Micron HBM4 mass production Q1
- Micron said on March 16 it began volume shipments of its 36GB 12-high HBM4 in the first quarter of calendar 2026. (investors.micron.com) - Micron said it has completed price-and-volume agreements for all calendar 2026 HBM supply, including HBM4, and named NVIDIA Vera Rubin as HBM4’s platform. (investors.micron.com) - Micron’s next disclosed step is HBM4 48GB 16-high customer sampling, alongside a fiscal 2026 capex plan above $25 billion. (investors.micron.com)
Micron Technology said on March 16 that it had begun volume shipments of its 36-gigabyte, 12-high HBM4 memory in the first quarter of calendar 2026, giving the company a new product in the race to supply memory for next-generation AI systems. The company said the part was designed for NVIDIA’s Vera Rubin platform and would deliver more than 2.8 terabytes per second of bandwidth with more than 20% better power efficiency than Micron’s HBM3E. (investors.micron.com 1) (investors.micron.com 2) Micron’s own disclosures also back the supply-tight part of the story. In fiscal first-quarter materials published on December 17, 2025, the company said it had completed agreements on price and volume for its entire calendar 2026 HBM supply, including HBM4. (investors.micron.com) In March, Micron raised its fiscal 2026 capital spending outlook again, with outside reports and company presentation material pointing to spending above $25 billion as it expands capacity tied to AI memory demand. The result is a clearer picture than the original market chatter suggested: the HBM4 launch is real, the NVIDIA design win is official, and the “sold out through 2026” claim is best understood through Micron’s statement that all 2026 HBM supply is already under price-and-volume agreements. (investors.micron.com) ### When did Micron actually move HBM4 into production? Micron said on March 16 that it had “begun volume shipment” of HBM4 36GB 12H in the first quarter of calendar 2026. The company used “high-volume production” in the headline of the same release, and its product page now describes HBM4 36GB 12H as being in high-volume production. (investors.micron.com) That matters because earlier Micron messaging, published about 11 months before the March announcement, had said the company planned to ramp HBM4 in calendar 2026 and had shipped samples to key customers. The March release marked the shift from sampling and planned ramp to volume shipment. (investors.micron.com) ### Is NVIDIA really the first named customer? Micron said the HBM4 36GB 12H part “is designed for NVIDIA Vera Rubin,” and the release tied other newly announced Micron products, including SOCAMM2 modules, to the same platform. NVIDIA, in a separate March 16 announcement, said Vera Rubin was in full production as part of its next AI infrastructure platform. (investors.micron.com) NVIDIA’s Vera Rubin platform combines the Vera CPU, Rubin GPU, networking and storage components into rack-scale systems for AI training and inference, according to NVIDIA. Micron did not name additional HBM4 platform customers in the March release. (investors.micron.com) ### What supports the claim that capacity is sold out through 2026? Micron said in its fiscal first-quarter 2026 presentation that it had “completed agreements on price and volume” for its entire calendar 2026 HBM supply, including HBM4. The company did not use the phrase “sold out” in that slide, but the statement indicates its 2026 HBM output had already been allocated under customer agreements. (investors.micron.com) That distinction matters because HBM includes more than one generation. Micron’s disclosure covered its entire 2026 HBM supply, while the March 16 product release separately confirmed that HBM4 had already entered volume shipment in the first quarter of calendar 2026. (nvidianews.nvidia.com) ### How large is Micron’s spending push behind AI memory? Micron said in December 2025 that it planned to raise fiscal 2026 capital expenditures to about $20 billion, primarily to support HBM supply capability and its 1-gamma DRAM node in calendar 2026. By March 2026, company-linked materials and third-party coverage described fiscal 2026 capex above $25 billion, with spending directed to additional capacity and cleanroom investments across multiple sites. (investors.micron.com) Micron has tied that spending to demand visibility. CEO Sanjay Mehrotra said in the December quarter release that the company was investing to support customers’ growing need for memory and storage, while the first-quarter presentation said HBM total addressable market could rise from about $35 billion in 2025 to around $100 billion in 2028. (investors.micron.com) ### What comes after the 36GB HBM4 launch? Micron said on March 16 that it had already shipped samples of a 48GB, 16-high HBM4 product to customers. The company said that package would raise capacity per HBM placement by 33% versus the 36GB 12-high offering. (investors.micron.com) Micron’s next milestones are likely to show up in its quarterly disclosures and product announcements. The company has already said fiscal 2026 capex is running above prior plans, and its disclosed customer agreements cover all calendar 2026 HBM supply, including HBM4. (futurumgroup.com) (investors.micron.com) (investors.micron.com)