Disclaimer: Unless otherwise stated, any opinions expressed below are solely the author’s.
Just last month, Apple, the last figure in the personal computer market, was forced to raise the prices of its computers and tablets between 10 and 30%, depending on the brand and model.
The Cupertino giant has been able to wait out AI-induced inflation thanks to its long-term contracts for memory chips and TSMC’s manufacturing capacity that was pre-booked for Apple silicon processors.
The fact that it has long enjoyed some of the highest positions in the industry must have helped, providing a buffer that allows it to absorb some of the costs from other areas.
Windows PC buyers had a rough time last year, as the AI revolution hit their devices first. Laptops are still available, but building a new desktop PC is currently almost impossible for average consumers after RAM prices explode by several hundred percent by late 2025.
This appeared over the prices of graphics cards, AI comes first, as hyperscalers like OpenAI, Anthropic or Google needed to secure their millions to train their artificial intelligence models.
That said, during the operation to beat Nvidia and AMD cards, RAM sticks and SSD storage, one part was not affected: the central processing unit (CPU).
Unfortunately, that is about to change.
AI doesn’t just work on GPUs
GPUs are best at doing many parallel calculations at once. This makes them very important for training artificial intelligence models, which is why they were important in the early years of the AI boom.
To build your AI model, you need GPUs—and LOTS of them.
But they don’t work independently.
CPUs still have to prepare and feed data to the accelerators, manage memory, handle networks, run jobs and coordinate all the other processes that happen around the model.
This is especially important during reasoning—the stage where a trained AI model responds to users.
Training can happen once or occasionally, but reflection happens every time someone asks ChatGPT a question, generates an image, writes code with Claude or tells the AI agent to complete a task.
As the number of AI users and applications grows, so does the need for logic.
In addition, the emerging generation of AI agents is very hungry for general-purpose computing.
Unlike a chatbot that generates a single response and stops, an agent may browse files, call external tools, execute code, check the results and repeat the process multiple times. Each of these actions creates work that CPUs are better suited to handle.
Until recently, AI companies needed about one CPU to eight GPUs, but that ratio is expected to decrease rapidly and could approach a 1:1 ratio by 2029.
Think of the millions of GPUs sold in the last two to three years. Now, their deployment may require three, four, maybe eight times as many CPUs. And there are new data centers being built as we speak.
Therefore, it is not only a new method that must fill the existing gaps, but also respond to future needs.
Intel is already running short
This wouldn’t be a problem if chip makers had a large amount of spare capacity. But they don’t.
Intel acknowledged that demand was already outstripping supply by late 2025 and warned that shortages would continue into 2026. Its data center business could not fully meet customer needs due to limited wafer capacity in its factories.
The company is now prioritizing the production of server chips, including its highly profitable Xeon processors, as the demand for AI grows.
This makes commercial sense. A single high-end server processor can cost thousands of dollars, while a CPU in a typical laptop might cost a manufacturer half that.
But Intel can’t just create more factory capacity overnight. If it produces more Xeons using captive production lines, something else may have to give.
And that something can be consumer processors.
Industry reports suggest that server CPU prices have already increased by around 20% since Mar, while consumer models are reported to be 5 to 10% more expensive in some channels.
More increases may follow later this year.
AMD is also benefiting. Data center revenue rose 57% year over year to US$5.8 billion in the first quarter of 2026, driven in part by strong demand for its EPYC processors.
Unlike Intel, however, AMD does not have its own flagship factories. It relies heavily on TSMC, which produces chips for Nvidia, Apple and many other competing companies with limited development capacity.
So, while Intel has to choose what to produce in its factories, AMD has to compete for space with someone else, the same Taiwanese company that everyone already relies on.
The bottle is getting tighter.
Regular consumers will end up paying too
Although servers and consumer CPUs are not always made with the same processes, and chip companies cannot freely change the entire production line from one product to another, there are several ways that pressure can reach consumers.
Manufacturers may prioritize their limited capacity, engineering resources and parts for more profitable business products. Computer makers may pay more for chips under their supply agreements. Lack of resources and installation parts can also increase the cost of the entire system.
And many high-end consumer processors can be used directly in business settings. While they may not perform the most important and critical tasks, they can serve in support roles to help maintain valuable server models where they are needed most.
That, in turn, could lift them from the consumer market and drag down the prices of all processors, as consumers turn to the next best option.
There is always another barrier
The AI boom started with the idea that the industry simply needed more GPUs. It soon became clear that it needed memory, storage and things outside of technical components, such as electricity, building materials or professional construction work.
Now, CPUs join the list.
This is what happens when hundreds of billions of dollars are invested in one industry at one time. Solving just one deficiency reveals the next bottleneck after that.
For consumers, the frustrating part is that they are competing with some of the richest companies in history, many backed by tacit or direct government support, as nations all see the use of AI as a strategic interest.
Hardware manufacturers will naturally sell their limited capacity where they make the highest return. Currently, it is becoming more and more common inside data centers for AI rather than the computers sitting on our desks.
That’s why, unfortunately, if you’ve been waiting for GPUs and RAM to get cheaper before buying your next PC, you may soon have something else to worry about. And there is no end in sight.
- Read other articles we’ve written about the rise of artificial intelligence here.
Featured Image Credit: Shutterstock
