Micron Technology is aggressively scaling its high-bandwidth memory (HBM) production capacity, with a clear goal of capturing a significantly larger share of the global AI memory market.

According to an industry source cited by Yonhap News Agency on the 4th, Micron plans to boost its monthly HBM output to roughly 100,000 wafers by the end of this year, nearly doubling last year's figure. This move is expected to reduce its capacity gap with Samsung Electronics and SK Hynix to about half of its current level. Simultaneously, the company is accelerating the production ramp-up of its latest-generation HBM4 12-layer products to meet the demands of NVIDIA's upcoming AI accelerators.

This development underscores the persistently robust demand for AI memory. AI chipmakers, led by NVIDIA, continue to exert intense demand for HBM, with supply still struggling to keep pace. For Micron, which holds an 18% market share and has long been the third-largest player, this expansion drive is both a critical battle for market position and a necessary step to convert AI memory demand into tangible revenue.

Production target: Monthly capacity to hit 100,000 wafers by year-end

An industry insider told media that Micron plans to add up to 60,000 wafers of monthly HBM capacity by the end of this year. As last year's monthly output was around 40,000 to 50,000 wafers, the new addition will exceed that base, bringing total capacity to approximately 100,000 wafers per month.

To achieve this goal, Micron has increased purchase orders with multiple HBM manufacturing equipment suppliers, with related machinery being continuously delivered to its facilities in Taiwan and Singapore. Currently, its HBM packaging operations are primarily carried out at the Taichung plant in Taiwan and the Woodlands plant in Singapore, with equipment investment also being prepared for the Hiroshima plant in Japan.

In terms of product mix, Micron is rapidly shifting toward HBM4 12-layer products. Its current HBM output is still dominated by HBM3E 12-layer, but the share of HBM4 12-layer is expected to rise from the 20% to 30% range seen at the start of this year to as much as 50% by year-end.

HBM4 12-layer is the companion memory for NVIDIA's latest AI accelerator, "Vera Rubin," and Micron initiated mass production of this product in the second quarter of this year. During the company's fiscal 2026 third-quarter earnings call in June, CEO Sanjay Mehrotra stated that the production ramp-up speed for HBM4 12-layer is approximately twice that of HBM3E 12-layer, and cumulative revenue from HBM4 shipments has already surpassed $1 billion.

Market landscape: The gap among the top three could narrow significantly

Despite being one of the world's three major HBM manufacturers, Micron still trails significantly behind Samsung Electronics and SK Hynix in terms of capacity. Industry consensus holds that the monthly HBM output of Samsung and SK Hynix is approximately 150,000 to 200,000 wafers each, representing three to four times that of Micron. Without closing this gap, Micron will struggle to alter its long-standing third-place position.

According to data from market research firm Counterpoint Research, SK Hynix led the global HBM market with a 50% share in the second quarter of this year, followed by Samsung Electronics at 32% and Micron at 18%. If Micron's capacity expansion targets are met as planned, the production gap between it and its two main competitors could be compressed to half of current levels, potentially triggering a reallocation of market share.