AI boom expected to keep DRAM market under pressure through 2027

The rapid expansion of artificial intelligence infrastructure is putting continued pressure on the global DRAM market, with industry analysts warning that memory shortages could persist well into 2027 and potentially beyond.

The latest assessment from EE Times highlights how the current memory cycle differs from previous DRAM shortages. While earlier periods of strong demand were largely driven by PCs and smartphones, AI data centres have created a new and rapidly expanding source of demand for memory.

Mike Howard, director of DRAM and memory markets at TechInsights, described the current market as the most undersupplied DRAM market in decades. He expects supply constraints to continue through 2027 because demand is growing faster than manufacturers can add production capacity.

AI changes the memory market

AI systems require large quantities of memory, particularly high-bandwidth memory (HBM), which is used alongside GPUs and other accelerators in data centres.

The rapid expansion of AI infrastructure means semiconductor manufacturers are increasingly allocating production capacity towards higher-value memory products. This is contributing to tighter availability of conventional DRAM products used in servers, PCs, smartphones and other consumer electronics.

Brad Gastwirth, global head of research and market intelligence at Circular Technology, told EE Times that much of the capacity expected to come online next year is already effectively committed, making a rapid return to oversupply unlikely.

The situation is also different from previous memory cycles because manufacturers have become more cautious about expanding capacity. Following earlier periods in which aggressive investment resulted in oversupply and collapsing prices, memory producers are taking a more measured approach.

Long-term supply agreements and customer-funded expansion projects are also helping manufacturers expand capacity while reducing the risk of another major inventory build-up.

Consumer electronics face increasing pressure

The consequences are extending beyond the data centre sector.

AI customers are able to pay premium prices for HBM and advanced DRAM, putting consumer electronics manufacturers further down the supply priority list.

That could translate into higher prices for PCs, smartphones and other products, while manufacturers may also have to reduce the number of devices they produce.

Gastwirth estimates that memory can represent around 30% of the bill of materials for some products, meaning substantial increases in memory costs can have a significant effect on manufacturers’ margins and retail prices.

The impact is already being seen in the wider electronics industry. Reuters recently reported that smaller smartphone and laptop manufacturers are preparing for a memory shortage that could continue through at least 2027, with companies having to redesign products and secure memory supplies further in advance.

AI could account for a huge share of DRAM revenue

One of the more significant forecasts highlighted by EE Times comes from TechInsights. Howard estimates that AI-related applications could eventually account for as much as 75% of total DRAM revenue.

That would represent a major transformation for the memory industry, effectively making AI infrastructure one of the dominant markets for DRAM.

Research from Dell’Oro Group also points to the scale of the current expansion. Worldwide data-centre IT semiconductor and component revenue for servers and storage increased 182% year-on-year in the second quarter of 2026, with DRAM and storage revenue growing at triple-digit rates.

Omdia has similarly increased its forecast for 2026 semiconductor revenue, citing unprecedented demand for DRAM from AI infrastructure. The company expects memory ICs to account for more than half of total semiconductor revenue during 2026.

New demand could extend the shortage

The pressure on DRAM may not end when the current generation of AI data centres is built.

Gastwirth points to robotics and physical AI as potential future sources of demand. Robots require multiple computing systems and large numbers of sensors, creating additional requirements for memory and storage.

This could create another wave of demand just as semiconductor manufacturers are attempting to increase supply for existing AI workloads.

At the same time, analysts acknowledge that the AI market remains subject to uncertainty. A significant slowdown in AI infrastructure spending could change the balance between memory supply and demand.

For now, however, analysts quoted by EE Times expect demand to remain strong enough to keep the DRAM market tight.

A changing semiconductor market

The current shortage demonstrates how AI is changing the semiconductor industry. Memory was traditionally regarded as a supporting component within computing systems, but increasingly it is becoming a limiting factor in the ability to deploy AI hardware.

HBM is particularly important because modern AI accelerators require extremely high memory bandwidth. Manufacturers therefore have strong incentives to allocate production capacity towards these products rather than conventional memory.

EE Times reports that bottlenecks are now appearing across HBM production, advanced packaging and semiconductor manufacturing capacity, with some of those constraints expected to remain until at least 2027.

For consumers, the result could be higher memory prices and continued pressure on the cost of electronic devices. For semiconductor manufacturers, the current environment represents a significant shift in the economics of the DRAM market.

Rather than the traditional boom-and-bust cycle, the industry is attempting to expand capacity cautiously while AI companies continue to consume increasing quantities of memory.

Unless AI infrastructure spending slows substantially or new manufacturing capacity arrives faster than expected, the DRAM market therefore appears likely to remain under considerable pressure throughout 2027.