According to Digitimes, TSMC and UMC's 22/28nm process capacity utilization will continue until at least 2022, according to the supply chain, as capacity vacated by some consumer chip customers will soon be used by companies including Sony and Infineon. Other customers fill in.
TSMC calls on customers: abandon 40nm and move to 28nm
When it comes to chip manufacturing, we talk more about leading edge nodes and the most advanced chips made using them, but in reality there are thousands of chip designs that were developed years ago that are still widely adopted by the industry today. Manufactured with mature technology. In terms of execution, these chips are still as perfect as when the first chip was made, which is why product manufacturers are increasingly using them for their manufacturing.
But on the manufacturing side, there is a hard bottleneck for further growth of these chips: that is all the capacity of the old nodes has been built -- and they won't be built any more. As a result, TSMC has recently begun to strongly encourage customers on its oldest (and lowest density) nodes to migrate some of their mature designs to its 28nm-scale process technology.
Today, about 25% of TSMC's revenue comes from making hundreds of millions of chips using 40nm and "older" nodes. For other foundries, mature process technologies have a higher revenue share: 80% of UMC's revenue comes from higher 40nm nodes, while 81.4% of SMIC's revenue comes from outdated processes. Mature nodes are inexpensive, have high yields, and provide sufficient performance for simple devices such as power management ICs (PMICs). But the cheap wafer prices for these nodes come from the fact that they were the leading nodes themselves a long time ago, and their construction costs were paid for by the high prices available on cutting-edge processes. That is, there is no profitability driver (or even equipment) to build new capacity for these old nodes.
That's why TSMC plans to expand capacity at mature and specialty nodes by 50% and focus on fabs with 28nm capabilities. As the last (and viable) generation of TSMC's classic pre-FinFET manufacturing process, 28nm is positioned as the new sweet spot for producing simple, low-cost chips. And, to consolidate the production of these chips into fewer and more widely available/scalable production lines, TSMC wants to keep customers using older nodes on the 28nm process.
"We don't have [expanded] capacity at the 40nm node right now," said Kevin Zhang, senior vice president of business development at TSMC. "You build a fab, and the fab doesn't come online in two or three years. So, you really need to consider where future products are headed, not today’s products.”
While TSMC's 28nm nodes are still generally subject to the same overall cost trends as chip fabs -- because they are more complex and more expensive than older nodes on a per-wafer basis -- TSMC hopes to convert customers by 28nm to balance this with the higher number of chips per wafer offered by the smaller node. So while companies will have to pay more, they will also get more total chips. And none of this takes into account the potential ancillary benefits of the new node, such as reduced power consumption and possibly greater clock speed (performance) headroom.
"So, many customers today are 40-nanometer or even earlier 65-nanometer," Zhang said. They are moving to more advanced advanced nodes. 20/28 nm will be a very important node to support future specialties. We're working with customers to accelerate [their transition]. [...] I think the customer will get the benefit, the economic benefit, the scaling benefit, you'll get better power consumption. But they already have a working chip. Why do this? Because going to the next node, you get better performance and better power, and overall you get system-level gains. "
In addition to multiple 28nm nodes designed for various customer applications, TSMC is also expanding its lineup of specialized 28nm and 22nm (22ULP, 22ULL) process technologies to address a wide range of chip types that currently rely on various obsolete technologies. As with the overall move to 28nm, TSMC is looking to attract customers to newer, higher-density process nodes. And, if not 28nm/22nm, customers also have the option to transition to more powerful FinFET-based nodes that are part of TSMC's N16/N12 series (eg, N12e for IoT).
TSMC expands capacity of mature and professional nodes by 50%
TSMC revealed at an earlier technology summit that by 2025, the production capacity of its mature and professional nodes will expand by about 50%. The plan includes the construction of a large number of new fabs in Taiwan, Japan and China. The move will further intensify competition between TSMC and chip foundries such as GlobalFoundries, UMC, and SMIC.
The surge in demand for various computing and smart devices in recent years has sparked a global chip supply crisis that in turn affects automotive, consumer electronics, PCs and numerous adjacent industries. Dozens of chips and sensors are already used in modern smartphones, smart appliances and personal computers, and the number (and complexity) of these chips will only increase. These parts use more advanced specialized nodes, which is one reason why companies such as TSMC must expand the capacity of otherwise "old" nodes to meet growing demand in the coming years.
But there's another market about to explode: smart cars. Cars already use hundreds of chips, and the semiconductor content of cars is growing. By some estimates, there will be around 1,500 chips per car in a few years — and someone has to make them. That's why TSMC's rivals GlobalFoundries and SMIC have been increasing their investments in new capacity over the past few years.
TSMC, which has the largest capex budget in the semiconductor industry (challenged only by Samsung), has been relatively quiet with its mature and specialized node production plans in recent years. But at the 2022 TSMC Technology Symposium, the company formally outlined its plans.
The company is investing in four new facilities for mature and professional nodes:
Located in the first phase of Fab 23 in Kumamoto, Japan, the semiconductor fabrication facility will manufacture chips using TSMC's N12, N16, N22 and N28 nodes and will have a monthly production capacity of up to 45,000 300mm wafers.
Fab 16 Phase 1B in Nanjing, China. TSMC currently manufactures its N28 chips in China, although there have been rumors that the new phase will be able to manufacture chips using more advanced nodes.
Increasing mature/specialized capacity by 50% over the next three years is a major shift for the company and will improve TSMC's competitive position in the market. Perhaps more importantly, the company's specialty nodes are largely based on its general-purpose nodes, allowing at least some companies to reuse IP they once developed for computing or RF for new applications.
"[Our] expertise is very unique because it's based on a common technology platform [logic technology platform], so our unique strategy is to have our customers share or reuse a lot of [common] IP," said TSMC VP of Business Development and Senior Engineer Kevin Zhang said. "For example, you have RF capability and you build RF on a common logic platform, but then you find 'hey, someone needs what's called a ULV capability to support IoT product applications.' You want to build it on a common platform so you can This allows different product lines to fully share IP, which is very important to our customers, so we really want to provide an integrated platform to meet the customer's market demand product perspective.
There are other advantages. For example, TSMC's N6RF allows chip designers to combine high-performance logic with RF, allowing them to build products such as modems and other more unique solutions. Many companies are already familiar with TSMC's N6 logical node, so now they have the opportunity to add RF connectivity to products that benefit from high performance.
GlobalFoundries has a similar approach, but since the US foundries have nothing to match TSMC's N6, TSMC has an indisputable advantage here.
