Artificial intelligence is changing warfare not only technologically but also economically. A new industrial ecosystem is growing around military AI. Current analyzes classify which companies are involved along the entire value chain. From cloud infrastructure to basic models to specific AI applications: In the following overview we will explain to you how the industry is structured.
Artificial intelligence is considered one of the key technologies of our age and therefore also of modern warfare. From analyzing large amounts of data to autonomous systems to supporting military decisions, it is already changing the way armed forces plan, operate and respond to threats.
At the same time, a new global industrial landscape is emerging in which defense companies, tech companies and specialized start-ups are increasingly working together. Researchers from the independent Stockholm International Peace Research Institute (SIPRI) have analyzed these connections.
A background paper was created that provides an overview of the military AI industry. This is intended to give political decision-makers, the media and the interested public an understanding of the diversity of the products, actors and relationships involved.
Where is AI used in the military?
Artificial intelligence is already being used in a variety of areas in our society. However, this is not a single technology, but rather a collective term for a variety of applications – and this also applies to the military sector.
AI applications are developed and provided by different companies. According to the SIPRI report, the military AI industry cannot therefore be clearly defined.
Many companies involved in military AI developments do not focus on creating AI models per se. Rather, numerous actors along the entire technological value chain interact with one another.
In their background paper, the SIPRI researchers have therefore divided the military AI industry into three areas. In a “stack” model, they arrange the companies and technologies involved along the different levels of the value chain.
These areas have a role for military AI
In their stack model, the SIPRI researchers represent the layered ecosystem of software and hardware that makes AI applications possible in the first place. The lowest level contains hardware components and the infrastructure.
This includes physical computing resources such as computer chips, data centers or fiber optic cables. But this also includes energy networks and storage in cloud infrastructures from Microsoft Azure or Amazon Web Services.
The second level includes basic software and programming. These include, for example, basic AI models such as large language models, frameworks for programming and training, and data pipelines that convert data and algorithms into machine intelligence. This layer provides the technological foundations on which downstream AI applications are built.
According to the researchers, this area is crucial because it determines factors such as the security and reliability of the resulting applications. Examples of these base models are OpenAI’s GPT model or Anthropic’s Claude model.
The third level of the stack model describes the scope of application. Here, AI is integrated into domain-specific systems designed to perform tasks for military end users.
It is no longer just basic models that are used here, but rather specialized AI. These include models that solve specific, clearly defined tasks and define the interaction between humans and AI systems.
Who makes military AI? Actors from the chip to the deployment system
“The military AI industry is not only diverse, but is also subject to rapid change,” write the SIPRI researchers in their background report. Military AI is an important business area in which many companies want to participate.
These include both established and new companies, which has created a diversified ecosystem. The researchers divide the players into four categories based on their business models.
1st category: Defense Primes – why the big defense companies are involved
The first category includes the so-called “defense primes”, i.e. large, established defense companies. These have direct, multi-year government contracts for military programs and capabilities.
They are behind complex supply chains and provide sophisticated systems such as fighter jets, armored vehicles, submarines and air defense systems. These major defense companies are eager to keep up with the latest advances in AI.
To do this, for example, they develop partnerships with start-ups and other companies with specific AI skills or commission them as subcontractors. An example of this is the Swedish company Saab, which took over the US company CrowdAI in 2023. Or the French arms company Safran, which took over the company Preligens, which specializes in AI-supported geodata analysis, in 2024.
2nd category: Neoprimes – develop start-ups instead of waiting
In the second category there are so-called “neoprimes”. These are start-ups that were founded much later than the established defense companies. These companies usually specialize in the provision of software and data products and offer corresponding services for the defense market.
The SIPRI researchers see the biggest difference between “Defense Primes” and “Neoprimes” in their approach to innovation. While the established defense companies wait for concrete needs from defense ministries, Neoprimes invest in their own research and development and market finished products to military customers.
These companies include, for example, manufacturers of autonomous drones such as Skydio or developers of technologies to defend against swarms of drones such as Epirus. Better-known companies include leading providers of battle management software that aggregates and analyzes data to support operational decision-making, such as Anduril, ShieldAI and Helsing.
3rd category: Big Tech – Cloud and infrastructure as military leverage
The third category is “Big Tech”, i.e. the globally dominant technology companies such as Microsoft, Google, Amazon, Meta and Oracle. These companies offer a wide range of digital products and services and operate in both civilian and military markets.
“Big Tech” primarily offers infrastructure for the military sector, such as cloud platforms or data centers. Meta and Google, for example, have laid deep-sea fiber optic cables to support their cloud computing services. They have also been linked to military applications of AI.
4th category: Foundation Models – from distance to partnerships
In the fourth category there are providers of so-called foundation models, i.e. basic models in the field of AI. These are companies that, like “neoprimes” and start-ups, are still relatively young in the defense sector.
Many base model vendors originally opposed military use of their AI tools, but that stance has changed over the past two years.
One of the most prominent examples of this category is OpenAI, which lifted the ban on using its products for “military purposes and warfare” in January 2024. Anthropic took a similar route that same year. To provide AI models for military use, the company partnered with Palantir and Amazon Web Services.
Also interesting: