IBM, Synopsys Move Toward 1.4-nm Node With Heat-Modeling Tech

2026-02-04 · Source: Big Data & AI News - EE Times · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Expert, medium

Summary

IBM, in collaboration with Synopsys and with support from DARPA, has developed a new machine learning (ML) tool for thermal modeling that aims to enable the 1.4-nm node. This technology, part of DARPA's Thermonat project, models chip thermal behavior down to the atomic level, achieving prediction accuracy within 1 degree Celsius and operating tens of thousands of times faster than existing simulation tools. The ML software was trained on IBM's semiconductor data and is expected to be crucial for performance at 1.4-nm and below, initially targeting data centers and high-performance AI applications. IBM plans to share this technology with chipmaking partners as 2-nm node manufacturing, led by companies like TSMC and Samsung, ramps up. Synopsys contributed a reduced-order modeling approach and a Fourier neural operator-based ML thermal solver, which delivers up to a 1,000x speed-up for designs with over 1 million transistors.

Key takeaway

For AI Scientists and chip designers working on advanced nodes, this new ML-powered thermal modeling tool offers a significant advantage. You should explore integrating such rapid, accurate thermal prediction capabilities into your design workflows to optimize chip layouts, improve cooling solutions, and achieve higher performance or lower power consumption, especially for 1.4-nm and 3D-IC architectures. This technology can help mitigate thermally driven failures and accelerate development cycles.

Key insights

A new ML-powered thermal modeling tool accelerates nanoscale chip design, enabling 1.4-nm nodes and beyond.

Principles

• Heat management is critical for nanoscale transistor performance.
• ML can drastically accelerate complex physics simulations.

Method

The method involves training ML software on semiconductor data to predict thermal behavior, utilizing a Fourier neural operator for solving partial differential equations and reduced-order modeling for computational efficiency.

In practice

• Use ML for thermally aware chip layout design.
• Apply reduced-order modeling to speed up simulations.
• Integrate thermal prediction into 3D-IC and packaging design.

Topics

• Nanoscale Transistors
• Thermal Modeling
• Machine Learning
• 1.4-nm Node Technology
• Advanced Packaging

Best for: AI Scientist, AI Hardware Engineer, AI Engineer, Research Scientist

Related on AIssential

• Thermonat: A machine learning tool for semiconductor modeling — High-accuracy, high-speed thermal simulation of individual transistors is crucial for advanced circuit design.
• IBM unveils world’s first sub-1nm chip with new nanostack architecture — IBM's nanostack architecture enables the world's first sub-1nm chip, significantly boosting transistor density and energy efficiency.
• Space-Station Tech Pivots to Cool AI Data Centers — Space-station derived liquid cooling and carbon nanotube TIMs are critical for managing extreme heat in next-gen AI data centers.
• PALTO: Physics-Informed Active Learning for Tri-Gate FinFET Design Optimization for Vertical Power Delivery — The PALTO framework uses physics-informed active learning to efficiently optimize GaN FinFET designs, outperforming traditional methods and industrial benchmarks.
• China Just Built What TSMC Said Was Impossible — The next leap in computing performance may come from optimizing data movement and 3D packaging, not just transistor shrinkage.
• This new chip survives 1300°F (700°C) and could change AI forever — A new memristor design withstands 700°C, enabling extreme environment electronics and efficient AI computation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Big Data & AI News - EE Times.