ClareNow
Search
ClareNow
Toggle sidebar
Semiconductors → Neutral

Memory Capacity Production Slowly Expanding To Meet AI Industry Demand

Samsung, SK hynix, Sandisk, Kioxia and Micron report building out new DRAM and NAND flash manufacturing and packaging facilities coming on line in late 2027 or 2028.

Forbes 1 min read 7/10
Memory Capacity Production Slowly Expanding To Meet AI Industry Demand
Key Takeaways
  • Samsung, SK hynix, Sandisk, Kioxia, and Micron are constructing new DRAM and NAND flash fabs with initial production slated for late 2027 or 2028.
  • High-bandwidth memory (HBM) — critical for AI accelerators — is the primary driver, with HBM3E and HBM4 generations already in development.
  • SK hynix dominates the HBM market with over 50% share, while Samsung and Micron are racing to scale their own HBM capacity.
  • NAND flash expansion targets AI inference storage, with 3D NAND layers exceeding 500 and new packaging techniques like hybrid bonding.
  • Industry forecasts suggest global memory bit supply will grow 15-20% annually through 2028, but demand from AI could outstrip supply until new fabs ramp.
The global memory industry is finally gearing up to meet the insatiable appetite of artificial intelligence — but the payoff is still two to three years away. Samsung, SK hynix, Sandisk, Kioxia and Micron are building new DRAM and NAND flash manufacturing and packaging facilities, with capacity coming online in late 2027 or 2028. This slow expansion reflects the massive capital investment and long lead times required to construct advanced semiconductor fabs, even as AI workloads push memory demand to historic highs.

Frequently Asked Questions

AI training and inference require enormous amounts of memory, particularly high-bandwidth memory (HBM) for accelerators and NAND flash for storage. Current supply cannot keep up with demand, prompting major memory makers to build new fabs.

Samsung, SK hynix, Sandisk, Kioxia, and Micron are constructing facilities that will begin production in late 2027 or 2028. Ramp-up to full capacity may extend into 2029.

The major players are Samsung, SK hynix, Micron, Sandisk (Western Digital), and Kioxia. Each is investing billions in DRAM and NAND flash manufacturing lines.

High-bandwidth memory (HBM) is critical for AI accelerators like GPUs. DRAM is used for system memory in servers, while NAND flash provides high-capacity storage for training datasets and inference.

Expanded capacity will help lower memory costs and remove supply constraints, enabling faster deployment of larger AI models. Until then, AI companies may face higher costs and limited availability of premium memory.

Original source

www.forbes.com

Read original

Discussion

Join the discussion

Sign in to post a comment or reply.

No comments yet. Be the first to share your thoughts!

Sign in
Enter your email to receive a one-time sign-in code. No password needed.
Email address