IBM Unveils World’s First Sub-1-Nanometer Chip Prototype: What It Means for AI

IBM’s research team has created the world’s first chip with technology below one nanometer, using a transistor design at the 0.7 nanometer level. This puts IBM at the leading edge of semiconductor development, as the AI hardware industry works to fit more power into smaller chips while using less energy. Despite the announcement, IBM’s stock fell early in the day due to a wider tech selloff. Here’s what this breakthrough means for the future of AI chips.

Today, most AI chips are made using 5-nanometer and 4-nanometer processes, mainly by Taiwan Semiconductor Manufacturing Company. NVIDIA’s Blackwell GPU uses a custom version of TSMC’s 4nm process, and the upcoming Rubin platform will use TSMC’s 3nm technology. TSMC has also started mass-producing 2nm chips for the next generation. IBM’s 0.7nm prototype goes further by using a nanostack design, which builds transistors vertically instead of sideways to fit more computing power into the same space.

This breakthrough uses a new method called nanostacks, which comes after nanosheet technology. Nanosheets are currently the most advanced technology, following the FinFET design that has powered top chips for the last ten years. IBM Research Director Jay Gambetta called nanostacks a completely new approach and compared it to IBM’s 2nm prototype from May 2021. That earlier design is now being made by major manufacturers, including Rapidus, a Japanese company that licensed IBM’s 2nm Gate-All-Around nanosheet technology.

IBM is not saying the chip is ready for commercial use. Gambetta made it clear that this news is only about research, and moving to production will depend on whether costs match the performance gains. He said early results suggest this could be possible.

IBM stopped making chips itself in 2015, handing over its factories to GlobalFoundries and focusing on licensing its chip research instead of manufacturing. The company is also putting a lot of resources into quantum computing through the Anderon Foundry, which has $1 billion in backing from IBM and the Commerce Department. Gambetta said that classical and quantum computing are both important for IBM, and that their work on transistors helps them advance quickly in quantum technology.