Micron's ¥1.5 Trillion Hiroshima HBM Expansion: What You Need to Know

Target Products and Customers

The Hiroshima expansion will focus on next-generation HBM, specifically HBM4 and beyond — the standard expected to succeed today's HBM3E . These chips are critical components for AI training and inference accelerators. According to Bloomberg and other reports, the primary target customer is Nvidia, whose H100, B200-class, and future GPUs depend on HBM for memory bandwidth . The plant positions Micron to supply Nvidia and other AI chip customers as they transition to next-generation memory architectures.

Production Timeline

• Construction start: May 2026, with site preparation already underway .
• First shipments targeted: Summer 2028, specifically between June and August 2028 .
• Full-scale production at maximum capacity: Projected for March to May 2030 .

This timeline means Micron's Hiroshima HBM output will arrive in the latter half of the decade, potentially capturing demand from the next wave of AI data center buildouts .

Japan's Subsidy Commitments

Japan's Ministry of Economy, Trade and Industry (METI) has committed substantial financial support:

• ¥500 billion (~$3.3 billion) directly allocated for this HBM expansion, covering roughly one-third of the capital cost .
• Combined with previous METI grants for Micron's Hiroshima operations, total Japanese government support now reaches ¥774.5 billion .
• Earlier reports also referenced a ¥536 billion ($3.6 billion) subsidy package announced in September 2025 .

This subsidy aligns with Japan's broader national strategy to revive its domestic semiconductor industry and secure advanced memory manufacturing on its soil .

How This Challenges SK Hynix's HBM Dominance

Current market context: SK Hynix has long dominated the HBM market as Nvidia's primary HBM3 and HBM3E supplier, enjoying a commanding lead in production volume and technology certification . Samsung trails in second place, while Micron has been a distant third for current-generation HBM.

Micron's competitive strategy with this plant:

• Dedicated HBM fab: Unlike SK Hynix's shared DRAM/HBM lines, this is a purpose-built HBM facility, which could yield higher efficiency and better yield rates .
• HBM4 leapfrog: By targeting HBM4 production from 2028, Micron aims to compete on the next technology node rather than try to catch up on current-generation HBM3E .
• Japanese government backing reduces capital risk and gives Micron a cost advantage compared to SK Hynix's Korean operations .
• Geopolitical diversification: The Hiroshima fab adds a Japan-based HBM supply chain, reducing reliance on Taiwan and South Korea — a factor increasingly important to U.S. and Japanese AI companies .

The challenge faces headwinds: SK Hynix is already developing HBM4 on an aggressive timeline, with mass production targeted as early as 2026-2027 . Micron's 2028 first-shipment date means SK Hynix could have a 1-2 year head start on the next-generation HBM cycle . However, within the broader AI memory supercycle — where HBM demand is projected to grow 40-50% annually through 2030 — there may be room for multiple winners, and Micron's late entry with a highly subsidized, dedicated fab could capture meaningful second-source share from Nvidia and other AI chip customers .