Micron Technology has laid out where its memory business is headed for the rest of the decade, and the answer is mostly AI. The company’s roadmap runs through 2030 and includes vertical high-bandwidth memory and flash storage that GPUs can access directly.
The roadmap came out of a Micron technology webinar on October 21, 2026, led by President Scott DeBoer. The headline commitment: a new generation of high-bandwidth memory (HBM) every year through 2030.
Micron’s sequence starts with HBM4 in 2026 and runs to HBM6 in 2030. Custom HBM and vertical HBM options are slated to begin in 2028.
The vertical HBM concept, which Micron calls vHBM, targets bandwidth of approximately 2 TB/s. The company is aiming for energy use under 1 picojoule per bit.
The other attention-grabber is GPU-accessible flash. Flash is the slower, cheaper, persistent storage that normally lives a few steps removed from the processor. Letting GPUs reach it more directly could, in principle, give AI systems a larger pool of usable data without paying HBM prices for all of it.
DRAM and NAND get their own upgrades
On the conventional DRAM side, Micron plans to bring its 1δ (1-delta) node into volume production in the second half of 2027. The company expects bit density improvements in the mid-20s percent range.
Micron says the 1δ node will support next-generation memory standards including LPDDR6 and HBM5.
NAND flash is getting a similar push. Micron’s G11 NAND is expected to begin its volume ramp in the second half of 2027, with more than 400 layers.
The company expects G11 to deliver around 2x the bit density of G9. It is also targeting I/O speeds up to 1.5 times faster and a reduction of more than 25% in energy consumed per data transfer.
A skipped generation and a strategic pivot
Not everything on the old plan survived. Micron is skipping a planned generation, choosing instead to advance its 6F² work and pursue 3D DRAM development.
Supply stays tight, and customers are locking in
Micron expects memory supply conditions to remain tight through fiscal year 2028. That outlook is reportedly backed by strategic customer agreements that account for nearly 35% of sales through 2030.
What this means for the memory market
For investors, the clearest signal is that Micron is organizing its entire product cadence around AI and high-performance computing. An annual HBM cadence is an aggressive schedule, and it puts the company on a treadmill that only works if AI demand keeps compounding.
The architecture shift deserves attention too. Moving away from traditional scaling toward 6F², 3D DRAM, vertical HBM and GPU-accessible flash points to a market where memory designs grow more complex and more specialized. Custom HBM starting in 2028 hints that large customers may increasingly want memory tailored to their own chips rather than buying off the shelf.
The risks are the usual ones for an ambitious semiconductor plan: manufacturing delays, yield problems on new processes, and the possibility that AI spending cools before 2030 arrives. Production timelines and actual adoption rates of these new products will be the metrics to watch as each milestone comes due.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.