The hidden economy behind every AI prompt

The majority of people associate AI with a chatbot, an image generator or a coding assistant. But these are merely the buildings that have a face. There is another economy behind them, an invisible economy that operates in the name of data, computing power, electricity, human work and public infrastructure. We use it every day, but do not fully appreciate who’s creating the value, who’s taking the value, who’s really paying for the value.

Each ‘prompt’ from the AI generates something bigger. But there is an invisible economy to getting that seemingly easy answer; chips and electricity, water and cloud infrastructure, data and human labour and billions of dollars of investment. AI is the next Industrial Revolution. However, there is an important point to the comparison. All industrial revolutions had their own infrastructure. The infrastructure of AI is not so visible. It has factories that are data centres. It’s using data as its raw material. It operates on electric fuel. It has an assembly line process, and a number of thousands of human workers train, test, and fine-tune intelligent systems.

This hidden infrastructure is growing exponentially. PwC has estimated that AI will have the power to add up to US$15.7 trillion to the global economy by 2030. These figures are often touted as the harbingers of the promise of AI. Much less consideration is given to the source of that wealth, the ownership of infrastructure that produces it, and who will benefit from it.

The solution is as simple as a question. When a user requests an email be drafted or a spreadsheet be analysed by AI, the request goes through extremely fast networks to data centers with specialized processors. These are not regular computers, but powerful graphics processors (GPUs) that can calculate trillions of calculations per second for thousands of dollars. The reason why the response is so quick is that gigantic investments have already been made in semiconductor manufacturing, cloud computing and networking infrastructure.

 It has just emerged that intelligence isn’t light. Nor is it free of pollution. According to the International Energy Agency (IEA), data centres used approximately 415 terawatt-hours of electricity in 2024, which amounted to about 1.5% of global electricity consumption. Each prompt thus comes with an invisible energy cost and connects AI policy with the discussions around renewable energy, grid resilience and sustainable infrastructure.

Another little-talked-about expense comes even further down the list: people. Although the popular notion is that AI generation will replace human jobs, it still relies on a huge amount of people to label images, check translations, moderate and filter out harmful content, and evaluate the responses of the models. Much of their work is hidden in plain sight, frequently spread out across the developing rest of the world, done away from the spotlight of AI success. All polished AI answers have a human judgment component which users are not aware of.

Traditionally, economists have called such uncompensated expenses externalities, and AI is ushering in a new era of digital externalities. Costs that the user may not have in mind include those in the electricity grid, water resources, cloud infrastructure, copyright issues and human labour. The costs of intelligence are distributed through a complex economic system and thus seem free. The invisibility applies to value creation as well. All interactions with AI create economic value. The accuracy of models is enhanced by users’ questions.

AI becomes a core part of businesses as a tool. Cloud providers rent computations. Manufacturers of chips are providing ever more advanced equipment. In this fledgling value chain, the most lucrative positions go to companies who own their data and infrastructure for computation. This is a chance and a strategic dilemma for India. Through Aadhaar, UPI and DigiLocker, the country has created one of the most revered ecosystems of digital public infrastructure in the world that can serve millions while scaling, showcasing how digital systems can scale. India also has more than 900 million Internet users, which produce one of the world’s most prolific sources of digital data. However, having a digital scale won’t guarantee AI leadership.

Without the ability to develop its own AI, India is at risk of becoming a data exporter and an AI importer. The next big step in digital policy needs to go beyond promoting the use of AI. It needs to prioritise the enhancement of domestic computing power, semiconductor manufacturing, providing affordable computing to start-ups, investing in AI research, establishing high-quality public datasets and establishing clear regulations that engender trust. The debate surrounding AI has mostly centered on the question of whether AI will lead to job displacement. That’s a very important question, but possibly not the most important one.

According to the International Monetary Fund, AI could impact approximately 40% of all jobs around the world. Equally important, though, is who will be the owners of the infrastructure on which AI creates wealth. Its long-term consequences will not only be related to innovation, but also to institutions, investment and public policy. Each AI prompt is thus more than just a digital conversation. It is an economic exchange that is part of an extensive global system of energy, infrastructure, investment and human effort. Although the technology is invisible, its economic impact is not. Each prompt has an invisible cost; its value is invisible, and its infrastructure is invisible. The hidden economy is the first step towards a technologically transformative, economically fair, and viable AI revolution.

The writer is a Professor in the Department of Commerce, CHRIST (Deemed to be University); Views presented are personal.