Samsung to Double HBM4 Output

Samsung Electronics is expected to more than double its output of HBM4 and HBM4E high-bandwidth memory chips next year, according to semiconductor industry sources reported on September 20. The signal comes from glass carriers, an essential material in HBM production, where Samsung is raising outsourced cleaning volume from 20'000 sheets per month this year to 50'000 sheets per month next year, a 2.5-fold increase.

Glass carriers are glass supports temporarily attached to the underside of HBM DRAM wafers to prevent bending or cracking while the wafer is ground thin. Because HBM requires stacking multiple DRAM dies within a limited thickness, thinning and warpage control become more critical as stack counts rise. The HBM4 and HBM4E products Samsung is scaling center on 12-layer and higher stacks.

From 10'000 to 50'000 in two years

The trajectory is steep. Glass carrier requirements stood at 10'000 sheets per month as recently as last year. That doubled to 20'000 this year and is set to rise 2.5-fold next year. Even accounting for the fact that glass carriers are reused after cleaning and that consumption varies by process loading and yield, a 2.5-fold increase makes it highly likely that HBM4 and HBM4E output will grow at least twofold from this year.

The product mix shift

Samsung began mass-production shipments of HBM4 in February 2026, using 10-nanometer-class sixth-generation (1c) DRAM and a base die built on a 4-nanometer process. In May, it provided 12-layer HBM4E samples to customers including Nvidia.

Industry analysts expect Samsung's total HBM production to grow nearly 40% to about 250'000 wafers next year, up from roughly 180'000 this year, measured by average monthly wafer input. By product shipment mix, the HBM4 family is projected to rise from around 40% this year to about 80% next year as HBM4E mass production ramps up. Samsung appears to be placing HBM4, a high-value product, at the center of its expansion strategy.

Why this matters

HBM4 is the memory architecture that powers the next generation of AI accelerators. Nvidia, AMD, and Google's TPUs all depend on it. The fact that Samsung is betting this hard on a 2.5x increase in production capacity signals confidence that AI compute demand will continue scaling aggressively through 2027 and beyond.

The Hacker News discussion (419 points, 271 comments) touched on the geopolitical implications: Samsung and SK Hynix dominate HBM production, both based in South Korea. TSMC manufactures the logic dies that pair with HBM stacks. The entire AI hardware supply chain runs through a narrow corridor of companies in East Asia, a fact that has not gone unnoticed by policymakers in Washington and Brussels.

For anyone running inference on consumer hardware, HBM4 is not coming to your desktop. It is a data center part, packaged and soldered directly onto accelerator boards. What trickles down is price pressure: as HBM4 capacity expands, older HBM3 stock gets cheaper, which eventually reaches the used and surplus markets. But the real story is the scale. When a company the size of Samsung more than doubles output of the most critical AI memory component in a single year, it tells you something about where the industry thinks the ceiling is.