Global semiconductor foundry TSMC has just confirmed that production of the 2-nanometer process node is on track. This means that computer chips of unprecedented power lie ahead.
According to current estimates, chips based on 2nm architecture will go into mass production in 2025.
The information comes from CC Wei, chief executive of Taiwan Semiconductor Manufacturing Company (TSMC), and was disclosed during an earnings call. TSMC first confirmed it was working on a 2nm process node in 2020, but it kept the details quite sparse. This time, the company revealed a little more about the architecture of the node and shared new updates on its planned roadmap.
A good amount of technical information was shared as part of the profit call. The new N2 node will rely on gate-all-around (GAA) transistors, which represents a change from the current Fin Field-Effect Transistor (FinFET) structure. The nodes are further fabricated based on extreme ultraviolet (EUV) lithography.
Such technicalities may not mean much to most end users, but TSMC has not shared many expectations when it comes to performance. However, it is not the first foundry to work on a 2nm process, and some of its competitors have already made serious progress. IBM unveiled its first 2nm chip last year.
IBM’s breakthrough was significant. The manufacturer said that using its 2nm process, it was able to fit as many as 50 billion transistors into a chip the size of a fingernail. That’s 20 billion more than the 5nm process had when IBM announced it in 2017.
All major manufacturers are slowly finding their way to smaller and smaller process nodes. Switching from 7nm to 5nm, and from 5nm to 3nm in the future, will bring significant improvements in performance and thermals. Smaller transistors consume less power, take up less space, and push the performance ceiling up because more cores can be stacked in a smaller chip.
The information was first shared by Tom’s Hardware. TSMC has consistently delivered process node upgrades about once every two years, with enhanced and modified versions of pre-existing nodes in between. This time it seems to have been delayed a bit. Wei confirmed that while “progress is in line with our expectation”, risk production will not start until around the end of 2024. After that, the chips will go into high volume production (HVM) in 2025, probably around the second half of the year, or even the end of it.
Switching to 2nm could result in performance we can’t imagine in consumer computers with today’s hardware, but we’ll have to wait. While the chips are on track to go into mass production in 2025, it’s unlikely we’ll see them on our PCs before the second half of 2026. It is a long way before the 2nm process is incorporated into products coming to the market.