An analysis of the semiconductor market in 2025/2026 indicates an important shift in the architecture of artificial intelligence systems. While the first phase of the AI revolution was mainly based on graphics processing units (GPUs), Intel’s financial data and the development of so-called Agentic AI are bringing general purpose processors (CPUs) back into the spotlight. With AI revenues up 40% year-on-year, the industry is faced with the question of cost optimisation. Why CPUs are becoming a key component of modern IT infrastructure.
AI drives server sales
For the past few years, Intel has had to measure itself against the reputation of being the company that ‘slept on’ the AI revolution. However, Q1 2026 results show a very different picture. The company’s revenue grew by 7% year-on-year and, more importantly, as much as 60% of its server segment revenue now comes from AI projects.
Demand for the latest Xeon processors is outstripping the available supply. This is a signal to the market that big tech companies (Big Tech) need classic computing power again. This shift is due to the fact that AI has ceased to be just a curiosity in the lab and has become part of everyday business operations.
From training to action: The role of the CPU in the age of Agentic AI
To understand why the CPU is coming back into favour, we need to distinguish between two stages of AI work: the training of models and their use.
- Training: This is where graphics processing units (GPUs) reign supreme. They can perform thousands of simple calculations simultaneously, which is essential to ‘feed’ the model with data.
- Operation and logic: this is where the CPU comes into play. Modern artificial intelligence is increasingly taking the form of ‘agents’. These are systems that not only answer questions, but can plan tasks, write code and make decisions themselves.
Agentic AI requires complex logic and fast switching between tasks. These are tasks where the Intel Xeon processor architecture performs much better than specialised GPUs. The processor becomes the ‘conductor’ that manages the flow of information and decides how to execute a complex user command.
Economics matter: Tenfold price difference
From a business perspective, the most important argument is cost. There is currently a huge shortage of GPUs on the market, which has pushed GPU prices to extreme levels. Companies have quickly noticed that using the most expensive graphics cards for simple logic tasks is simply not cost-effective.
Many IT leaders have made the mistake of trying to solve every problem with the GPU. Meanwhile, many AI processes leave most of the GPU cores idle, generating huge waste. Given that a server processor can be up to ten times cheaper than its graphics counterpart, the economic sense of reverting to the CPU is obvious. For CFOs and IT managers, this means being able to perform the same AI tasks at a much lower cost.
Market confirms trend: Not only Intel is investing in CPUs
The fact that general-purpose processors are important again is evidenced by the moves of the biggest players in the market. Intel is not the only one betting on the revival of this architecture. Nvidia, the previous GPU market leader, is developing its own Vera processor. Arm is introducing chips dedicated to AI, and giants such as Meta, Google and Amazon are investing billions in their own processor designs.
The technology industry seems to be correcting the mistake of a few years ago, when the foundations were forgotten in the pursuit of GPU power. Today, no one treats CPUs as relics anymore. They have become the indispensable foundation without which AI infrastructure cannot operate efficiently and cheaply.
What does this mean for business?
For companies investing in technology, there are three specific lessons from this situation:
- Cost-effectiveness: It is worth checking which in-house AI tasks can be performed by CPUs. This will allow you to make big savings without sacrificing quality.
- Resource planning: The shortage of Intel Xeon processors shows that server hardware purchases need to be planned well in advance.
- Investing in logic: the future belongs to AI agents, and these need strong processors to manage tasks efficiently.
By basing its strategy on the rebirth of the CPU, Intel has hit the point of the needs of today’s market. While GPUs will continue to be essential for training the largest models, CPUs such as Xeon will be responsible for their practical business use.
