It’s time to stop using older nodes and move to 28nm
We tend to discuss leading-edge nodes and the most advanced chips made with them, but there are thousands of chip designs developed years ago that were created using what are now mature process technologies that are still widely used by the industry. In terms of performance, those chips still do their job as perfectly as the day the first chip was made. Therefore, product manufacturers continue to use them more and more. But on the manufacturing side, there’s a hard bottleneck for further growth: all the old hub capacity that will ever be built has been built – and they won’t build anymore. As a result, TSMC has recently begun strongly encouraging its customers at the oldest (and least dense) nodes to migrate some of their mature designs to its 28nm class process technologies.
Today TSMC earns around 25% of its revenue by making hundreds of millions of chips with 40 nm and larger nodes. For other foundries, the revenue share of mature process technologies is higher: UMC gets 80% of its turnover at 40 nm higher nodes, while: 81.4% of SMIC’s revenue comes from outdated processes. Mature nodes are inexpensive, have high efficiency, and provide sufficient performance for simplistic devices such as power management ICs (PMICs). But the cheap wafer prices for these nodes stem from the fact that they were once, long ago, leading nodes themselves, and their construction costs were paid for by the high prices that an advanced process can deliver. That is, there is not the profitability (or even the equipment) to build new capacity for such old nodes.
Therefore, TSMC’s plan is to increase production capacity for mature and specialized nodes by 50%, targeting 28nm capable fabs. As the last (viable) generation of TSMC’s classic, pre-FinFET manufacturing processes, 28nm is positioned as the new sweet spot for producing simple, low-cost chips. And in an effort to consolidate production of these chips around fewer and more widely available/expandable production lines, TSMC wants to move customers using legacy nodes to the 28nm generation.
“We are currently not [expanding capacity for] the 40 nm node,” said Kevin Zhang, senior vice president of business development at TSMC. “You build a fab, fab doesn’t come online [until] two years or three years from now. So you really have to think about where the future product is going, not where the product is now.”
While TSMC’s 28nm nodes are still subject to the same general cost trends as chip factories in general – in that they are more complex and expensive per wafer than even older nodes – TSMC wants to convert customers to 28nm by balancing that against the much larger number of chips per wafer that the smaller node provides. Therefore, while companies will have to pay more, they will also be able to get more in terms of total chips. And none of these take into account possible additional benefits of a newer node, such as lower power consumption and potentially greater clock speed (performance).
“So, many customers’ product these days is at, say, 40nm or even older, 65nm,” Zhang said. They go to lower advanced nodes. 20/28 nm will be a very important node to support future specialty. […] We work with the customer to accelerate [their transition]. […] I think the customer is going to get an advantage, economic advantage, economies of scale, you have better power consumption. but they already have a chip that works. Why? Oh, then you could say why we use advanced technology. Yes. Yes. I mean, it’s, uh, the nature of the peak is you go to the next node, you get better performance and more power and overall you get a system-level advantage.”
In addition to multiple 28nm nodes designed for different client applications, TSMC is expanding its range of dedicated 28nm and 22nm (22ULP, 22ULL) process technologies to address a variety of chip types that currently rely on various legacy technologies. As with the overall shift to 28nm, TSMC is trying to encourage customers to use the newer, higher-density process nodes. And if not 28nm/22nm, customers also have the option to move to even more capable FinFET-based nodes, which are part of TSMC’s N16/N12 family (e.g. N12e for IoT).