The five layers of the AI economy

Artificial intelligence is often discussed as a single technology, but the industry that supports it spans a much broader economic ecosystem. According to the United Nations Conference on Trade and Development's Technology and Innovation Report 2025, the global artificial intelligence market is projected to expand from $189 billion in 2023 to $4.8 trillion by 2033, accounting for one-third of the market for frontier technologies – a category of advanced technologies expected to generate new opportunities for economic development, sustainability and governance. 

The scale of investment reflects the complexity of that ecosystem. According to Stanford University's April 2026 AI Index Report, global corporate AI investment reached $581.7 billion in 2025, while private companies produced nearly 90 percent of notable AI models. Beneath the applications familiar to consumers lies a stack of interconnected layers, each performing a different function and populated by a distinct set of companies, industries and countries. Examining those layers offers a clearer understanding of how artificial intelligence operates and why governments from the Gulf to East Asia are competing for a place within this rapidly expanding ecosystem.

Energy powers the entire system

Artificial intelligence begins with electricity. Before a model can generate text, recognize images or write code massive amounts of energy are required to train and operate the systems behind it.

According to the International Energy Agency's April 2025 report Energy and AI, data centers consumed approximately 415 terawatt-hours of electricity worldwide in 2024. By 2030, consumption could reach about 945 terawatt-hours, more than the current annual electricity use of Japan. Data centers are expected to account for almost half of the growth in U.S. electricity demand over that period. 

This layer is occupied less by technology firms than by utilities and countries with abundant energy resources. The United States benefits from extensive natural gas production, while Canada and Norway possess extensive hydroelectric power. Gulf states are also trying to turn their energy advantages into AI assets. Saudi Arabia and the United Arab Emirates have announced multibillion-dollar initiatives aimed at attracting data centers and building domestic AI capabilities. Without access to large and reliable supplies of electricity, none of the higher layers of the AI stack can function.

Semiconductors provide the computing power

Electricity alone does not create intelligence. AI systems depend on highly specialized chips capable of performing enormous numbers of calculations simultaneously.

Graphics processing units, or GPUs, have become the engines of modern artificial intelligence. Nvidia has emerged as the dominant supplier, briefly becoming the world's most valuable publicly traded company in 2025. Much of its advanced chip production relies on Taiwan Semiconductor Manufacturing Company, which produces more than 90 percent of the world's leading-edge semiconductors. Memory chips from South Korea's Samsung and SK Hynix form another essential component of the ecosystem. 

According to the Semiconductor Industry Association's 2025 State of the U.S. Semiconductor Industry report, the semiconductor industry recorded $630.5 billion in global sales in 2024, a record high. The strategic importance of chips has transformed them into a geopolitical issue. American export controls have attempted to limit China's access to advanced AI processors, while Beijing has invested heavily in domestic alternatives. Yet no single country controls the entire semiconductor supply chain. Chip design, manufacturing, equipment production and packaging are distributed across multiple regions, creating deep interdependencies among the United States, Taiwan, South Korea, Japan, China and Southeast Asia.

Cloud infrastructure scales AI

Individual chips must be combined into vast computing clusters capable of training frontier models. This role falls to cloud infrastructure.

McKinsey's September 2025 report estimated that AI-related data center investment could reach $7 trillion by 2030, with approximately 60 percent devoted to hardware and the remainder split between mechanical and electrical systems. These facilities contain hundreds of thousands of processors connected through high-speed networks and supported by sophisticated cooling systems.

Amazon Web Services, Microsoft Azure and Google Cloud dominate this layer. Together, they operate hundreds of data centers across North America, Europe, Asia and the Middle East, providing computing resources that many companies cannot afford to build independently. Oracle is also expanding aggressively, while Gulf countries have begun investing heavily in regional data center projects. Because few firms possess the capital required to build and maintain large-scale computing infrastructure, cloud providers have assumed a central role in the AI ecosystem, connecting the hardware that powers artificial intelligence with the applications that bring it to users.

Foundation models supply the intelligence

Above the physical infrastructure sit the foundation models that have come to define the current AI boom.

OpenAI's GPT models, Google's Gemini, Anthropic's Claude, Meta's Llama and China's DeepSeek are among the most prominent examples. According to Stanford University's April 2026 AI Index Report , a small number of firms account for a disproportionate share of frontier AI development. In 2025, OpenAI released 20 notable models, followed by Google with 14 and Alibaba with 11, underscoring how concentrated this layer has become.

The model ecosystem is also divided between open and proprietary approaches. Open-source models, such as Meta's Llama and China's DeepSeek, make their model parameters available to researchers and developers, allowing others to study, customize and deploy them. Proprietary systems developed by companies such as OpenAI and Anthropic, by contrast, remain under the control of their creators and are generally accessed through products like ChatGPT or licensed for use in third-party software.

Applications bring AI to everyday life

At the top of the stack are the products that consumers and businesses interact with directly.

ChatGPT, Microsoft Copilot and Perplexity are among the most recognizable examples, but thousands of specialized applications are also emerging. AI tools are being integrated into customer service, software engineering, healthcare, finance, logistics and scientific research. Rather than developing foundation models from scratch, many startups focus on adapting existing systems to particular industries, languages or use cases. 

The spread of these applications is also transforming labor markets. According to the Stanford report, demand for AI-related talent continued to rise in 2025. Singapore recorded the highest concentration of AI-related job postings, with 4.69 percent of listings requiring AI skills, followed by Hong Kong, Luxembourg and Spain. By comparison, AI-related roles accounted for 2.6 percent of job postings in the United States and 1.9 percent in the United Kingdom.

The application layer is where consumers and businesses encounter artificial intelligence most directly. Yet every AI-generated answer, image or recommendation ultimately depends on the electricity, semiconductors, computing infrastructure and foundation models beneath it, illustrating how closely interconnected the components of the AI stack have become.