Key Pointers:
- Built on a 0.7nm architecture
- Uses Nanostack 3D transistor design
- Nearly 100 billion transistors on one chip
There comes the shift in the semiconductor industry, as IBM has officially unveiled the world's first sub-1 nanometer (nm) chip technology, which is developed at 0.7 nm and packs nearly 100 billion transistors into a small, human fingernail-sized device. This announcement happened at the VLSI 2026 Symposium, which was placed to represent the restructuring of silicon design.
Mike Murphy, an IBM researcher, stated, “With these sorts of power gains, the potential for 7 angstrom devices is sky high, with a massive potential impact on the world of AI. Today’s popular AI accelerators can produce about 1,500 TOPS (or trillions of operations per second), and IBM researchers estimate one using 7 angstrom technology could deliver about six times more, or around 9,000 TOPS. So if 7 angstrom chips were used to train today’s massive, frontier-model LLMs, we could drastically cut a typical training time from around three months to a couple weeks."
Reinventing Silicon with Nanoistack
IBM Research has evolved the chip from the traditional horizontal architectures to a sophisticated design called 'Nanostack.' Unlike other chips, this Nanostack architecture has vertical layers and transistors that use 3D sequential integration. This vertical scaling allows us to offer more processing power on the same size while moving different material combinations to separate optimized performance and electrical resistance. According to IBM’s published technical metrics, processors designed on this 0.7 nm platform are made to offer a forward move in their capability over the current 2 nm nodes.
- With this, its performance speed will increase by 50% at the same power levels.
- There will be a 70% reduction in energy consumption while maintaining the performance.
AI Memory Wall
Apart from processing, the 0.7 nm node specifically targets high-speed data delivery. IBM's testing has confirmed that the Nanostack architecture successfully reduces random-access memory (SRAM) by 40%. As more AI models are demanding more computing power, smaller chips can play a major role in improving AI preference. The SRAM sits just adjacent to the processor, it serves as the high-bandwidth short memory which can relieve high-density memory for heavy data processing for generative AI infrastructure and data centers.
In 2021, IBM introduced its 2 nanometer chip technology, but its new sub-1-nm architecture nearly doubles transistor density compared to the previous generation. This shows how IBM continues to work and push semiconductor research despite manufacturing industry challenges.
Disclaimer: At Jagran Reviews, we prioritize providing up-to-date information on the latest trending commodities and goods. Links in this article are affiliate links. Please note that our product recommendations are based on independent research and analysis however, Jagran Reviews shall not be held liable for any claims, damages, or disputes arising under applicable laws, including the Consumer Protection Act, 2019. Any issues regarding product quality, delivery, or warranties must be directed to the respective manufacturer or seller. Furthermore, the products featured in this article are organized for reader convenience and are listed in no particular order of priority or ranking.