Global DRAM manufacturing capacity will rise 41 percent by 2030, according to SEMI, the industry association that tracks fab construction and equipment spending worldwide. The projection covers capacity across all major producing regions through the end of the decade.

The figure anchors SEMI's outlook for memory production infrastructure. A 41 percent expansion implies that manufacturers will add wafer capacity at a sustained pace over roughly the second half of the decade, rather than in a single construction cycle.

Who is behind the forecast?

SEMI serves as the semiconductor industry's principal association for equipment and materials supply chain data. Its capacity forecasts draw on announced fab projects, equipment shipment tracking and member reporting. The 41 percent figure represents a forward-looking projection, not a confirmed build-out. Individual company plans can shift with memory pricing cycles.

DRAM capacity additions have historically moved in waves. Manufacturers expand aggressively when prices are strong, then slow construction when oversupply erodes margins. A multi-year projection of this scale suggests SEMI expects the current investment phase, driven by demand for memory used in AI accelerators, servers and data centers, to persist rather than reverse.

Why capacity growth matters now

The DRAM market is concentrated among a small number of producers. Samsung Electronics, SK hynix and Micron control the overwhelming share of global output. Any industry-wide capacity figure of this magnitude therefore reflects capital spending decisions at a handful of companies.

Capacity growth and price direction do not move together automatically. The memory industry has repeatedly added capacity into soft demand, producing price declines that punished producers' margins. Conversely, restrained capacity growth during demand surges, most recently for high-bandwidth memory, has supported firm pricing for suppliers able to qualify advanced products.

SEMI's projection does not, on its own, distinguish between capacity for conventional DRAM and capacity for advanced products such as HBM, which carries different pricing and qualification dynamics.

What remains uncertain?

Forecasts of this length carry clear limits. Fab construction timelines run three to five years from announcement to volume output. Projects announced today may not yield wafers until late in the forecast window, and projects under way can be re-timed.

Demand assumptions also drive the outlook. AI server build-outs have absorbed a rising share of memory supply, and producers have allocated more wafer capacity to HBM at the expense of conventional DRAM. Any change in AI infrastructure spending would ripple through these plans.

The projection also arrives as memory pricing has strengthened. Producers have reallocated capacity and moderated supply growth, and buyers have reported tighter availability for certain products.

What comes next?

SEMI's figure sets a benchmark against which actual build-out can be measured. If the 41 percent expansion holds, the added supply will test whether AI-driven demand growth can absorb it without repeating the oversupply cycles that have defined the DRAM industry's past.