AI growth drives chip demand, but memory shortages and supply chain woes threaten progress.

Generative AI has quickly become the new buzz word (and for good reason) in the expansive world of technology, as tech giants and startups alike aim to claim their slice of the proverbial AI pie. But how much value can companies truly garner from this AI boom if the systems used to power them lack the necessary components and ability to perform?

It is true that, in response to the ever-growing demand for AI workloads, semiconductor sales have thus increased. Global semiconductor sales reached a figure of $364 billion in in the first half of 2025, a 18.9% increase compared to the same time period last year, according to the World Semiconductor Trade Statistic’s (WSTS) latest report on global semiconductor sales.

Following an unexpectedly robust third quarter, the outlook for chip sales remains optimistic. According to WSTS data, the global semiconductor market could grow by more than 25%, reaching an estimated $975 billion. Analysts credit this increase in sales to the boosted demand for data center infrastructure following the emergence of initial AI edge applications.

However, despite growth trends, it is important to remain cautiously optimistic, as the semiconductor industry can be a fickle mistress.

A report from Deloitte published in February of this year paints a more reserved outlook for semiconductor sales: “while gen AI chips and associated revenues (memory, advanced packaging, communications, and more) are responsible for outsized revenues and profits, they represent a small number of very high-value chips, meaning that wafer capacity—and therefore utilization—for the industry as a whole isn’t as high as it might appear.”

“It’s worth reminding readers that the chip industry can be notoriously cyclical. The industry has flipped from growth to shrinkage nine times in the last 34 years,” the report added. “So, it may seem that the industry is seeing less extreme growth or shrinkage in the last 14 years, compared to 1990 to 2010, but the frequency of contractions seems to have increased. The year 2025 looks solid for now, it’s hard to tell what 2026 will bring.”

(Source: Deloitte)

So, will the semiconductor industry remain up to snuff when it comes time to deliver on the advances of AI, especially given the expected growth in sales? Truth is that the industry already seems to be struggling.

Statistics provided by McKinsey & Company predicts that the total generative AI compute demand could reach 25×1030 FLOPs by 2030, which exceeds the performance of current supercomputers and AI systems outfitted with today’s latest silicon, as previously reported in an article contributed to EE Times earlier this year.

“The adoption of gen AI use cases in the B2B sector is significantly influenced by the sufficiency and cost of semiconductor chip supply,” according to the Mckinsey Report. “Enterprises must be capable of rationalizing their investment in compute infrastructure, ensuring that the cost of service is lower than the company’s willingness to pay.”

The industry remains at an inflection point: the success of AI depends greatly on the semiconductor market’s ability to provide the chips to power this new age of intelligence. And while the Covid-era chip shortage seems to have abated, new constraints continue to choke critical supply chains and pose a significant hurdle to AI’s continued proliferation: most notably, memory.

As it stands, the industry has hit a roadblock: while demand for advanced AI chips skyrockets, supply of traditional memory chips (DRAM, HBM, flash) is tightening sharply, causing noticeable shortages and price hikes. In fact, a report from market research firm Counterpoint expects memory prices to increase by as much as 20% in 2026 as a result of the expansion of AI, as reported in November by EE Times correspondent Alan Patterson.

The culprit? A lack of legacy LPDDR4 memory after heavyweights like SK Hynix and Micron shifted output to more advanced high-bandwidth memory (HBM) to meet the AI boom, according to Counterpoint. Add in U.S. tariffs and competition from China, and this is surely a recipe for disaster.

However, some players are making moves to meet this shortage in memory.

For instance, Marvell Technology announced on Dec. 2 a massive acquisition that could make waves: it is buying startup Celestial AI, featured in this year’s Silicon 100 report published by EE Times, in a $3.25 billion deal. The move positions Marvell to compete more directly with top AI-chip and networking firms thanks to Celestial AI’s specialization in photonic-fabric data-center technology.

Meanwhile, reports of a $9.6 billion investment from Micron Technology to build a new HBM chip manufacturing facility in Hiroshima, Japan, could help lessen the strain on the global memory market.

Additionally, Samsung has set its sights on a strategy to accelerate mass production of its memory tech, having announced a new semiconductor research and development complex (NRD-K) at its Giheung campus in South Korea. This new R&D initiative will allow Samsung to “set up with High NA extreme ultra-violet lithography and new material deposition equipment aimed at accelerating the development of next-generation memory semiconductors such as 3D DRAM and V-NAND with more than 1,000 layers,” the company said in prepared remarks.

And a recent memorandum of understanding signed between United Microelectronics Corp. (UMC), a manufacturer of IC processing technologies and manufacturing solutions that includes embedded NVM solutions, and Polar Semiconductor, a foundry that specializes in high-voltage, power and sensor semiconductor manufacturing, represents a push to boost U.S. domestic chip production for sectors like automotive, aerospace and defense—critical sectors that could be negatively impacted by a shortage in memory.

This article was originally published on EE Times.