Physical AI and the Future of Aerospace and Defense Manufacturing

Goldman Sachs expects the next wave of artificial intelligence investment to move beyond software and into the physical economy. For aerospace and defense manufacturers, the opportunity lies in connecting AI with factories, equipment, supply chains, and the skilled workers who produce mission-critical systems in the real world.

For much of the recent artificial intelligence boom, attention has centered on large language models, digital assistants, software development, and the companies supplying the computing power behind them. That may represent only the first stage of AI’s economic impact. In a June 26, 2026, Axios article, Madison Mills reported that Goldman Sachs bankers expect the next major wave of AI investment to spread into the “physical economy.” This includes factories, utilities, mines, energy facilities, data centers, and other asset-intensive parts of the global economy.

Goldman Sachs estimates that approximately $7.6 trillion could be invested globally in AI infrastructure between 2026 and 2031, including computing capacity, data centers, and power generation. Infrastructure, however, is only part of the story. The larger industrial opportunity may emerge as companies apply AI to the physical systems that manufacture products, move raw material, maintain equipment, and support critical infrastructure. For aerospace and defense companies, this transition could be especially significant because these industries operate complex production systems, manage extensive supplier networks, and manufacture products that must meet demanding standards for quality, reliability, security, traceability, and regulatory compliance.

The federal government has also made AI leadership a national economic and security priority. In 2025, President Donald Trump signed four major executive orders that form a central part of the administration's AI agenda. The first directed the development of a national action plan to strengthen American leadership in AI. Three additional orders issued alongside America's AI Action Plan focused on accelerating federal permitting for data centers and related infrastructure, establishing standards for the federal government's procurement of large language models, and promoting exports of the American AI technology stack.

Together, these actions reflect an effort to ensure that the United States leads the global AI race not only through software and advanced models, but also through the energy systems, semiconductor capacity, data centers, industrial infrastructure, workforce, and international partnerships required to deploy AI at scale. That policy direction reinforces the importance of the physical economy and creates direct implications for aerospace, defense, energy, construction, and advanced manufacturing.

The first phase of generative AI demonstrated how technology could help analyze information, generate content, write code, and automate administrative work. Physical AI expands those capabilities into environments where digital intelligence interacts with machinery, equipment, materials, and industrial processes in the real world. That does not mean every factory will soon be filled with autonomous robots. In aerospace and defense, the most immediate applications are likely to be more practical. AI can help manufacturers understand what is happening across their operations, anticipate problems, and make better decisions before those problems affect cost, quality, delivery, or customer satisfaction.

An AI system could analyze machine data to identify signs of an upcoming equipment failure. It could review production performance to recognize the early stages of a bottleneck or examine supplier, inventory, and program information to determine which parts are most likely to delay final assembly. The technology creates value when it helps an organization reduce costs, prevent downtime, improve quality, protect a production schedule, or respond more effectively to a supply chain disruption.

Aerospace and defense manufacturing environments are particularly difficult to manage. Production volumes may be lower than in consumer manufacturing, while product complexity and customization are substantially higher. A single aircraft, spacecraft, missile, vehicle, or defense system can depend on thousands of components produced across multiple tiers of suppliers. AI-supported production planning and inventory management could help manufacturers account for changing material availability, labor constraints, machine capacity, order priorities, raw material requirements, and supplier performance.

Instead of relying entirely on static schedules or manually assembled spreadsheets, planners could use AI to identify potential disruptions and evaluate alternative production scenarios. These tools could help determine which shortage will have the greatest effect on the production schedule, where work-in-process is beginning to accumulate, which orders are at risk of missing their delivery dates, and whether production could be shifted to another machine, facility, or supplier. AI will not eliminate production constraints, but it can help leaders identify problems earlier and respond with better information.

Effective inventory management is particularly important because manufacturers must balance competing risks. Carrying too much inventory ties up working capital, increases storage expenses, and creates the possibility that specialized components will become obsolete. Carrying too little inventory leaves programs vulnerable to supplier delays and material shortages. AI can analyze demand, lead times, consumption patterns, production schedules, and supplier performance to recommend appropriate inventory levels. Better decisions can produce cost savings while reducing the risks associated with shortages and excess inventory.

Quality is another promising application for physical AI. Aerospace and defense components must satisfy strict engineering requirements, and defects can create substantial financial, operational, and safety consequences. Computer vision systems can assist inspectors by analyzing images for potential surface defects, dimensional inconsistencies, assembly errors, or other irregularities. AI can also review historical quality data to identify recurring problems connected to particular machines, suppliers, materials, processes, or production conditions.

These technologies should not be viewed simply as replacements for experienced quality professionals. Their more immediate value may come from helping those professionals concentrate on the highest-risk issues. A system that rapidly evaluates large volumes of inspection information can flag potential problems for human review and make patterns easier to recognize. When properly validated and integrated into an established quality-management system, AI can improve inspection consistency, strengthen traceability, and support a process that ensures compliance with customer, contractual, regulatory, and certification requirements.

Predictive maintenance offers another practical application. Unplanned equipment downtime can disrupt an entire manufacturing schedule, particularly when a facility depends on specialized machines with limited backup capacity. Predictive maintenance uses information such as vibration, temperature, pressure, electrical consumption, and operating history to recognize patterns that may indicate equipment deterioration. AI can provide continuous monitoring of these signals and estimate when maintenance may be required, allowing a manufacturer to service equipment during a planned production window instead of responding after a failure.

This approach could extend the useful life of industrial equipment, improve capacity utilization, and reduce costs associated with emergency maintenance and lost production. The same principles extend beyond the factory. Aerospace and defense operators can use aircraft, vehicle, and system-performance data to support maintenance planning, improve readiness, and reduce unnecessary downtime. Identifying potential failures before they occur could generate cost savings while improving the availability of critical assets.

Supply chain management may represent one of the greatest opportunities for AI in aerospace and defense. A delay involving one casting, forging, electronic component, or specialized raw material can affect multiple production programs. Many companies already possess large amounts of supplier data, but that information may be divided among procurement platforms, quality records, inventory systems, emails, spreadsheets, and individual business units. This fragmentation makes maintaining data difficult and prevents leaders from developing a complete view of supply-chain performance.

AI can help organize and analyze information from these sources, but its effectiveness depends on strong data governance, data standardization, and data accuracy. Companies must establish clear rules for collecting, storing, updating, protecting, and using operational data. Data standardization allows information from different facilities, programs, and suppliers to be compared consistently, while maintaining data accuracy ensures that AI systems are working with dependable information.

Creating a single source of truth for supplier, inventory, production, and quality data can improve decisions throughout the organization. When procurement, operations, engineering, and program-management teams work from the same reliable information, they can identify risks earlier and coordinate their responses more effectively. Greater visibility could help companies conduct continuous monitoring of supplier lead times, delivery performance, quality trends, financial conditions, available capacity, and geographic exposure.

Manufacturers could then select suppliers based on a broader set of operational factors rather than price alone. Quality, delivery reliability, cybersecurity, financial health, capacity, location, regulatory compliance, and the ability to support future production rates may all influence sourcing decisions. AI can help evaluate these factors, but experienced procurement and operations professionals must remain responsible for interpreting the results and making final decisions.

The objective should not be to create another dashboard for employees to monitor. The real value comes from giving decision-makers enough warning to act before a supply chain disruption reaches the production line. That could mean placing an order earlier, qualifying an additional supplier, adjusting inventory, supporting a critical vendor, or redesigning a component to reduce dependence on a constrained source. Better visibility can also strengthen supplier relationships by allowing manufacturers and suppliers to address capacity, quality, and delivery problems collaboratively.

AI could also increase the value of digital twins, which create virtual representations of physical products, equipment, or production systems. When combined with AI, these models can help engineers evaluate alternative designs, production configurations, maintenance strategies, and capacity investments before making changes in the real world. A manufacturer could use a digital twin to analyze how a production line might respond to a higher build rate, determine whether additional tooling or labor would be required, and identify where bottlenecks might emerge.

At the product level, AI-supported modeling could help engineering teams evaluate design alternatives more quickly. Nevertheless, aerospace and defense companies must maintain rigorous control over design authority, configuration management, testing, documentation, and certification. A well-designed governance process ensures compliance and keeps qualified employees accountable for final engineering decisions. AI can support professional judgment, but it does not remove the need to verify that every product meets its applicable requirements.

Although the promise of industrial AI is significant, companies should resist treating it as a shortcut around fundamental operating problems. AI cannot compensate for inaccurate inventory records, inconsistent processes, poor data governance, unstable production schedules, weak supplier relationships, or limited supply-chain visibility. If the underlying information is incomplete or unreliable, an AI system may produce recommendations that appear sophisticated but are operationally flawed.

Data accuracy is therefore not simply an information-technology concern. It is an operational requirement. Maintaining data across complex aerospace and defense programs requires clear ownership, consistent definitions, disciplined processes, and regular verification. Establishing a single source of truth reduces the risks created by conflicting spreadsheets, outdated supplier information, and disconnected business systems.

Before scaling AI, manufacturers need clearly defined operational problems, reliable data, standardized processes, appropriate cybersecurity controls, measurable objectives, and employees who understand how to use and challenge the system. This foundation is especially important in aerospace and defense, where controlled technical data, intellectual property, national-security requirements, and regulatory compliance create risks that may not be present in other industries.

The organizations that benefit most from physical AI may not be those that launch the greatest number of pilot programs. They will be those that select valuable problems, prepare their operations, and scale the applications that produce measurable results. Instead of announcing a broad objective to adopt AI, a manufacturer might begin by reducing inspection time for one component, improving inventory management, increasing schedule adherence at one facility, or predicting downtime for a constrained machine.

Results should be evaluated using measures such as lead time, first-pass yield, equipment availability, inventory levels, labor productivity, on-time delivery, customer satisfaction, and the cost of poor quality. Human oversight must remain central throughout the process. Employees with direct knowledge of the product, production process, customer requirements, and supplier relationships should help design, validate, and improve the system. Continuous monitoring will also be necessary to confirm that AI systems remain accurate, secure, and effective as operating conditions change.

Goldman Sachs’ physical-economy thesis points toward a broader shift in how companies think about artificial intelligence. The next stage will not be defined solely by which organization has access to the most advanced model. It will depend on which companies can connect technology to physical operations and convert its capabilities into better performance in the real world.

For aerospace and defense manufacturers, that means applying AI to engineering, production, quality, maintenance, inventory management, supply chain management, supplier relationships, and workforce decision-making. The opportunity is substantial, but technology alone will not determine the winners. Successful adoption will require strong processes, reliable data, secure systems, capable suppliers, workforce involvement, and disciplined execution.

The next AI boom may reach the factory floor. Aerospace and defense companies that prepare their operations today will be in the strongest position to turn that investment into greater productivity, cost savings, customer satisfaction, supply-chain resilience, and long-term industrial competitiveness.