IBM has built a working computer processor with 50 billion transistors in an area the size of a fingernail, making it the smallest and densest computer chip architecture yet produced. If it is adopted for use in data centres and consumer electronics, it could slash the vast amount of energy required for computing.
The computer chip manufacturing industry names generations of chip technology by size class, though these don’t refer to the actual size of the chip. New commercial devices tend to use 7-nanometre or 5-nanometre chips. For instance, Apple’s M1 chip in its current range of computers and its A14 chip that powers current iPhones both use 5-nanometre chips.
IBM claims that it has created the first working 2-nanometre chip. The transistors in this prototype are 12 nanometres wide, which is just 24 silicon atoms across. Decreasing the size of transistors can make chips smaller, faster and more efficient. For mobile devices this can mean a longer battery life.
IBM says that 2-nanometre chips can be used in everything from phones and tablets to very high-performance server chips and supercomputers, and could achieve 45 per cent higher performance and 75 per cent lower energy use than the 7-nanometre chips in production today.
Steven Freear at the University of Leeds in the UK says that denser chips would bring a range of benefits beyond power efficiency, including improved reliability due to the ability to include more security and error-checking logic.
Mukesh Khare at IBM says the prototype chip was made at the company’s headquarters in Albany, New York, and that the process involves “hundreds and thousands of steps” in a 9300-square-metre clean room that runs 24 hours a day, every day of the year. IBM says that the chips will go into production towards the end of 2024 but that it will outsource this to factories elsewhere.
Read more: There’s a global shortage of computer chips – what’s causing it?
Chips as we know them today can be traced back to 1959 at Bell Laboratories in the US, where transistors were made by carefully oxidising a piece of silicon. These tiny devices are the building blocks of modern electronics and all advances since then have essentially been down to miniaturising similar technology.
The miniaturisation of chips is famously governed by Moore’s Law, which states that the number of transistors on a processor will double every two years. This proved to be true for decades but has slowed in recent years.
Kieran Brophy at Imperial College London says that improved chip design based on the transistor is rapidly approaching fundamental physical limits: “It’s common knowledge that Moore’s Law is breaking down. You fundamentally can’t just keep going to smaller and smaller scales. You can’t change physics.”
The global chip industry publishes a timeline of expected technological jumps in miniaturisation so that companies can work together to develop the required manufacturing and testing processes. Currently, this roadmap states that 2-nanometre chips will be launched in 2023, with companies like Taiwan Semiconductor Manufacturing Company expected to begin manufacturing commercial products using it in 2024.
The EU is also pushing for 2-nanometre technology and is investing £125 billion for research and development, and to build manufacturing capacity.
Adoption of this technology could potentially make a big dent in climate change by reducing the energy demands of the vast data centres that power email, social media, banking and video streaming services. Data centres like this currently account for 1 per cent of the world’s electricity use.