For most of the 20th century, an arms race was easy to picture. Two superpowers accumulated weapons, a small group of governments set the requirements, and a handful of enormous defence contractors built the machines. The Cold War produced a relatively legible military-industrial order, dominated by the United States and Soviet Union and supported by industrial giants with the money, factories and security clearances needed to build advanced weapons. That structure has not disappeared, but it is being surrounded by something far more fragmented.

Global military spending reached nearly $2.7trn in 2024, up from about $1.25trn at the start of the century, while more than 100 countries increased their defence budgets. The more consequential change is happening beneath those headline figures: military technology, industrial capacity and strategic influence are spreading across a much larger field of countries and companies.

The result is a new kind of arms race, one that is becoming increasingly decentralised. Countries that once depended on American, Russian or European suppliers are building domestic defence industries and competing for export markets. Turkey has turned drones into a strategic export industry. India is pursuing defence self-reliance while trying to become a supplier in the Indo-Pacific. South Korea has emerged as a major exporter of tanks, aircraft and artillery. At the same time, the technology entering military systems increasingly comes from companies that would not have been considered defence contractors a generation ago. Space companies, satellite operators, semiconductor manufacturers, artificial-intelligence firms, software developers and commercial drone makers now occupy parts of the military supply chain. The arms race therefore has a growing number of participants, each controlling a different piece of the puzzle.

This is partly a consequence of technology. The weapons of the industrial age rewarded scale. A fighter aircraft, tank or ballistic missile required enormous factories, specialised machinery and long procurement cycles, creating natural advantages for a small number of established contractors. Many of the technologies driving today’s military competition have lower barriers to entry. Sensors, autonomy software, communications systems, artificial intelligence and electronic-warfare tools can be developed by companies whose capital requirements are tiny compared with those of a traditional aircraft manufacturer. A small firm can therefore become strategically important by solving one problem exceptionally well. The modern military system increasingly resembles an ecosystem of specialists rather than a single industrial machine.

The technology beneath the weapon is becoming as important as the weapon itself. An advanced military system may depend on a sensor from one supplier, processors from another, communications equipment from a third and software developed by several others. The wider technology stack runs through energy, raw materials, semiconductors, cloud computing, data centres, capital and specialist talent. Research from Cambridge’s Bennett School of Public Policy describes the AI supply chain in similar terms, with different countries and companies controlling distinct points of leverage rather than one actor possessing the entire system. A country can therefore become strategically important by mastering one layer of the chain, while a small company can become indispensable without ever producing a complete weapon.

Consider the semiconductor industry, where strategic influence can reside in a highly specialised company that most people have never heard of. ASML’s lithography machines are critical to the production of advanced chips, giving the Dutch firm leverage far beyond what its size might suggest. The Cambridge research points to a similar pattern across the photonics supply chain, where France’s Soitec, Israel’s Tower Semiconductor, America’s Lumentum and Sweden’s Sivers Semiconductors occupy different technical niches. None controls the entire chain. Their importance comes from the fact that the chain cannot easily function without a combination of specialised capabilities. Defence technology is increasingly developing along the same lines, creating a market in which a company can matter enormously without ever producing a complete weapon.

The rise of the specialist is changing how militaries can innovate. A traditional weapons programme might require a government to commit billions of dollars to a major contractor before it knows whether the underlying concept will work. Smaller technology companies can test narrower ideas, adapt their products quickly and compete against established suppliers. Drones have provided the clearest demonstration of this dynamic. The war in Ukraine has shown the military value of producing large quantities of relatively simple systems, with commanders able to accept rapid technological turnover because inexpensive platforms can be modified or replaced. The advantage can come from the ability to iterate and manufacture at scale rather than from owning the most sophisticated platform in the catalogue.

Ukraine has effectively turned this into a new form of military-industrial competition. Battlefield requirements can change faster than a conventional procurement programme can respond. A drone that performs well in one phase of a conflict can become vulnerable after an adversary develops a countermeasure, forcing manufacturers to alter software, communications, sensors or airframes. The feedback loop between combat, engineering and production is consequently becoming much shorter. That creates opportunities for companies that can move quickly, particularly those built around software and commercial manufacturing methods. It also gives militaries a reason to maintain relationships with many suppliers simultaneously rather than betting their future on a single programme.

The geographic spread of defence production reinforces this decentralisation. Turkey, India and South Korea illustrate three different versions of the same strategic impulse: countries that once relied heavily on foreign suppliers increasingly want to manufacture advanced weapons themselves and sell them abroad. Turkey’s Bayraktar TB2 has become a prominent export product and an instrument of Ankara’s foreign policy, while India’s BrahMos missile has opened a market in the Philippines and supported New Delhi’s effort to become a regional security provider. South Korea has gone further, using its industrial capacity to secure major export contracts in Europe and elsewhere. These countries are competing for customers, influence and strategic relevance alongside the established suppliers.

Defence exports are therefore becoming a form of diplomacy for a much larger group of states. Selling a missile or drone creates a relationship that can persist through training, maintenance, upgrades, spare parts and future procurement. It can also create political influence because the customer becomes tied to the supplier’s technology and industrial ecosystem. Turkey’s willingness to sell military systems with fewer political conditions than some Western suppliers has helped make its products attractive to governments with limited diplomatic options. South Korea’s ability to deliver sophisticated equipment quickly while maintaining compatibility with NATO standards has given it another advantage. The weapons market is becoming a market for political relationships as well as military hardware.

The same process is taking place inside individual countries. Governments are increasingly acting less like customers choosing from a fixed catalogue and more like investors assembling technological ecosystems. They fund start-ups, subsidise production, protect critical suppliers and try to secure access to foreign components. Cambridge’s research describes Britain’s Sovereign AI unit as an example of this approach, operating partly as a venture-capital mechanism to invest in companies that could strengthen a particular national advantage. Vietnam is pursuing another niche by attempting to expand its role in semiconductor assembly, packaging and testing. Such policies reflect a broader understanding of strategic autonomy: control over every part of the supply chain is unrealistic, but control over valuable parts can create leverage.

This creates a different relationship between states and defence companies. The government remains the ultimate customer and often the source of capital, but the boundary around the defence industry has become porous. A satellite operator can provide military communications without being a conventional defence contractor. A cloud company can host data used for military operations. An AI company can develop software for intelligence analysis or targeting. A commercial drone manufacturer can find its products adapted for combat. As private-sector assets become embedded in military systems, the distinction between civilian and military infrastructure becomes harder to maintain. Analysts cited by Anadolu describe this erosion of the traditional boundary as one of the clearest characteristics of modern conflict.

The strategic consequence is a shift from competition between industrial giants towards competition between networks. A European military might combine American software, Israeli sensors, Korean platforms and components produced elsewhere. A Middle Eastern state can source different capabilities from several suppliers while developing its own specialised technologies. A smaller country can gain influence by controlling one valuable component while relying on partners for the rest. Cambridge’s research describes this broader model as “balanced interdependence”, in which countries specialise in different parts of a technological system and use those positions to build collective bargaining power. In defence, such networks could become as important as traditional alliances because they determine who can build, maintain and upgrade a military system.

For governments, the attraction is obvious. A diversified supplier base can reduce dependence on a single contractor or foreign power. It can also create competitive pressure among companies, encourage faster innovation and allow militaries to replace underperforming technologies without abandoning an entire weapons programme. For companies, meanwhile, the opportunity is enormous. Defence budgets are expanding, countries are seeking domestic industrial partners and military customers are increasingly willing to test technologies that would previously have spent years inside laboratories. The result is a much larger addressable market for firms operating at the edges of traditional defence.

The risks are equally substantial. More producers mean more potential exporters, while more customers seeking strategic autonomy can intensify regional competition. The Atlas Institute argues that the parallel development of military-industrial capabilities by emerging powers can inflame regional rivalries and increase the availability of weapons outside the traditional supplier networks. It also points to the possibility that countries with fewer export restrictions could provide governments with alternatives to established Western suppliers. As more states acquire the capacity to produce and sell advanced systems, the number of actors capable of influencing a regional security balance grows.

Arms control becomes harder in such an environment because the object being controlled is increasingly difficult to define. A treaty can regulate a missile, aircraft or warhead. It is much harder to regulate a software update, an AI model, a commercial satellite constellation or a semiconductor that can be used in both civilian and military systems. The problem is compounded when private companies own critical infrastructure. Governments may impose rules on states, but they cannot easily control every commercial actor whose technology can influence a conflict. The institutional architecture of arms control was built around a world of governments and weapons programmes. The new arms race is increasingly organised around supply chains and technological ecosystems.

The numbers suggest that this fragmentation is developing alongside a broader global military buildup. Global military expenditure has reached its highest level on record, with spending increasing across almost every region. Europe alone recorded a 17% increase in 2024, while the Middle East rose by 15%. More than 100 countries increased their defence budgets, and the ten largest spenders still accounted for most global expenditure. Yet beneath this concentration of money sits a much wider distribution of industrial ambition, with emerging powers building their own production capabilities and private companies competing to supply increasingly specialised technologies.

The defining feature of the next arms race may therefore be its lack of a clear centre. America and China will remain enormously important, and the world’s largest defence contractors will continue to dominate major programmes. But around them is forming a dense layer of states, start-ups, specialist manufacturers and technology companies that can alter military capabilities without controlling an entire arsenal. Turkey can change the economics of drone warfare. South Korea can reshape the European weapons market. A semiconductor company can become a strategic chokepoint. A small software firm can alter how a military uses a weapon built by a much larger contractor. The competitive landscape is becoming broader because military power itself is becoming more modular.

The Cold War produced an arms race of superpowers and industrial giants. The emerging era is producing an arms race of ecosystems, in which countries compete to build capabilities and companies compete to supply the pieces. The market is becoming more dynamic, with more routes into defence, more suppliers competing for military customers and more countries using industrial capacity to pursue strategic influence. It is also becoming harder to map. There may no longer be two arsenals staring at each other across a divide. Instead there are hundreds of companies, dozens of states and thousands of technological dependencies pulling the system in different directions.

The question for the next decade is therefore less who will win the arms race than who will control the networks through which the race is fought.