia control the world’s critical minerals. We can produce Titanium, Magnesium, Aluminum, Zirconium & Hafnium in the U.S. more profitably than any existing process, and carbon neutrally via co-production. Carbochlorination is the primary process for producing critical metals like titanium, hafnium, and zirconium, and was once the standard for magnesium. The problem: conventional carbochlorination relies on purchasing carbon feedstocks, wastes valuable byproducts, and is highly CO₂-intensive. Duranium changes this with a novel reactor that recycles CO₂ emissions back into feedstock (CO) while generating a valuable co-product (bleaching agents). This closed loop slashes emissions and delivers Chinese-level metal pricing without subsidies — unlocking the full potential of carbochlorination for U.S. critical mineral production.

We build humanoid robots that do your chores at iPhone prices. Just like the millions of cars generating data for self-driving, we need millions of humanoids generating real-world data for general-purpose robotics. Humanoid companies today can only manufacture hundreds a year because motors are expensive and slow to assemble. We use artificial muscles. Each muscle is just a tube wrapped in braided fibre, and the whole robot is powered by a single pump — which means we can mass-produce humanoids by the millions. Our design is built for that scale from day one and is cheap enough to enter billions of homes.

Parametric builds autonomous physical businesses.

We build AI models for controlling robot arms, enabling them to learn new skills in just hours. Our robots help businesses cut costs by automating back-of-house tasks with an innovative pay-by-the-hour business model. Our customers include a biotech unicorn, a leading biotech non-profit, and a direct to consumer apparel brand.

Semi-humanoid robots that do real physical labor. Building hardware and training models, each driving the other. Already deploying in hospitality.

DeepAware (Robotics Center of Silicon Valley, https://roboticscenter.ai/) is the fastest way for enterprises and researchers to get robots and robotics parts in the US — 72-hour delivery or Bay Area pickup. Beyond hardware, we help teams collect teleoperation data, build reinforcement learning environments, and deploy robots into production. Customers include AI labs, industrial operations, research teams, and event producers.

Physical Turing rolls out humanoids in the real world and captures failures for post-training and benchmarking. We procure, staff, and operate a wide range of target environments for real-world rollouts, helping robot makers, model companies, and enterprises continuously evaluate where humanoids succeed, where they fail, and what needs to improve before each deployment.

Nox Metals is an AI-powered metals supplier powering modern American manufacturing at scale. We are building automated factories that turn raw metal into production-ready material faster, cheaper, and more reliably than legacy service centers.

Flywheel AI converts any existing excavators for contractors to enable remote ops to increase safety and productivity, and use robotics context dataset to train autonomous policies.

Boost Robotics is building autonomous mobile manipulation robots to remotely perform manual tasks in data centers. The surge in AI applications is pushing data centers to scale faster than they can manage, leading to suboptimal operations and staffing shortages amid rising power and compute demands. We save data centers millions in insurance premiums and SLA violations by increasing the frequency of inspection and maintenance while reducing the time it takes to triage and address failures.

We manufacturing instruments and reagents that are purpose built for AI, automated labs, and personalized medicines.

Teach robots new skills through natural language and run reliably in production in minutes. Mbodi combines generative AI and agent orchestration, turning everyday language and quick demos into precise, reliable robot actions for industrial automation.

We're building an AI powered platform for engineers that are building new robots and machines. We allow users to invent and modify machine designs in a matter of minutes, not years. Ayna and Andy, the co-founders, met during their time at Harvard and started this in 2024 as their programs were wrapping up. We're commercializing 10+ years research done by our team at Harvard and MIT on this space. Our team wrote the very first papers on applying GenAI and optimization algorithms to compose robots from parts and applying computational design to fine tune machine designs. We're bringing in our deep knowledge of modifying equipment and integrating it into factories keenly understanding what makes hardware hard. We take in parts under consideration, constraints and tasks at hand and our algorithm formulates and fine tunes from 100s of possible design options. All in a few clicks, so engineers can push hardware to market faster and companies can benefit from the inventions being done by their R&D teams. We already have two paying customers in the OEM and system integrator spaces. We are venture backed and supported by a GBP 2.5MM grant by ARIA, the UK Government's invention agency.

We build robots for physical AI. Meet Axol - A dual-arm robot designed for real-world tasks using AI.

We’re building AI-powered systems that automate manual work in scientific labs using robotics. Scientists simply type their experiment in plain English. Zeon translates it into code and runs it on robotic arms (no coding required). By controlling robots with natural language, we’re enabling automation for every lab, every scientist, and every experiment. Zeon is lab-aware. It detects equipment, adapts to protocols, and moves autonomously between instruments to execute full workflows. We’re already working with labs at Stanford and UCSF to accelerate discovery - from running overnight nanoparticle fluorescence measurements to safely disposing of hundreds of bacterial samples. Zeon reclaims time for what matters most: the science.

We capture and structure real factory workflows at scale by combining first-person industrial video with SOP-level process knowledge. This enables robotics and AI labs to train on real production data, not just controlled lab environments.

Inversion Semiconductor is developing the next-generation chip fabrication machine, to create the most powerful chips 15x faster. Our vision is to reshore advanced chip fabrication capabilities in the West. We're building a lithography machine. Lithography uses light to pattern circuit features on silicon. We're scaling transistors to their physical limits by shrinking particle accelerators 1000x, to create a high power light source. Using our light source we will double transistor density for a given numerical aperture and 3x throughput.

Pave Robotics builds robots that seal cracks in asphalt, helping paving companies save time and money. Our robots work 24/7 and deliver higher-quality results than humans.