Physical AI and Autonomy of Things ( AoT® )implementations with drones (airborne, underwater, ocean), robots and autonomous vehicles needs a secure chain for components - batteries, motors, fans, cables and sensors.

Major manufacturers of robots include China, Japan, Russia and the United States. Approximately 500,000 industrial robots were installed in 2025 ( 54% in China, 20% in the rest of Asia, primarily Japan, 16% in Europe and 9% in the U.S.). They perform well-structured tasks, including layout, factory automation, and task specific applications such as drilling, fastening, and material handling. China also dominates the manufacture of humanoid robots (87%) with the U.S. in second place (13%) . Humanoid robots are still not able to perform in less structured and constantly changing home and construction environments. For home construction, t he strategy employed is to build with robots in a factory environment and then install on-site. Boston Dynamics is probably the oldest robotics company in the U.S . Founded in 1992 as a spin-off from MIT, the company today fields quadricep ( Spot ), purpose-built ( Stretch ) and humanoid ( Atlas ) robots for security monitoring, warehouse and factory production tasks.

The drone ecosystem is composed of high-volume consumer imaging applications (dominated by China), enterprise autonomy (for example, monitoring critical infrastructure ), defense, logistics and aerial logistics and mobility. U.S. defense companies lead in the manufacture of high performance aerial drones. Turkey, India, China, Ukraine, Iran and Russia are active in the high volume, low cost, expendable drone space (for obvious reasons like war and border tensions). DARPA recently released a solicitation for next generation underwater drones that are able to move near the ocean floor with dramatically lower cost, size and weight, and able to work in challenging and contested communication, sensing and navigation environments. Technology approaches and regulatory environments are critical, as are costs (especially for expendable drones).

As countries compete to lead and dominate in physical AI, supply chain resiliency and self-sufficiency is paramount, given the global political and security environment, national security needs and economic/trade/tariff policies. Developing technology and controlling the manufacture of the end product is important. Equally critical is the development of a secure supply chain and a domestic manufacturing base to make components that build and operate physical AI systems.

Investments in vertical integration and volume manufacturing for its own robot drone and EV product make China a dominant supplier of the parts that go into these products as well - motors, batteries, printed circuit boards, fans, etc. It has used its dominance to cut off key supplies in many cases - like suspending battery exports to US drone manufacturer Skydio . AXT, a NASDAQ-traded materials company headquartered in Silicon Valley, manufactures its semiconductor wafers in China through a subsidiary, which needs Chinese government export licenses for Germanium and Gallium Arsenide wafers (the work horse materials for thermal imaging, optical detection and semiconductors). As AI demand accelerated, China also instituted license controls on InP wafer export s. InP is a critical material that powers lasers for fiber and free space optics for communications, data centers, physical AI and LiDAR (Light Detection and Ranging) sensing. China also recently denied or delayed visas for Indian drone company executives trying to visit to establish their product lines and supply chains, impacting their ability to move ahead and compete. This has been a driver for the country to establish its domestic manufacturing and supply chain for telecommunications, AI and semiconductors. Similar pressures exist in Europe, Canada and other parts of Asia like Japan and South Korea, all of who are making serious investments in developing domestic supply chains and manufacturing capabilities.

The need for a sovereign, secure supply chain for physical AI implementations has become an urgent issue for most countries. In the U.S. for example, the FCC (Federal Communications Commission) recently issued a memo titled “ National Security Determination on the Threat Posed by Foreign-Produced Advanced Robotic Devices”. It argues that physical AI technologies like robotics and drones for commercial and defense applications need to be developed and built domestically, including establishing a secure supply chain of components, sensors and batteries to support this effort. This was needed to maintain security, industrial competitiveness and defense leadership. J.P. Morgan recently initiated a $1.5 T “Security and Resiliency Initiative to boost critical industries”. In general, Wall Street is investing heavily in these areas, with the likes of Softbank, Goldman Sachs and Blackstone securitizing investments in physical AI and materials technologies.

On the defense front, scaling manufacturing capacity for critical weapons like Patriot missiles and the Joint Advanced Tactical Missile (JATM) is becoming critical as global tensions intensify. However, the bottleneck to this are some of the key precision components that make these missile work, for example, precision ball screws that convert rotary motion into linear motion for missile guidance require precise machining and are in short supply. These screws are used to control fin motion of the missile (Figure 1) for precision guidance. At present, the backlog is severe, limiting the manufacturing scale-up. Evaluating simpler screw designs would be useful, as would developing a more robust U.S. manufacturing base for this and other components (Figure 1):

Establishing Manufacturing and Design Competencies

On-shoring manufacturing capability is not easy - because of know-how, experience and the ability to match the low costs that Chinese suppliers provide (because of their manufacturing volumes, capital investments and years of experience). EnerVenue, a U.S. based battery company recently reversed its plan for volume manufacturing of its battery cells in Kentucky to China, primarily because of high costs, and manufacturing capital investments that would be required in the U.S. Lauf, a Norwegian designer of high performance road bicycles set up a U.S. based manufacturing plant in Virginia to avoid high U.S. import tariffs - great idea, except the supply chain for all the critical parts is based in China - forcing steep tariff payments anyway. Similar examples exist for critical physical AI technologies like drones and robots. There are drone motor suppliers in the United States, although the main focus is in defense and security, and driven by the need to be NDAA/Blue UAS compliant regulations for defense contracts. The supply chain for low cost consumer and enterprise drones is limited.

Before you can even engage in manufacturing, you need design expertise. After years of relying on established foreign suppliers, design know-how and expertise has to be built from scratch. Folding in supply chain control, manufacturing and cost constraints needs new and innovative thinking. The use of physics and geometry based AI designs is filling in this gap - to innovate quickly, consider practical constraints, evaluate a large number of design options and zero in on a couple of final options for initial prototyping and testing.

Monumo - AI-based Optimization of Engineering Systems

Monumo, based in Cambridge, United Kingdom, was founded in 2021. Apart from Cambridge (which houses the AI and physics groups), the company also has a location in Coventry that does prototyping and testing of designs (results of which are fed back into its AI engine). The company currently has ~30 employees and is revenue-generating.

Dr. Jarek Rzepecki is the CEO and co-founder of the company, which specializes in AI-driven approaches to solve complex multi-physics design challenges, with electric motors as a flagship application. Typically, the parts that go into such motors are designed separately. This makes it difficult to incorporate design trade-offs and overall system optimization. Monumo addresses this with its proprietary design agent, Anser® AI.

Figure 2 shows the workflow:

  1. Customer Requirements: customer, outlining the criteria they want to optimize and the manufacturing constraints a design must meet. For example, in the case of a new electric motor, a customer might want to minimize cost, reduce permanent magnet content, and maximize efficiency, while maintaining customer-specific manufacturing constraints.
  2. Calibration of results on the reference design: if the customer already has an existing motor or a new motor design in development, it is modeled within the Anser engine to match it’s performance characteristics. The customer verifies this aspect. If no reference design exists, Monumo is able to use its own internal design as a starting point.
  3. Anser AI-driven optimization and verification: at this point, the Anser AI engine kicks in, evaluating millions of design variations in a single optimization session lasting up to 72 hours. The designs evaluated are not random. Anser’s intelligence layer selects them to ensure the best design is found within the available time.
  4. Delivery of optimal results and trade-off data: Anser AI explores the relationships between different properties of the motor system and plots the resulting trade-offs across multiple Pareto fronts. This allows the customer to choose the optimal solution for their use-case. Appropriate CAD design files are then supplied to the customer for review and prototyping.

Dr. Rzepecki highlights that apart from cost reduction, performance optimization and motor component design, a key focus has also been to reduce the amount of rare earth materials for the motor magnet design. He also clarified that Anser AI is now transitioning into a mode where a customer can engage with it directly to generate and zero in on their designs. At this point, Proof-of-Concept (POC) demonstrations are moving to commercial contracts with significant subscription revenues in 2026. In addition to a seed round in 2024 and grants funding, the company has raised ~$14M to date and intends to do a Series A round in the near future. The initial focus on electric motor design was driven by the intent to focus on areas that can deliver impactful returns on investment. By way of reference, electric motors:

  1. Consume half the world’s electricity - small gains in motor performance, multiplied across billions of motors worldwide, translate into enormous reductions in energy use and carbon emissions.
  2. Exploiting the system-level gap: as opposed to traditional approaches where motor components have been designed in silos, Monumo designs the motor as an integrated system, unlocking performance gains that conventional engineering approaches cannot achieve.
  3. Reducing reliance on rare earth materials: ~ 80% of EV motors depend on rare earth magnets - expensive, scarce, and environmentally costly to mine. Monumo’s approach has already demonstrated > 20% reduction in magnet usage without sacrificing efficiency or performance.

Monumo recently completed a project with Hyundai Cradle on the design of electric motors for automobiles. It demonstrated how Anser AI, following the design flow in Figure 1 was able to deliver a motor design that reduced costs, accelerated innovation, reduced rare earth material content, and very importantly, ~20% lower costs. According to Dr. Rzepecki, “Hyundai’s clear commitment to staying at the forefront of innovation is paying dividends. Our most recent POC with the R&D team delivered alternative, fully optimized, lower-cost motor designs on a timeline impossible with conventional methods, proving that physics-grounded AI can add real commercial advantage even to a world-class engineering team’s process". According to Hyundai Cradle, "this successful collaboration marks an important step in evaluating how AI can be integrated into future mobility development. Beyond technical validation, it also creates opportunities for potential long-term collaboration and reinforces Hyundai’s commitment to staying at the forefront of next-generation EV innovation".

Neural Concept - Engineering Intelligence

A previous article discussed how Neural Concept (NC) uses its physics and experience- based AI design co-pilots to solve complex aerodynamic and car-cabin designs. The co-pilots use past designs and their performance as a base to generate new designs with different performance, style, cost and manufacturability constraints. Key automotive OEM customers include JLR and General Motors, while Mahle, a Tier 2 automotive supplier is also a customer. It also counts General Electric Vernova, Leonardo Aerospace, Eaton, Safran, Renault Group, and multiple Formula 1 teams as customers. Over the past 18 months, NC has generated 4X growth in revenues, and is growing its global presence including offices in Munich, New York and the Asia-Pacific region. It currently has ~100 employees. NC recently raised $100M in a Series C round, led by Growth Equity at Goldman Sachs Alternatives, with participation from existing investors.

In a quest to broaden its customer base, Neural Concept also recently established offices in South Korea (to tap into the vast automotive, electrification, shipbuilding, electronics and and semiconductor industries) and India (to accelerate engineering design efforts in automotive and industrial autonomy applications). Traditionally, the AI Co-Pilot is extremely efficient on simulations and design solutions with large complexity, applications like airflow modeling and low drag, stylistic automotive designs. The DARPA program for next generation underwater drones (mentioned above) is very well suited to Neural Concept capabilities. More recently, it is also engaging in design of less complex but critical parts for physical AI.

NC’s design co-pilot workflow is shown in Figure 3.

  1. An engineering request (input, left side of Figure 2) can come from a user, a connected PLM, CAD or CAE environment, enterprise knowledge, or a third-party AI model.
  2. The AI Agent Gateway routes the request into the Neural Concept platform, where the appropriate skills define the engineering workflow, tools execute the required geometry, physics and optimization tasks, and the infrastructure manages the applications, models, compute and data.

Skills : include reusable engineering know-how that defines how a task is performed. This includes design copilot skills, industry and project-specific methods, and skills embedded in CAx environments.

Tools : are the technical capabilities used to execute the workflow, including geometry tools, hybrid physics and geometry AI, optimization, embedded CAx and engineering software connectors .

Infrastructure: is the underlying management layer for applications, machine-learning and design operations, compute orchestration and engineering data.

3. Results return through the gateway for user review and validation, creating an iterative, human-in-the-loop workflow.

4. The AI Agent Gateway c onnects users, engineering systems, company knowledge and third-party AI models to the platform, then returns results for validation.

NC’s design co-pilot also works when no prior design knowledge exists. Using requirements as a starting point, the Skills and Tools components of the agent are able to generate multiple geometries through CAD programs (like CATIA) and apply FEA (Finite Element Analysis) to create design options that meet the requirements. This is especially critical as countries develop sovereign supply chains for electro-mechanical parts used in robots and drones, and for which they have no prior knowledge or experience.

Dr. Thomas Von Tschammer is a co-founder of Neural Concept, and General Manager of its U.S. Division. Per Dr. Von Tschammer, “we’re seeing AI reshape who owns critical design and manufacturing knowledge. Embedding AI directly into customer design and simulation workflows allows manufacturers become more vertically integrated on critical components, innovate faster, and grow know-how exponentially, instead of depending on external partners for it. In a moment where supply chains are under constant pressure, that ownership becomes a critical asset”.

One of NC’s recent customers is Cooper Standard , a global, publicly traded company, headquartered in Michigan, United States. It has annual revenues of ~$3B and employs ~22,000 employees globally. It specializes in molded and extruded parts for a variety of end applications (automotive is a primary market), and has codified its specialist engineering knowledge into NC’s design co-pilot that cuts custom tooling design time for its molded parts by 50%. This helps manufacturing companies to onshore and retool production locally without relying on a small number of experts or external tooling capacity. Figure 4 shows the types of systems the company specializes in, and the use of NC’s design co-pilots to rapidly design and build custom tooling to fabricate these parts.

GLĪD TECHNOLOGIES - Sovereign Physical AI For U.S. Logistics

Glīd builds dual-mode road-to-rail systems that move freight across road, rail, yard, port, and defense environments. Founded by Kevin Damoa, an Army veteran, the company was founded in 2022 and is headquartered in Riverside, California. A previous article described the company’s unique solution for a logistics vehicle that can “glide” frictionlessly between rail and road corridors for first and last mile logistics.

Injecting autonomy into both rail and road modes is a key focus of the company, enabling critical logistics operations in defense and commercial environments. Glīd’s platform carries a load over road and over rail, point to point, on infrastructure that exists today: no locomotive, no crane, no transfer. Glīd’s EZRA-1SIX, command and control layer, decides what moves, on which network and when, and is designed to tie the machine to the rules of the railroad it is running on. The advantages are immense - it makes short-distance, small batch cargo-on- rail economically feasible, reducing road congestion, fuel costs and pollution. Seamless road autonomy make first and last-mile commercial and residential deliveries viable, without any cargo transfers.

According to CEO Kevin Damoa, “ the vehicle, autonomy stack and intelligence are all developed through U.S. based partners and manufacturers. Production sourcing strategy is built around 100% U.S.-based supplier sourcing by planned BOM spend. The focus is on building the platforms through the American industrial base, from the chassis and fabrication through power, propulsion, rail systems, controls, autonomy hardware, sensors, compute, communications, electrical architecture, and final integration”. To be clear, not every semiconductor, battery cell, or subcomponent is physically manufactured in the United States today, although the supply chain is deliberately structured so that the companies that sourcing and integrating through are here in the United States. “ The point is to use the American industrial base to build the platform, then use the platform to strengthen the American industrial base” .

The AoT® revolution is proceeding rapidly in various defense and commercial applications that use robots, drones and vehicles. While leadership in the end systems is critical, equally important is the need to institute secure, sovereign supply chains to support this revolution, and maintain a competitive edge in this fast growing area of physical AI. Governments and industry across the globe are investing in establishing design and manufacturing infrastructure to support these supply chains.

AoT® is a Registered Trademark of Patience Consulting LLC. Visit aot.llc for more details.