单双哈希( - China has painted itself into a chip corner



Chip worker shws a 300mm silicon wafer at a German semiconductor plant

AS Washington embarks on a multi-billion dollar, decade-long semiconductor development campaign, Beijing is reckoning with its own 20-year effort that’s largely failed to deliver.

Both will need to grapple with wasted funds and misguided goals as they play catch-up to Taiwan and South Korea.

Architects of China’s ambitious efforts may be facing the music for having not produced world-beating technology, Bloomberg News reported this week.

Multiple corruption probes announced by authorities stem from anger among the nation’s top leaders over an inability to develop semiconductors that could replace American components, it reported.

Two of the most scrutinised areas are the US$9bil (RM40bil) bailout of Tsinghua Unigroup Co, and the National Integrated Circuit Industry Investment Fund – known as the Big Fund.

For all intents and purposes, China has failed to achieve its semiconductor goals, and those tasked with realising them are being brought to account.

Beijing won’t be smarting at the loss of money – it’s been willing to burn cash – but at the lack of progress such expenditure was supposed to buy. Those looking at China’s achievements are mostly finding what they seek, and ignoring the rest.

Semiconductor Manufacturing International Co (SMIC), for example, got a lot of attention recently when industry analysts TechInsights wrote: “SMIC has been able to fabricate features that are small enough to be considered 7nm.”




That “nm” figure refers to nanometers, a metric for the size of connections within a chip (smaller is better), and these days is as much a marketing term as a scientific one.

China chip cheerleaders see this as an incredible breakthrough, bringing the Shanghai-based firm closer to the capabilities of world leaders Taiwan Semiconductor Manufacturing Co and Samsung Electronics Co.

But it’s not, really.

What SMIC appears to have done is produce a somewhat standard chip, used for bitcoin mining, and with little evidence it can churn these out with good yields or at scale.

Any budding chef trying to make souffle will eventually pull off a couple of good samples. But mastery can only be claimed when almost all attempts are successful (yield), and can be done consistently and in large quantities (scale).

By contrast, TSMC and Samsung rolled out mass production at 7nm four years ago – a demonstration that they’d nailed the process.

The question of how SMIC managed this, while being cut off from Western manufacturing equipment, also remains.

As TechInsights notes, it’s not impossible to make 7nm using older machinery, just a lot more complicated. And it’s an established fact that TSMC has in the past sued SMIC for stealing technology. The company has not been accused of wrongdoing with respect to 7nm processes.