Micron Technology is challenging industry enthusiasm around Micron 4F2 DRAM, arguing that the design offers little density advantage over alternatives while adding cost and complexity. President and Chief Technology and Products Officer Scott DeBoer said DRAM density depends primarily on capacitor spacing, not the cell layout, calling 4F2’s claimed advantage over 6F2 “a bit of a myth.”
Micron can manufacture 4F2 DRAM at older process nodes, but DeBoer said doing so is more expensive and complex. At the leading edge, he described fundamental scaling problems that could lead to 3D DRAM replacing 4F2. Micron plans to keep 3D DRAM as a parallel development path between its 1-zeta and 1-eta nodes, ready for implementation when its economics improve.
Micron 4F2 DRAM: Why Micron Is Rejecting Its 4F2 Commercial Path
Asked by UBS analyst Tim Arcuri about competitors pursuing 4F2, DeBoer said Micron invented the design and has studied it for two decades, yet chose not to commercialize it. The core issue is scaling: although 4F2 may appear compact in a cell layout, the spacing between capacitors remains the decisive constraint on density.
The distinction matters because 4F2 and 6F2 describe the area occupied by each memory cell. Micron’s position is that cell-layout terminology alone does not establish a meaningful density gain. The company sees 3D DRAM as the more viable route once the economics work, rather than continuing to pursue 4F2 at advanced nodes.
Micron’s roadmap includes 1-delta DRAM, its fifth consecutive node, scheduled to begin ramping in the second half of 2027 in Boise. The node is expected to improve bit density by a mid-20% percentage and is designed for HBM5 and vertical HBM. Development of 1-epsilon DRAM, which uses wafer bonding, is also underway.
HBM4 and Vertical Memory Define the Next Roadmap
Micron said its HBM4 memory is already in high-volume production, with each 12-high stack delivering 2.8 terabytes per second of bandwidth. DeBoer said roughly 100 stacks could match the world’s total internet bandwidth, illustrating the scale of the bandwidth requirements surrounding AI systems.
The company is also developing vertical HBM, which places memory directly on top of a processor. Micron says the approach can increase bandwidth through co-design while continuing to use HBM around the edges of many XPUs. Hybrid bonding, which connects chips at much tighter spacing, will be essential to achieving the required performance.
Another product, vertical edge inference memory, stacks low-power DRAM on a logic die for applications such as robots and cars. Micron says it can provide more bandwidth than processing-in-memory designs, which embed computing functions inside the memory chip. The company is pursuing both approaches.
Micron is also moving from Gen 9 to G11 NAND, skipping a generation. G11 is scheduled to ramp in the second half of 2027 with more than 400 layers, twice Gen 9’s bit density and 1.5 times its input/output speed. It uses wafer-on-wafer bonding.
For mobile memory, Micron pointed to a Meta white paper on LPDDR6, which the company presented as an example of its research work. The broader product discussion also included DDR6, designed to deliver twice the bandwidth, and LPDDR6, which is expected to provide 35% better power performance.
Tighter Supply Could Last for Years
Micron expects memory supply to remain tight for at least two years. CEO Sanjay Mehrotra said the company does not see a clear point when supply and demand return to balance. DeBoer characterized memory as the AI bottleneck and said more memory leads to better system performance.
The supply outlook is complicated by the new vertical products. Micron said both use more wafers per bit than conventional DRAM, and that the relevant trade ratios continue to increase. As advanced architectures demand more wafer capacity for each bit of memory, the shift could add further pressure to an already constrained supply chain.
Micron is using AI internally to accelerate development, cutting the time from silicon development to tape-out for a new data-center SSD controller by half. The company also developed a data-center SSD about 60% faster than the previous generation. Micron reported record quarterly revenue of $54.23 billion and forecast $61.5 billion for the next quarter.
For DRAM buyers, the abandonment of Micron 4F2 DRAM production marks a clear signal that capacity will be redirected toward AI-oriented products rather than commodity memory. Customers that had been evaluating Micron 4F2 DRAM for cost-sensitive workloads may need to reassess sourcing options, particularly as the company indicated that vertical architectures consume more wafers per bit than conventional designs. This could tighten availability for traditional DRAM buyers even before the broader supply-demand imbalance resolves.
From a regulatory and antitrust perspective, the pivot raises questions about concentration in advanced memory. As Micron narrows its commercial focus to HBM and 3D DRAM, competitors such as SK hynix and Samsung are pursuing similar roadmaps, potentially leaving fewer alternatives for buyers of leading-edge memory. Any future scrutiny of AI-related supply chains is likely to examine whether such strategic convergences reduce buyer leverage.
For developers and system architects, the message is that future platforms will be designed around stacked memory and HBM4 rather than planar DRAM scaling. Software and firmware teams building for AI training and inference workloads should plan for increased bandwidth and capacity per device, alongside tighter power envelopes. The deprioritization of Micron 4F2 DRAM underscores that conventional density gains alone are no longer the industry’s primary path forward.
For investors, the key question is whether Micron’s decision to move away from commercial Micron 4F2 DRAM reflects a differentiated technology strategy or a competitive risk. The company is betting that 3D DRAM, vertical HBM and HBM4 will capture the next wave of AI demand, even as its roadmap targets 2027 and supply remains constrained. Micron 4F2 DRAM may remain a valuable research platform, but DeBoer’s comments make clear that Micron does not expect it to define the company’s commercial future.
Source: https://finance.yahoo.com/technology/articles/micron-says-rival-dram-design-161841951.html