IBM on Thursday introduced its first processor built with the company’s newest semiconductor technology, cramming nearly 100 billion transistors into a piece of silicon no larger than a fingernail. Increasing transistor density on a die of the same or smaller footprint is crucial for boosting both power efficiency and speed.
The new processor is fabricated at a 0.7‑nanometer node, which is even smaller than IBM’s 2‑nanometer chip unveiled in 2021. However, the circuit layout has been dramatically altered. The older, larger process placed transistors flat in structures IBM Research called nanosheets, as Gfaloe reported in 2021. The 0.7 nm chip now employs IBM’s newly developed nanostack architecture, stacking the nanosheets vertically.
According to IBM, this architecture delivers superior performance. In internal tests the chip showed up to a 50 % performance gain and a 70 % improvement in energy efficiency compared with the 2 nm version.
IBM also says the nanostack design shrinks SRAM die size by about 40 %. SRAM – static RAM – retains data without a constant power supply and, being faster than DRAM, is highly sought after for AI workloads.
The processor is not yet ready for market. IBM is still collaborating with its manufacturing partner Rapidus, a Japanese foundry, to scale production. The company “sees a path to production” within five years, while demand for energy‑efficient compute continues to rise.
Chips from IBM, Nvidia, AMD and others form the backbone of the AI industry. As AI developers such as OpenAI and Google push for ever‑more advanced models, they require massive compute resources, which in turn consume large amounts of electricity, water and land for data centers.
“Everyone demands more performance, but no one wants to pay for the bill for the power,” said Huiming Bu, vice president of IBM semiconductor R&D. “The new chip’s energy efficiency is a very critical component for AI.”
Developing more efficient hardware is a key piece of the AI‑driven future envisioned by tech leaders. Current shortages in memory, processors and other components have created supply gaps for laptops and other gadgets. New research prototypes like this could help companies—and their customers—get more value out of each device.
“Fundamentally, it comes down to, can we make transistors more efficient?” said Jay Gambetta, director of IBM Research. “This is a platform that can be customized, so our expectation is it’ll impact everything from the logic to the SRAM, and as it scales we’ll see larger, more efficient AI accelerators.”
