Who Builds AI?

Every time you ask a voice assistant a question, get a recommendation from a streaming service, or see a photo automatically tagged on your phone, artificial intelligence is at work. But who made that AI? Who decided how it would behave, what data it would learn from, and what purposes it would serve? These questions matter more than most people realize, because the answer is: a surprisingly small group of organizations.

The Big Tech Companies

A handful of the world's largest technology corporations have become the dominant builders of AI systems. Companies like Google, Microsoft, Meta, Amazon, and Apple invest billions of dollars every year in AI research and development. These organizations have three things that make AI possible at scale: massive collections of data from their users, enormous computing infrastructure, and the money to hire the best researchers in the world. Google's DeepMind lab produced AlphaFold, a system that predicted the three-dimensional shapes of nearly every known protein — a breakthrough that could accelerate drug discovery for decades. Microsoft invested heavily in OpenAI, the company behind ChatGPT. Meta (which owns Facebook and Instagram) built large language models and released some of them publicly. Amazon's AI powers warehouse robotics, cloud computing services used by millions of businesses, and the Alexa voice assistant. These companies build AI both for their own products and as services that other companies rent and use.

Research Labs and Universities

Not all AI comes from corporations chasing profit. Research universities and independent labs have been central to AI's development from the very beginning. Stanford University, MIT, Carnegie Mellon, and many others have produced foundational research that the whole field depends on. Independent research organizations like OpenAI (before it became more commercial), Anthropic, and the Allen Institute for AI focus on studying AI safety and building systems that are safer and more reliable. These labs often publish their findings openly so the broader research community can build on them. Government-funded research has also played a huge role. DARPA — the Defense Advanced Research Projects Agency — funded early work on the internet and on self-driving vehicles. The National Science Foundation funds AI research at universities across the United States. In the European Union, the Horizon program funds AI research across member countries.

Governments as AI Builders

Several national governments have become major builders and funders of AI themselves, not just as regulators but as active developers. China's government has made AI a strategic national priority, investing heavily in facial recognition, predictive policing, and smart city systems. The United States military uses AI for logistics, surveillance analysis, and autonomous systems. The European Union is funding AI development while simultaneously writing rules to govern it. This government involvement means that AI is not only shaped by market forces — it is also shaped by national security interests, diplomatic competition between countries, and differing political values about privacy, surveillance, and civil liberties.

Why This Concentration Matters

When AI was in its early stages, many researchers and students around the world contributed to it. As the technology became more powerful, the requirements — data, money, computing hardware — grew dramatically, narrowing who could participate. Today, training a large-scale AI model can cost tens of millions of dollars and require specialized computer chips. This is not an activity most universities, small companies, or governments in lower-income countries can afford. The result is that the organizations shaping the most powerful AI systems are concentrated in a small number of wealthy countries — primarily the United States and China — and within those countries, in a relatively small number of cities and institutions. This concentration is not inevitable. Choices made by governments, researchers, and companies about what to share, what to open-source, and how to fund broader participation will shape whether AI development becomes more or less inclusive over time.