IBM Unveils Nanostack Chip to Extend Moore's Law for Another Decade
IBM has unveiled a groundbreaking nanostack chip with 100 billion transistors, promising to extend Moore's Law for another decade. This vertical stacking technology will lead to significantly faster and more energy-efficient computers.
A
··2 min readAgent
Newsroom

IBM has unveiled a groundbreaking chip technology that promises to extend the trajectory of Moore's Law for another decade, paving the way for significantly faster and more energy-efficient computing. The tech giant's new prototype chip boasts an astonishing 100 billion transistors packed onto a surface area no larger than a fingernail, effectively doubling the density of its previous cutting-edge technology from 2021. This innovative design, dubbed "nanostack," represents a pivotal shift in chip architecture, moving beyond traditional horizontal scaling to a vertical stacking approach.
For over half a century, the semiconductor industry has driven computational power forward by adhering to Moore's Law, which dictates the doubling of transistors on a chip roughly every two years. This was primarily achieved by continually shrinking the size of individual transistors. However, in recent years, transistors have approached their physical limits, reaching sizes where quantum mechanics begins to interfere with their functionality, making further miniaturization impossible. To overcome this fundamental barrier, engineers across the industry are now adopting a strategy akin to urban planning: building upwards rather than outwards.
IBM's nanostack architecture is a prime example of this vertical integration. The new design vertically stacks transistors in two distinct layers on a single silicon chip. Jay Gambetta, director of IBM Research, emphasized the significance of this achievement, stating it's "not just an incremental step... It’s a meaningful leap forward." Industry experts like Dan Hutcheson, vice chair of TechInsights, concur, calling it "transformational" and predicting it "puts another 10, 15 years on the roadmap." IBM reports that chips built with this new approach can perform up to 50% more work in the same timeframe and achieve up to 70% greater energy efficiency, a crucial factor for data centers grappling with power consumption.
The fabrication process for IBM's nanostack chip involves building layers akin to a cake. Transistors are first created on one silicon layer, then another silicon layer is placed on top, and a second set of transistors is fabricated directly above the first. Electrical connections are then established between these two layers. This vertical stack, combining two types of transistors, is known as a complementary field-effect transistor (CFET). While other major chip manufacturers like Intel, Samsung, and TSMC, along with research labs like Imec, are also exploring CFETs, IBM distinguishes its design by staggering the transistors in the second layer relative to the first, which simplifies wiring and offers other advantages over alternative two-tiered approaches like AMD's 3D V-Cache.
Despite the promise, scaling this technology further presents significant challenges. While chipmakers might aim for more tiers, manufacturing errors become more problematic; a fault in any layer renders the entire chip unusable, leading to higher failure rates and increased costs. Another critical hurdle is the "thermal budget," requiring engineers to develop each subsequent layer without damaging the connections of the layer beneath, necessitating manufacturing processes below 400 °C. IBM has reportedly overcome this thermal challenge for its two-layer design, though its specific methods remain undisclosed. Academic research, such as that by Professor Qing Cao's group, is also contributing to solutions for low-temperature stacking, ensuring the future of advanced chip design continues to evolve.




