Miért összpontosítunk a mesterséges intelligenciára?

Artificial intelligence enables computers to perform tasks that traditionally require human intelligence. Businesses use AI to analyse information, predict demand, automate routine tasks and develop new products.

These applications rely on different areas of AI, including machine learning, natural language processing (NLP), computer vision, generative AI, AI robotics and autonomous & sensor technology.

As companies introduce AI into their operations, they purchase software and computing services from technology providers. The share of surveyed organisations using AI increased from 55% in 2023 to 88% in 2025.^{Source 1}

As more businesses develop and use AI systems, demand for computing power is increasing. AI infrastructure consists of the physical equipment needed to develop and run artificial intelligence. This primarily includes servers and specialised processors, memory, data storage and networking equipment.

Companies such as Microsoft, Amazon and Google are expanding their computing capacity to operate AI services. This requires powerful servers equipped with processors, including NVIDIA's GPUs, memory supplied by companies such as Micron, and equipment to store and transfer data. Many of these chips are manufactured by TSMC.

Worldwide AI infrastructure spending more than doubled from $153 billion in 2024 to $318 billion in 2025, an increase of 108%.^{Source 2} This growth was largely driven by investment from major cloud providers, increasing demand for AI servers and national AI infrastructure projects.

Building a data centre is expensive and takes years. Most businesses therefore rent computing power from cloud providers and pay for what they use, in a similar way to paying for electricity. This is known as Infrastructure as a Service, or IaaS. The broader IaaS market is forecast to reach $367 billion in 2027.^{Source 5}

The three largest providers are Amazon Web Services (AWS), Microsoft Azure and Google Cloud. As AI workloads grow, they are adding servers designed specifically for training and running AI models. Spending on this AI-optimised cloud capacity is expected to rise by 96% in 2026.^{Source 3}

Computing infrastructure provides the power needed to run AI, but businesses need software to put that power to use. This software relies on AI models, which are systems trained on data to understand information, identify patterns and generate responses.

A foundation model is an AI model that can perform a wide range of tasks. Examples include GPT, developed by OpenAI; Gemini, developed by Google; Claude, developed by Anthropic; and Llama, developed by Meta. Businesses can use these models to build their own applications or purchase ready-made AI software, including ChatGPT and Microsoft Copilot.

Companies earn revenue by charging customers for access to their models and applications. This can take the form of monthly subscriptions, software licences or fees based on usage. Worldwide enterprise spending on foundation models reached approximately $12.5 billion in 2025.^{Source 6}

As AI becomes useful for more business activities, the market for these products is expected to expand. The worldwide generative AI market was estimated at $288.4 billion in 2025 and is forecast to reach approximately $1.34 trillion by 2032.^{Source 7}

Most AI tools today answer one request at a time. Agentic AI goes a step further: given a goal, it plans and carries out a sequence of tasks, such as gathering information, filling in forms and checking the result, with limited human input. The more complex the processes, the more the demand for AI software and computing capacity will increase.

A further development is Physical AI, which combines artificial intelligence with machines capable of sensing their surroundings, making decisions and performing physical tasks. Examples include warehouse robots, autonomous vehicles and industrial robotic arms. Robotics is one emerging market within this area. Potential applications include manufacturing, warehousing, logistics and autonomous vehicles.

These developments reinforce demand for technologies already used in AI infrastructure. NVIDIA develops high-performance processors and robotics platforms, Micron supplies memory, and companies including Microsoft, Amazon and Alphabet provide cloud computing services.

The worldwide AI robotics market was estimated at $74.49 billion in 2025 and is forecast to reach $180.63 billion by 2032, a projected growth of +142.5%.^{Source 8} While the commercial adoption of autonomous machines is still developing, robotics opens up another potential market for companies supplying AI hardware, computing infrastructure and software.

The growth of artificial intelligence is creating demand across several connected markets. Businesses purchase AI software and services, cloud providers supply computing capacity, and semiconductor manufacturers produce the processors and memory needed to operate these systems.

Worldwide spending on artificial intelligence was estimated at $1.79 trillion in 2025 and is forecast to increase to $2.67 trillion in 2026 and $3.64 trillion by 2027. More than half of the projected spending in 2027 relates to AI infrastructure, reflecting the substantial investment required to develop and operate these systems.^{Source 4}

As AI technology develops, new applications are expected to create further demand for computing infrastructure and software. Autonomous AI agents, which can plan and perform sequences of tasks, and Physical AI, which enables machines to interact with their surroundings, are two areas of particular interest.

This development creates opportunities for companies operating across the AI industry, including semiconductor and memory manufacturers, cloud computing providers and developers of AI applications.