This article is more than 1 year old
TSMC’s 2025 timeline for 2nm chips suggests Intel gaining steam
Semiconductor veteran says x86 titan could catch up with Asia-Pacific rivals in three years
TSMC said it won't start production at its 2nm node until the second half of 2025 or possibly the end of that year, which could signal a shift in the competitive landscape.
The Taiwanese chip foundry revealed the timeline for its 2nm node, known officially as N2, during a conference call [PDF] last week for its first-quarter financial results. With a mid- to late-2025 production timeline, after late-2024 risk production, TSMC's 2nm production dies will likely land in the hands of their designers in volume in 2026, which, in turn, means those chips could, at the earliest, be available for phones, PCs, and servers that year.
TSMC made the disclosure only a few days after Intel, which is revitalizing its competing foundry business, revealed that its next-generation 18A node will be ready for manufacturing in the second half of 2024, months ahead of the previously given 2025 timeline. As the A is short for ångströms, Intel's 18A label suggests it will be a 1.8nm process (see Register passim for caveats about node sizes.)
Now here's the fun part: does this mean Intel will beat TSMC to the market with a comparable manufacturing process? We should remain skeptical, though there is reason to believe competition is growing tighter and Intel could catch up or take the lead by 2025, assuming things keep going smoothly.
Intel CEO Pat Gelsinger has maintained that his x86 goliath will become the leader in process performance with its 18A node, and the company's recent disclosure that production will begin several months sooner shows the chipmaker is feeling confident right now.
At the same time, TSMC CEO CC Wei said on the earnings call that his company expects its 2nm node to provide the "best technology, maturity, performance and cost" for chip designers, though it is apparently entering production later than Intel's 18A node.
In a recent analysis, Scotten Jones, head of semiconductor consulting firm IC Knowledge, said Intel's recent acceleration in node development has led him to believe that the chipmaker could "flip the script" and leapfrog rivals TSMC and Samsung in performance with its 18A node.
If it actually happens, this would mark a reversal in the competitive landscape after Intel fell behind TSMC and Samsung in leading-edge manufacturing due to blunders with its 10nm and 7nm processes. At the very least, it marks a reversal in Jones' opinion as he previously believed Intel had no chance in catching up.
Intel eyes subscriptions to grow software sales from 2021's $100m
READ MORE"In conclusion we believe Intel has been able to significantly accelerate their process development at a time when the foundries are struggling," he wrote. "Although we don't expect Intel to regain the density lead over the time period studied, we do believe they could retake the performance lead."
Jones based his analysis on the most recent timelines given by Intel, TSMC, and Samsung as well as the performance improvements each corporation expects for new nodes. At the time Jones' analysis was published early last week, TSMC had not yet given the official timeline for its 2nm node, and he said he assumed it would arrive in 2025, roughly in line with what TSMC said in its earnings conference call.
Intel plans to introduce a total of four new nodes between now and 2024, starting with the Intel 4 node in the second half of 2022, followed by Intel 3 in 2023, Intel 20A in early 2024 and Intel 18A later that year.
TSMC and Samsung, on the other hand, only plan to introduce two nodes each in this timeline, with both of their 2nm nodes due in late 2025. TSMC, for one, is known for making components for AMD, Nvidia, and Apple.
Now it's important to remember that Intel renamed its next-generation manufacturing processes last year, which was done in an attempt to marry up [PDF] its node labeling with those of TSMC and Samsung. For example, Intel's 10nm Enhanced SuperFin node is now known as Intel 7, which means Intel believes the node is equivalent to its competitors' 7nm processes. Therefore, it follows that Intel 4, previously known as Intel's 7nm node, is equivalent to the 4nm nodes of rivals and so on.
However, it's ultimately the performance and efficiency of the products enabled by these nodes that will matter the most, so we will only truly know how Intel's next-generation nodes compare to TSMC's and Samsung's when they power competing chips available in the market. ®