Artificial intelligence is leaving the lab. Robots are unloading trailers in distribution centers, cleaning solar panels on commercial rooftops, and autonomously sweeping Fortune 500 campuses. AI-native propulsion systems are navigating satellites in orbit without human inputs. Machine learning is predicting electricity demand at grid level in real time, days in advance. Satellites are being tasked from iPhones by high school students. The companies doing this most consequentially are not building demos; they are deploying hardware and intelligence in environments where failure has real consequences, and they are winning commercial contracts because their systems actually work. From NL-driven task robots to AI-powered orbital propulsion, from earth intelligence platforms to sustainable AI infrastructure, these are the scale-ups building the physical layer of the AI economy.
Companies are listed in alphabetical order. This list is non-exhaustive.
1. AMPERON
Headquarters: Houston, TX | Total Funding: ~$30M
The electrical grid is the most complex physical system in human civilization, and it is operating under more stress than at any point in its history. AI data centers are adding unprecedented load. Renewable generation is introducing volatility that the grid was not designed to absorb. Extreme weather is making demand spikes less predictable and more dangerous. In this environment, the accuracy of electricity forecasts, at the grid level and the meter level, is not an operational nicety; it is the difference between grid stability and blackout.
Amperon is the leader in AI-powered electricity forecasting for energy market participants. Founded in 2017 by CEO Sean Kelly and Abraham Stanway, the platform uses machine learning trained on 70 million smart meters across the U.S. to deliver short-term and long-term demand forecasts that are up to three times more accurate than conventional methods, as well as price forecasting at 5-minute intervals for energy trading desks. The company raised a $20 million Series B led by Energize Capital in 2023 and has since received strategic investments from National Grid Partners, Samsung Ventures, and Acario, bringing total funding to $30 million. Clients include Ørsted, HSBC Asset Management, and major utilities and retail energy providers across North America and Europe. The company launched its AI-driven energy forecasting on the Snowflake Marketplace in February 2026, extending its reach to enterprise data teams. In a grid that is being asked to do more than it was ever designed to do, Amperon is the intelligence layer that makes it manageable.
2. ANORI
Headquarters: San Francisco, CA | Total Funding: ~$26M
Getting a building approved and built is one of the most expensive, opaque, and litigation-prone processes in the developed world. Between the moment a developer decides to build and the moment the first shovel hits the dirt, two to four years typically pass, consumed by navigating municipal permitting systems, reconciling conflicting feedback from structural engineers, insurers, architects, and city planners, and managing a coordination process so fragmented that projects routinely die in pre-development without any technical reason to fail. The U.S. needs millions more housing units. The regulatory and coordination infrastructure for building them has not evolved to support that demand.
Anori is the platform fixing pre-development, the most underserved phase of real estate construction. Spun out of Alphabet’s X moonshot factory in March 2026, the company raised $26 million in its first funding round led by Prologis, one of the world’s largest real estate owners, and Builders VC, with participation from X’s dedicated spinout vehicle Series X Capital. Anori’s platform brings all stakeholders, including the developer, architect, structural engineer, insurer, contractor, and city agency, onto a unified system from the start of a project, surfacing compliance conflicts within weeks instead of months or years. CEO Adrian Walker, who spent over nine years at Ford before joining X, is targeting three-to-six-story multifamily buildings of 5 to 100 units as the initial category, a housing type the world needs at enormous scale and is most confused about how to build. Alphabet head Astro Teller, who has watched X attempt this problem twice before, describes the market signal this time as fundamentally different: industry stakeholders said they did not want to see a finished product. They said they wanted to help build it.
3. BLINK TECHNOLOGIES
Headquarters: San Francisco, CA | Total Funding: Undisclosed (early-stage)
Eye tracking technology has existed for decades and has consistently failed to cross the threshold into mainstream adoption because every implementation has required active, cumbersome hardware: cameras mounted on screens, specialized glasses with embedded sensors, or IR illuminators that limit where and how the technology can be used. The applications that matter most, gaze-based accessibility tools, fatigue detection for drivers and operators, attention measurement for education and training, and cognitive load monitoring for high-stakes environments, all require eye tracking that works passively, continuously, and without requiring users to think about it.
Blink Technologies is building passive eye tracking: the ability to infer gaze, attention, and cognitive state from standard cameras and ambient sensors without any specialized hardware or active illumination. Its AI models are trained to extract meaningful gaze signals from inputs that existing computer vision systems ignore, enabling applications in industrial safety, workforce wellness, automotive, and consumer electronics that do not require the user to wear anything, configure anything, or be aware they are being tracked. The platform targets OEM integration with device manufacturers and industrial equipment providers, positioning passive eye tracking as an embedded intelligence layer rather than a standalone product. For the industries where knowing whether a person is alert, attentive, or cognitively overloaded is the difference between safe and unsafe, Blink Technologies is building the sensor layer that has always been missing.
4. CONTORO ROBOTICS
Headquarters: Austin, TX | Total Funding: ~$22M
Unloading floor-loaded trailers and shipping containers is one of the most physically grueling, injury-prone, and labor-constrained jobs in logistics. Unlike palletized freight, floor-loaded boxes come in every size, orientation, and weight, stacked unpredictably by human packers in conditions that vary with every trailer that backs up to a dock. Traditional automation approaches have consistently failed here because the variability is too high for rigid pre-programmed systems and the edge cases are too frequent for AI that has not been exposed to enough of them to be reliable.
Contoro Robotics solves this with a human-in-the-loop model that combines AI with teleoperation to deliver reliability that neither approach achieves alone. Its robotic system, featuring a patented DuoGrasp gripper and a mobile base that moves dock to dock via remote control or forklift, uses AI trained on customer-specific data to handle the majority of boxes autonomously. When the AI encounters a genuine edge case, a remote teleoperator steps in via the Symphony exoskeleton interface, contributing a demonstration that immediately improves the model for next time. Founded in 2022 by CEO Youngmok Yun, a spin-off from Harmonic Bionics, the company raised a $12 million Series A in March 2025 from Doosan, Coupang, the Amazon Industrial Innovation Fund, and IMM, bringing total funding to $22 million. Early deployments have doubled unloading speed, reduced manual labor dependency, and saved hundreds of person-hours per month for customers including Go! Retail Group. Amazon’s Industrial Innovation Fund described trailer unloading as “one of the most labor-intensive and manual processes in the warehouse.” Contoro is the company making it autonomous.
5. DEXMATE
Headquarters: San Francisco, CA | Total Funding: Undisclosed (early-stage)
General-purpose mobile robotics has been a horizon that has retreated as the field has advanced: every year the hardware gets better and the software gets smarter, and yet robots still struggle to operate reliably across the unstructured, changing environments that define most real workplaces. The gap is not primarily in sensing or actuation; it is in the intelligence layer that allows a robot to understand a new environment, adapt to unexpected obstacles, handle manipulation tasks that were not explicitly programmed, and receive new instructions in natural language without requiring a specialist to reprogram it.
Dexmate is building general-purpose mobile robots from the ground up with that intelligence layer as the primary design constraint. Its platform combines mobile manipulation hardware with a multi-modal AI system that enables robots to navigate, interact with, and operate in diverse real-world environments without environment-specific programming. Natural language task instruction, on-device learning from observation, and real-time scene understanding allow Dexmate robots to take on new tasks in new settings with minimal operator configuration. The company is targeting light industrial, logistics, and commercial facility environments where the need for flexible, reprogrammable robotic labor is acute and the existing solutions are either too rigid or too expensive to deploy at scale. For the many industries that need robots that can actually adapt to the job rather than requiring the job to adapt to the robot, Dexmate is building the platform that makes that possible.
6. HALCYON
Headquarters: New York, NY | Total Funding: Undisclosed (early-stage)
Energy data is fragmented, siloed, and underutilized. Utilities, grid operators, energy retailers, industrial consumers, and asset managers all generate enormous volumes of operational data, including meter readings, load profiles, weather correlations, equipment telemetry, and market price signals, but most of it sits in disconnected systems that were not built to share information or be queried by AI. The result is that the energy industry is making billion-dollar investment and operational decisions with significantly less data fidelity than the decisions warrant, and the tools to change that have not kept pace with the urgency of the energy transition.
Halcyon is the AI platform built to change that. Its system aggregates, normalizes, and structures energy data from across the operational technology stack, and deploys AI models that surface actionable intelligence for asset operators, grid managers, and energy market participants. The platform connects operational data from utilities, renewable assets, storage systems, and demand response programs into a coherent analytical layer, enabling queries and forecasts that are impossible when the data lives in separate systems. For the energy companies and grid operators navigating an increasingly complex and volatile supply and demand environment, Halcyon is the data intelligence layer that turns operational data into operational advantage. The transition to a renewable-heavy grid is generating more data than the industry has ever had to manage. The question is whether the tools exist to act on it. Halcyon is building those tools.
7. LUCID BOTS
Headquarters: Charlotte, NC | Total Funding: ~$34M
Building exteriors are among the least glamorous maintenance challenges in commercial real estate and among the most persistently manual. Window washing, surface cleaning, paint removal, and coating application on high-rise and mid-rise structures require workers on scaffolding, harnesses, and rope access systems, performing repetitive work at height in conditions that generate high injury rates, inconsistent output quality, and significant liability exposure for property owners. The labor market for this work is tightening. The safety record is poor. And the economics of traditional exterior maintenance do not scale with the growing commercial real estate portfolio that property managers are being asked to service.
Lucid Bots is replacing that model with autonomous exterior cleaning platforms designed for the structures and surfaces where human labor is most dangerous and least efficient. Its Sherpa drone platform hovers outside building exteriors, performing painting, cleaning, inspection, and coating tasks with modular attachments that allow it to switch functions between jobs. Its wheeled Lavo Bot handles horizontal surfaces including sidewalks, patios, and parking areas. Founded in 2018 and backed by Y Combinator, Cubit Capital, and Idea Fund Partners, the company raised an oversubscribed $20 million Series B co-led by Cubit Capital and Idea Fund Partners in March 2026, bringing total funding to $34 million. The Sherpa flew nearly 28,000 missions in 2025, logging over 4,100 hours in the air, and operators report 1.5x to 2.5x productivity gains. The company generated as much revenue in 2025 as in its first seven years combined. Customers include Disney, Sunbelt Rentals, and commercial facilities managers across the U.S. Fast Company named Lucid Bots one of 2026’s most innovative companies for making robots tackle the kind of tough, tricky outdoor work that humans should not have to do.
8. MOJO AI
Headquarters: San Francisco, CA | Total Funding: Undisclosed (early-stage)
Industrial worksites generate safety incidents at a rate that no other sector of the economy tolerates. Construction alone accounts for roughly one in five worker fatalities in the U.S. annually, and the incidents that precede those fatalities, near-misses, PPE violations, unsafe proximity to equipment, unauthorized zone entry, and worker fatigue, are predictable with the right sensing and AI. The problem is not that the data to predict them does not exist; it is that most worksites have no systematic way to capture, interpret, and act on that data in real time before the incident occurs.
Mojo AI is the AI platform for industrial worksite safety, built to make predictive safety monitoring accessible to the construction, energy, and industrial sectors that need it most. Its system integrates with existing jobsite cameras, wearables, and IoT sensors to detect unsafe behaviors, hazardous conditions, and anomalous patterns in real time, and surfaces alerts to safety managers and site supervisors before they escalate into incidents. The platform applies computer vision and machine learning to the continuous visual data stream that modern jobsites generate but rarely analyze, converting passive surveillance infrastructure into an active safety intelligence layer. For every safety manager responsible for a site where incident rates are both a human cost and a financial liability, and where the gap between a near-miss and a fatality is often the speed of intervention, Mojo AI is the AI layer that closes the gap.
9. MORPHEUS SPACE
Headquarters: El Segundo, CA / Dresden, Germany | Total Funding: ~$43M
Low Earth orbit is becoming congested at a pace that regulation cannot keep up with. Thousands of new small satellites are being launched annually, driven by the economics of commercial satellite constellations for broadband, imaging, and signals intelligence. In that environment, a satellite that cannot maneuver is a satellite that cannot comply with collision avoidance requirements, cannot adjust its orbital slot to serve a customer’s changing needs, and cannot execute the controlled deorbit that regulators increasingly require. Electric propulsion is the only technology that can provide the sustained, efficient thrust that small satellites need to remain maneuverable throughout a multi-year mission lifetime.
Morpheus Space is the leading manufacturer of scalable electric propulsion systems for the small satellite market. Its GO-2 thruster, based on Field Emission Electric Propulsion technology, uses a metallic propellant that eliminates the leakage risks of chemical and pressurized systems, provides redundancy through 40 individually controllable thrusters, and delivers high total impulse in a compact, modular package designed for satellites from 10 to 250 kilograms. Founded in 2018 at the Technical University of Dresden by Daniel Bock and co-founders, the company demonstrated the GO-2 in orbit in July 2025 and raised a $15 million Series A+ from Alpine Space Ventures, the European Investment Fund, Lavrock, Pallas, and Vsquared in February 2026, bringing total funding to $43 million. Its Reloaded Facility in Dresden produces 100 units per year, with capacity to reach 500. The company’s AI-powered Sphere constellation management software enables “objective-first” mission planning, where AI continuously optimizes the constellation’s orbital configuration to serve user objectives with full automation. For the satellite operators building the next generation of commercial constellations, Morpheus is the propulsion infrastructure that makes them viable.
10. ORAN DEVELOPMENT CO.
Headquarters: New York, NY | Total Funding: Undisclosed (early-stage)
Mobile networks are software now. The shift from proprietary, closed-stack telecom hardware to Open RAN architecture, where network functions run as software on general-purpose hardware, is one of the most consequential structural changes in telecommunications infrastructure in a generation. It is also creating an AI deployment problem that the industry has not yet solved: the computational demands of running AI natively in the radio access network, optimizing spectrum allocation, managing interference, and personalizing network experience at scale, are enormous, and the tooling to deploy and manage AI workloads in that environment does not yet exist at production quality.
ORAN Development Co. is building the AI-RAN infrastructure that makes intelligent, software-defined mobile networks operational at scale. Its platform provides the orchestration, management, and optimization tools that allow network operators to deploy AI models natively within Open RAN environments, enabling real-time network intelligence without the latency of routing decisions through centralized cloud infrastructure. The addressable market is every mobile network operator transitioning to Open RAN architecture, a transition being driven by government mandates, security concerns about legacy vendor concentration, and the economic pressure of reducing network operating costs. As AI-RAN becomes the standard architecture for 5G and 6G networks, the infrastructure company that makes that architecture operationally viable is building at the center of one of the largest technology transitions in global telecommunications history.
11. PROMETHEUS HYPERSCALE
Headquarters: Austin, TX | Total Funding: Undisclosed (growth-stage)
AI data centers are among the most energy-intensive infrastructure assets ever built, and the energy grid that powers them was not designed for this. A single hyperscale AI training cluster can consume as much power as a small city. The combination of power density, heat output, and 24-hour operational requirements creates an infrastructure challenge that conventional data center design cannot address efficiently. At a moment when both the regulatory environment and the investment thesis for AI infrastructure are placing sustainability at the center of new builds, the technical and economic case for a new approach to AI compute infrastructure has never been stronger.
Prometheus Hyperscale is building sustainable AI infrastructure designed specifically for the compute density and power requirements of frontier AI workloads. Its approach addresses the full stack of data center sustainability: power architecture optimized for AI accelerator loads, advanced cooling systems designed for the thermal profiles of GPU clusters, and operational efficiency systems that reduce both the energy cost and the carbon footprint of AI compute at scale. The company positions itself at the intersection of two converging pressures: the hyperscalers’ commitment to net-zero data center operations and the AI industry’s insatiable demand for more compute. For the enterprises, cloud providers, and sovereign AI programs that need massive compute capacity but face growing regulatory and reputational pressure around energy consumption, Prometheus Hyperscale is building the infrastructure that resolves the contradiction.
12. SENSERA SYSTEMS
Headquarters: Golden, CO | Total Funding: ~$40M+
Construction sites are among the most data-rich environments in the physical world and among the least instrumented. Thousands of workers, dozens of subcontractors, heavy equipment, material deliveries, and concurrent workflows create a coordination and safety challenge that project managers typically navigate through a combination of daily site walks, paper sign-in sheets, and after-the-fact incident reports. The gap between what is happening on a jobsite in real time and what the project team knows about it is where schedule overruns, safety incidents, and cost overruns originate.
Sensera Systems is the market leader in AI-powered jobsite intelligence for the construction industry. Its self-contained, solar-powered, wirelessly connected cameras can be installed on any jobsite in 15 minutes without external power or network infrastructure, and its SiteCloud platform uses AI to interpret the continuous visual data stream from those cameras in real time. SiteCloud Insights delivers plain-language summaries, safety alerts, dumpster status notifications, and progress milestones directly to project teams on mobile devices, without requiring anyone to manually review footage. The company raised a $27 million Series B led by 10 Atlantic Group in February 2026, with participation from Egis Capital Partners and MUUS Asset Management, taking total funding to over $40 million. Clients include TRITEC Real Estate, which deployed Sensera cameras across its $113 million Station Yards mixed-use development in New York, and commercial facility managers and developers across the United States. CEO Robert Garber has described the Series B as purpose-built to advance the company’s AI strategy. For construction teams where the cost of a missed safety event or a delayed decision is measured in lives and millions of dollars, Sensera is the intelligence system that keeps the site visible.
13. SKYFI
Headquarters: Austin, TX | Total Funding: ~$12.7M+
Satellite imagery has been transforming how governments, militaries, and commercial enterprises understand the physical world for decades. The problem has never been the data; it has been the access. Tasking a satellite, retrieving imagery, and generating actionable insights from it required contracts with large commercial providers, specialized GIS teams, and multi-week procurement processes that most organizations could not navigate. The result was that satellite intelligence was available to large institutions with dedicated earth observation programs and largely inaccessible to anyone else.
SkyFi is the self-service earth intelligence platform that removes every one of those barriers. Its web platform, mobile app, and API allow any user to task satellites, access archived imagery from over 50 geospatial data partners, and run built-in AI analytics without GIS expertise, without contracts, and with transparent per-image pricing. The platform supports defense, government, commercial energy, finance, agriculture, and infrastructure use cases, with flexible ordering options calibrated to the specific needs of each sector. Founded by CEO Luke Fischer and co-founder Bill Perkins, SkyFi raised a $12.7 million Series A in January 2026 co-led by climate-focused Buoyant Ventures and dual-use investor IronGate Capital Advisors, with participation from DNV Ventures, Beyond Earth Ventures, and TFX Capital. The round oversubscribed its initial $8 million target following a record year for defense-related investments. SkyFi was also selected, from 3,600 applicants across 24 NATO countries, for NATO’s DIANA 2026 Challenge Programme. Fischer’s measure of the product’s accessibility: his teenage daughters task satellites for their homework from their iPhones.
14. TRENER ROBOTICS
Headquarters: San Francisco, CA | Total Funding: Undisclosed (early-stage)
Industrial and warehouse robots have traditionally required specialist programming for every new task. Adding a new product SKU, reconfiguring a picking workflow, or adapting to a new pallet format means halting operations, calling in an integrator, and spending days reprogramming the system. For organizations that operate in dynamic environments where the task mix changes constantly, this constraint has made general-purpose robotics economically unviable: the cost and time of reprogramming exceeds the value of the automation.
Trener Robotics is eliminating that constraint by building robots that can be instructed in natural language. Its platform enables operators without robotics expertise to assign new tasks, adjust workflows, and reconfigure robotic behavior through ordinary language commands, with the robot’s AI interpreting the intent and executing the task without specialist intervention. The natural language interface is backed by a learning system that improves task execution with every deployment, narrowing the gap between instruction and performance over time. The addressable market is every warehouse, manufacturing floor, and logistics operation that needs task flexibility without programming overhead, which is to say every organization that has decided not to automate because the cost of programming outweighed the cost of the human labor it would replace. For those organizations, Trener Robotics is building the interface that makes the economics of robotics work.
15. VIABOT
Headquarters: Sunnyvale, CA | Total Funding: ~$9M+
Large property portfolios, corporate campuses, logistics facilities, and commercial real estate have an outdoor maintenance problem that has historically been solved with rotating crews of manual labor performing sweeping, debris removal, and security monitoring on schedules calibrated to cost rather than need. The labor market for this work is tightening, the cost is rising, and the service quality is inconsistent. Outdoor maintenance is, in the company’s own framing, one of the least glamorous industries in the world. It is also one of the most operationally important for the property managers responsible for the condition and security of large outdoor footprints.
ViaBot is the robotics company that has built a commercially viable solution to this problem by solving the Swiss Army knife challenge that has prevented outdoor robotics from scaling: building a robot that is multifunctional enough to be economically worthwhile without being so complex that it becomes unreliable. Its RUNO platform is a self-charging, self-emptying autonomous outdoor sweeper that handles cleaning, debris detection, and passive security monitoring simultaneously, capable of scanning license plates, flagging suspicious activity, and escalating alerts to security teams without requiring a separate system. Founded in 2016 by CEO Gregg Ratanaphanyarat and Andre Ding and backed by Baseline Ventures, Morado Ventures, Grit Ventures, and SOSV, ViaBot has deployed robots across Fortune 500 campuses and formed a strategic partnership with Cushman & Wakefield for Bay Area commercial properties. Era Ventures, GS Futures, and RXR ARDEN Digital Ventures are among later investors. For property managers who need outdoor maintenance performed continuously, automatically, and at a cost that manual labor cannot match, ViaBot is the platform that finally makes autonomous outdoor robotics commercially sensible.
This was a brief overview of the rapidly evolving AI robotics, industrial, and physical AI landscape. If there is a company you think belongs on this list, reach out to our editorial team and we will make sure they are included on the next one.
