Pentagon wants to fix how America makes the steel for its weapons

DARPA, the Pentagon’s advanced research agency responsible for some of the most consequential technological breakthroughs in American military history, has issued a request for information targeting metals producers, ceramics manufacturers, and composites makers with an unusual proposition: share your production data, and we will give you AI tools that make your factory perform better than it ever has, at no cost to you.

The request for information, published June 25, 2026, under the designation DARPA-SN-26-81 and titled “Revolutionizing Industrial Scale Materials Processing,” seeks engagement from industrial producers of cast, rolled, and forged metals, bulk ceramics, and composite materials to help DARPA design a future research program aimed at eliminating the unpredictability that currently plagues large-scale materials manufacturing. Responses are accepted through July 24, 2026.

The problem DARPA wants to solve is one that sits at the intersection of manufacturing economics and national security in ways that are not immediately obvious. When engineers design aircraft, missiles, ships, or armored vehicles, they cannot use the actual average performance of the materials they specify, because any given batch of aluminum plate or titanium forging will vary slightly from the mean depending on subtle fluctuations in composition, heat treatment temperature, rolling pressure, and dozens of other interdependent production variables. To guarantee that a finished weapon system will not fail, designers must use the minimum expected property values, the worst-case performance numbers published in materials standards documents, rather than the typical or peak performance the material actually delivers. That conservative approach is structurally necessary but operationally costly: it means that virtually every piece of structural metal in an American fighter jet, missile airframe, or naval vessel is stronger, lighter, or tougher than the design actually required, because the design had to assume the worst.

DARPA’s solicitation illustrates the scale of this waste with a concrete example drawn from aerospace aluminum. Many alloys in the 7000 series, the family of high-strength aluminum used extensively in aircraft structures and missile bodies, have a potential performance margin exceeding 15 percent above minimum specification values at fixed plate thicknesses and standard heat treatments, according to the DARPA document. That 15 percent gap represents weight that could be removed from every aircraft and missile built from that material, or performance headroom that designers cannot currently access because the statistical uncertainty in the manufacturing process makes using it too risky. Across the full volume of American defense production, that margin is not academic: it translates directly into aircraft that fly faster or carry more payload, missiles that fly farther or carry larger warheads, and vehicles that weigh less without sacrificing protection.

The reason that margin cannot currently be closed is that large-scale materials processing remains, in DARPA’s characterization, more art than science at the production floor level. The complex, interconnected nature of production variables, ranging from slight composition fluctuations and heat treatment temperatures to rolling pressures, forging temperatures, and even environmental conditions inside the production facility, creates an optimization problem that has defeated conventional engineering analysis. Because these factors are highly interdependent, improving one step in the process does not guarantee a better final product if a subsequent step introduces new variations. The practical result is that the tacit, intuitive knowledge that experienced operators develop over decades of watching how a specific furnace behaves on a cold morning or how a particular batch of raw material responds to a specific rolling sequence represents the primary mechanism for managing process variability at most large-scale producers, and that knowledge cannot be replicated, scaled, or transferred to new workers as the existing generation of skilled operators retires.

DARPA’s proposed solution is a physics-informed computational modeling approach that would analyze multi-year production data from large-scale manufacturers, identify the causal relationships between process variables and product outcomes, and translate the tacit knowledge of experienced operators into quantitative models that any trained technician can apply consistently. The agency describes this as using material informatics and physical modeling to demystify the art of industrial scale production, with the explicit goal of producing tools tailored to each manufacturer’s specific equipment, processes, and historical data that can be implemented within existing facilities without capital investment. DARPA is asking producers who manufacture more than 10 metric tons (22,046 lb) per year to participate, targeting the scale of production where defense-relevant materials actually get made.

America’s ability to produce advanced military hardware at scale depends on a domestic industrial base that can reliably manufacture high-performance structural materials faster and cheaper than adversary industrial systems. China has invested heavily in its own materials manufacturing capacity, and the gap between Chinese and American production economics for aerospace-grade aluminum, titanium, and high-performance composites is one of the less-discussed but consequential dimensions of the two countries’ defense industrial competition. A DARPA program that could push American producers to the frontier of what their current equipment and processes can deliver, without requiring new capital investment, would strengthen that competitive position without the long lead times that building new production capacity requires.