Soon after Taiwan Semiconductor Manufacturing Co. (TSMC) decided to double its 28nm capacity at its Nanjing fab, some within China’s semiconductor industry feared that the move was more strategically motivated to impede China’s semiconductor autonomy than purely commercially driven by the stringent demand for auto chips.
An investment of US$2.8 billion and expecting mass-production in 2023, it represents TSMC’s major investment in China since the foundation of its Nanjing fab in 2015.
TSMC aims for the choke point?
According to Xiang Li Gang, a veteran observer of China’s ICT industry, TSMC’s move was driven by a few reasons. Taiwan’s recent water and power shortages, combined with the low profitability to construct fabs in the US, have driven TSMC to expand capacity at Nanjing, where such supplies are relatively stable.
Most important of all, Xiang feared that TSMC’s expanded capacity at 28nm would directly compete with China’s homegrown foundries, especially Semiconductor Manufacturing International Corp. (SMIC), that seek to consolidate its competitive edge at the same technology node.
Earlier this March, SMIC also announced to expand its 28nm capacity and expect a monthly output of approximately 40 thousand units in 2022.
SMIC stuck at 28nm
To date, the 28nm node remains a watershed to cross for SMIC. At the height of Sino-American trade war, the Dutch company ASML, citing expired export license, delayed the shipment of EUV (Extreme Ultraviolet) lithography machine to SMIC, originally due for delivery in 2019. Without the key equipment, SMIC has been stalled at 28nm, unable to progress down the nodes to manufacture at 7nm.
Meanwhile, yield rate at the 14nm node remains incomparable with that of TSMC, despite the company repeatedly claiming otherwise, citing a yield rate jump from 3% to 95% – something that industry insiders disputed.
ASML already applied for a new license from the Dutch government, but progress has been hindered, owing to geopolitical tensions.
28nm – back in the spotlight
It is a peculiar time for TSMC to boost its 28nm capacity – the technology node is not only central to the latest auto chip crisis, but also cements TSMC’s impregnable status in the past decade.
Back in 2009, amid the global financial crisis, TSMC’s legendary founder, Morris Chang, made the crucial decision to raise capital expenditure and increased its 28nm capacity, after foreseeing the future demand of the smartphone market. The decision also helped TSMC to defeat its competitor GlobalFoundries around 2011, while becoming TSMC’s most profitable technology node.
Shortly before the chip shortage crisis and US sanctions on SMIC, however, oversupply in 28nm nodes actually dragged TSMC’s profit margins, with capacity only 80% full in the past two years.
As sanctions came amid trade tensions, followed by the pandemic, uncertainty prompted IC companies to transfer orders to non-Chinese foundries. Seizing this opportunity against their Chinese counterparts, TSMC, UMC, Samsung and other foundries subsequently expanded their 28nm capacities.
Against this backdrop, TSMC’s recent announcement to expand the fab in China indeed came as a surprise and raised some eyebrows.
Uncertain road ahead for China
Especially when China’s national semiconductor strategy focuses on achieving economy of scale at the 28nm node, so as to finance the improvement of the 14nm process and beyond – a strategy also pursued by some Chinese industry leaders. Recently, Hu Weiwu, the head of Chinese IC designer Loongson Technology, contended that China needed to consolidate the 28nm and 14nm nodes before looking further.
Last year, the Chinese government, in a bid to upgrade its IC industry, just introduced a ten-year tax holiday for foundries that have been in existence for 15 years and manufacturing at the 28nm and more advanced processes. SMIC would be a primary beneficiary.