The development of artificial intelligence continues to create a paradoxical situation for the semiconductor industry. Demand is growing so rapidly that companies are losing potential revenue due to a shortage of components and production capacity. Intel CEO Lip-Bu Tan acknowledged that the company can currently meet no more than 50% of the demand for its processors, while also warning that the memory shortage may persist until at least 2028.
One of the reasons is the rapid change in the structure of the AI market. As the industry shifts from model training to inference, demand is growing not only for GPUs but also for central processors, which are used to manage accelerators and applications, as well as to handle general‑purpose computing. For Intel, this creates a rare opportunity to boost sales of server CPUs, but the company is not yet able to fully capitalise on this demand. Some investors are already calling this a new gold rush, drawing parallels with past investment booms and even turning to the gold price chart to identify similar market patterns.
Meanwhile, almost all critically important infrastructure elements are becoming more expensive. According to Tan, memory prices have already increased five to sevenfold, and many companies are forced to postpone projects because they cannot secure the necessary volumes of chips. This is compounded by shortages of cooling systems and electricity for new data centres. Intel is trying to partially address the cooling problem through its own investments in liquid cooling technologies, but most of these constraints are beyond its control.
This shortage simultaneously gives the company an opportunity to raise prices. As early as October 5, Intel is expected to increase the price of PC processors for the third time since the end of last year, by approximately 10%. Previously, prices rose in the first quarter, and in July, some client and server chips became more expensive, with the price increase for the latter in some cases exceeding $1,000.
However, limited supply also affects Intel’s promising area of advanced chip packaging. The company is facing a shortage of substrates and is forced to make advance payments to suppliers to secure priority. In parallel, the yield for some EMIB substrates from Japanese partners is only about 45%, although an improvement to 55% is expected by January. All of this means additional costs precisely at the moment when Intel is trying to regain competitiveness in its manufacturing business.
The next major milestone is the 14A process, the production of which the company expects to begin in the first quarter of 2027. Alongside this, efforts to improve the yield of 18A chips continue. For Intel, successful scaling of these technologies is critically important. High demand does little to help financial performance if the company is physically unable to produce enough products to meet it.
Against this backdrop, Intel continues to free up capital by divesting non‑core assets. Acquired in 2015 for $16.7 billion, Altera is preparing for an IPO after Intel sold 51% of the company to Silver Lake last year for $4.46 billion, retaining 49 %. At that time, the entire Altera business was valued at approximately $8.75 billion, which is almost half the original acquisition price.
The upcoming IPO will allow Altera to independently finance further development, while Intel will reduce its own financial burden and potentially monetise the remaining stake. The timing for the offering looks favourable. The AI boom has dramatically revitalised the US IPO market, and investor demand for companies related to semiconductors and artificial intelligence remains high.
Moreover, the scale of available capital is well illustrated by OpenAI. Despite abandoning an IPO this year, the startup is discussing another private round with a potential valuation of $1.2 trillion, compared to $852 billion in March. Despite annual expenses of approximately $34 billion against revenue of around $40 billion, investors continue to fund the expansion of AI infrastructure.
For Intel, this investment race is both an opportunity and a challenge. The more capital AI developers receive, the higher the demand for processors, packaging, and infrastructure. But the same boom drives up the cost of memory, substrates, cooling, and data centre construction.
As a result, the main constraint for Intel is no longer the lack of buyers, but the ability to convert that demand into actual output. If the company can increase production and successfully scale up 18A and 14A, the current shortage will give it the opportunity to increase volumes and prices at the same time. If not, a significant portion of the demand generated by the AI boom will remain lost revenue.
