Memory has become the “king” of the semiconductor industry, according to one analyst, and he believes “its reign is here to stay.”

Investments into artificial-intelligence data centers have put chip-industry revenue on pace to double to $1.5 trillion this year, and memory chips have been ”the primary driver” of that jump, Susquehanna analyst Mehdi Hosseini said, citing data from the Semiconductor Industry Association on global shipments and average sales prices in July.

Memory chips now make up between 50% and 55% of overall semiconductor revenue, he said in a Tuesday note to clients, adding that memory has historically made up about 20% to 30% of total industry revenue.

Demand has surged for memory chips in the AI boom, and that has led to severe supply shortages and rising prices. Although average prices across the semiconductor industry are already trending upward, Hosseini said the growth would be “more moderate” without counting the memory segment.

The acceleration in demand and pricing for memory chips has made Micron Technology one of the best-performing stocks in the S&P 500 so far this year. The stock is up 250% on the year. Sandisk, which focuses on NAND flash memory, has also been a major beneficiary of AI-driven demand for storage. That has sent its shares up 632% so far this year. Hosseini has positive ratings on both stocks.

Hosseini said he expects average prices across dynamic random-access memory to increase 50% sequentially in the current quarter, and to rise a further 20% in the fourth quarter of 2026. He sees average prices for NAND climbing 60% sequentially this quarter, and another 25% in the following quarter.

Earlier this month, a report from South Korean news site Electronic Times said Micron is aiming to double its production capacity of high-bandwidth memory by the end of the year. The advanced memory technology, which involves stacking layers of DRAM, is in high demand from chip makers such as Nvidia to handle intense AI workloads. The silicon-intensive process of making HBM has also contributed to shortages of DRAM.

Meanwhile, Hosseini noted that logic chips, such as central processing units, have also benefitted from rising average prices, though “recent trends suggest a period of stabilization.” He attributed that to delays in the ramp of Nvidia’s upcoming Rubin AI platform, which includes both its Vera CPUs and Rubin graphics processing units.

Hosseini said that while future generations of AI hardware will carry “a higher upfront cost,” the falling prices of tokens, or the pieces of data processed by AI models, coinciding with an increase in production of tokens helps to support “the economic value proposition of AI at scale.”