Samsung plans to double its HBM4 chip output next year

Samsung is planning to more than double its production capacity of HBM4 and HBM4E memory next year to feed hungry AI chipmakers.

According to industry sources reported by SE Daily, Samsung is scaling up output for its sixth-generation HBM4 and seventh-generation HBM4E memory chips. The telltale sign comes from the supply chain: Samsung plans to increase outsourced cleaning for glass carriers—supports that hold silicon wafers flat during grinding—from 20,000 sheets a month to 50,000 next year. Overall HBM wafer input is projected to rise 40% to roughly 250,000 wafers a month, with HBM4 making up about 80% of total shipments.

Why it matters: Stacking memory 12 layers high is physically difficult. Wafers bend easily when ground down ultra-thin. Doubling down on HBM4 shows Samsung is confident it can manage those manufacturing hurdles to supply customers like Nvidia, who received 12-layer HBM4E samples in May.

Here's the gist on glass carriers: think of them like temporary scaffolding. A glass support attaches to the DRAM wafer while it gets drilled and ground thin, preventing cracks before being washed and reused.

Next year, the AI race runs on very thin, very crowded silicon.