IBM Develops 0.7nm-Class Process with Nanostack Transistors for Major Efficiency Gains

According to Tom's Hardware, IBM has produced a test chip using its 0.7nm-class process, leveraging a dual-wafer nanostack architecture to deliver significant performance and efficiency gains.

Chips & Climate Desk
Updated Nov 7, 2026 2 min read
IBM Develops 0.7nm-Class Process with Nanostack Transistors for Major Efficiency Gains

According to Tom's Hardware, IBM has successfully produced a test chip utilizing its new 0.7nm-class, or 7 angstroms, fabrication technology, marking a notable milestone as the industry's first sub-1nm manufacturing process. Announced on Thursday, the concept process relies on nanostack transistors to deliver substantial power, performance, and area gains relative to the company's existing 2nm-class node. To realize these nanostack structures, IBM deviates from standard practices by employing two wafers instead of one, supplemented by ultra-thin dielectric bonding in an arrangement that has not been deployed previously.

Tom's Hardware reports that this 7A-class process yields up to 50 percent higher performance and 70 percent higher energy efficiency compared to IBM's 2nm-class node, which utilizes nanosheet gate-all-around transistors introduced in 2021. Furthermore, the underlying architecture delivers 40 percent higher SRAM density alongside significant density improvements for logic transistors. These gains stem from the nanostack design, which conceptually resembles complementary field-effect transistors by separating n-type and p-type transistors into vertically bonded tiers rather than housing them side-by-side in a single standard-cell layer.

By transforming the traditional two-dimensional layout into a three-dimensional stacked configuration, IBM achieves roughly double the transistor density of its 2nm research node without relying on conventional planar shrinks, as noted by Tom's Hardware. Instead of fabricating complementary transistors on the same wafer using identical materials, IBM builds them on separate wafers and integrates them via ultra-thin dielectric bonding. This methodology allows engineers to optimize n-type and p-type channels independently, tailoring process conditions, channel materials, strain engineering, or geometries for each tier.

Despite these architectural advantages, Tom's Hardware outlines several caveats associated with using two active device tiers instead of a single-tier logic node. The primary manufacturing challenges involve alignment and bonding yields, as two advanced logic wafers require extreme precision to prevent interface defects from ruining the stack. Additional complications include more complex power delivery and routing, increased thermal resistance since the active tier sits farther from the heat sink, and higher overall production costs driven by dual advanced front-end-of-line wafers and extra thinning steps.

Given these economic and technical hurdles, Tom's Hardware notes that the approach is likely best suited for heavy-duty data center artificial intelligence solutions near reticle size rather than mainstream client processors. Crucially, IBM completed the test chip at its Albany, New York research facility using proven Low-NA EUV lithography rather than High-NA EUV tools. Because IBM's developments serve as pre-competitive intellectual property and research-and-development know-how rather than turn-key high-volume manufacturing processes, the company believes nanostack technology could potentially enter mass production within the next five years.

Sources

  1. 01 IBM goes sub-1nm, develops 0.7nm-class technology — offering up to 50% higher performance and 70% higher energy efficiency compared to IBM's 2nm-class node — Tom's Hardware