TradingKey - According to industry insiders cited by Yonhap News Agency on September 4, Micron (MU) plans to increase its monthly high-bandwidth memory (HBM) production capacity to about 100,000 wafers by the end of 2026, nearly doubling last year's scale. If the target is achieved as scheduled, the capacity gap between Micron and Samsung Electronics and SK Hynix could narrow to around half of its current level.
Meanwhile, Micron is accelerating the volume production ramp of its latest-generation 12-layer HBM4 products to meet demand for Nvidia's next-generation AI accelerators.
Micron’s Monthly HBM Capacity May Increase to 100,000 Wafers
Industry sources revealed that Micron plans to add up to 60,000 wafers of monthly HBM capacity. The company's monthly HBM output last year was about 40,000 to 50,000 wafers, and after the new capacity goes into production, the total scale is expected to reach approximately 100,000 wafers.
To achieve this goal, Micron is placing additional orders with several HBM manufacturing equipment suppliers, with relevant equipment being shipped to its factories in Taiwan and Singapore. Currently, Micron's HBM packaging business is mainly being carried out at its Tongluo plant in Taiwan and Woodlands plant in Singapore, while equipment investment for its Hiroshima plant in Japan is also under preparation.
Behind the capacity expansion is sustained demand for HBM from AI servers and AI accelerators. Because HBM provides higher memory bandwidth, it has become a crucial component in generative AI training and inference systems. AI chipmakers, led by Nvidia, continue to scale up their procurement, and the HBM market remains in short supply.
Share of HBM4 12-Layer Products Set to Rise
In addition to expanding capacity, Micron is also adjusting its HBM product mix. Currently, the company's HBM output remains dominated by 12-layer HBM3E products, but the share of 12-layer HBM4 products is expected to rise from 20% to 30% at the beginning of 2026 to up to 50% by the end of the year.
12-layer HBM4 products are regarded as the complementary memory for Nvidia's next-generation "Vera Rubin" AI accelerator. Micron launched volume production of the product in the second quarter of 2026.
Micron Chief Executive Sanjay Mehrotra stated during the third-quarter fiscal 2026 earnings call in June that the mass production ramp-up rate for 12-layer HBM4 was roughly twice that of 12-layer HBM3E. By then, Micron's cumulative shipment revenue for HBM4 had surpassed $1 billion.
Micron Still Needs to Catch Up with Samsung and SK Hynix
Although Micron Technology is one of the world's top three HBM manufacturers, its production capacity remains significantly below those of Samsung Electronics and SK Hynix. Industry estimates generally indicate that Samsung Electronics and SK Hynix each have a monthly HBM capacity of approximately 150,000 to 200,000 wafers, three to four times Micron's current scale.
In terms of market share, Counterpoint Research data shows that in the global HBM market in the second quarter of 2026, SK Hynix ranked first with a 50% share, while Samsung Electronics accounted for 32% and Micron Technology held 18%, remaining in third place over the long term.
If Micron Technology can raise its monthly HBM capacity to approximately 100,000 wafers by the end of the year and successfully increase the proportion of 12-layer HBM4 products, the capacity gap between Micron and its two major rivals is expected to narrow significantly. For Micron, this is not only a key action to compete for HBM market share, but also an important step in converting AI memory demand into revenue growth.
Find out more