Reports have emerged this week that SMIC, the largest foundry in China, is set to start mass production using its in-house developed 14 nm FinFET manufacturing technology in the first half of this year. Notably, this comes at least a couple of quarters earlier than was initially expected, indicating that SMIC is apparently ahead of schedule. Meanwhile the company is already working on its post-14nm processes, as development of its 10 nm and EUV-enabled 7 nm fabrication processes are currently underway.

Based on various reports from China and Taiwan technical media, SMIC’s yields at 14 nm have reached 95%, which is more than sufficient to start mass production. Consequently, the foundry is gearing up for volume production of a 14 nm smartphone SoC in the first half of 2019. While SMIC naturally does not disclose name of their first 14nm customer, the company’s key clients are HiSilicon, Qualcomm, and Fingerprint Cards (FPC, which produces fingerprint sensors), so it's a relatively short list of potential candidates.

Analysts say that SMIC’s 14 nm capacity will be relatively small when compared to the industry leaders, all of whom run multiple leading-edge fabs. SMIC currently has two fabs that can process 300 mm wafers using 28 nm and larger fabrication processes. The same fabs will be used for 14 nm projects too, but given their capacities and SMIC’s very high fab utilization rate (94.1% in Q2 2018), do not expect them to make loads of 14 nm SoCs. And for these reasons, along with prepping 14nm for its current fabs, the company is building a large $10 billion fab that will be used for its leading-edge manufacturing technologies in the future.

“SMIC is getting $10 billion to build capacity for 14nm, 10nm and 7nm. They will have capacity for 70,000 wafers a month by Q4 in 2021,” said Handel Jones, chief executive of International Business Strategies (IBS). “The building is huge. They have bought some equipment, but nothing significant yet.”

That said, do not expect SMIC to produce SoCs using leading-edge FinFET process technologies in quantities that are comparable to other makers of semiconductors in the foreseeable future. Even if the company can line up the capacity, lining up the demand could prove trickier. 14 nm chips are expensive to design and build the masks for, which is why so much chip volume is still at 28 nm and larger

Overview of SMIC's Fabs
  Process Technologies Capacity
Wafer Starts per Month
Location
BJ 200mm 90 nm - 150 nm 50,000 Beijing, China
300mm 28 nm - 65 nm 35,000
SH 200 mm 90 nm - 350 nm 120,000 Shanghai ,China
300 mm 28 nm - 65 nm 20,000
SZ 200 mm 90 nm - 350 nm 60,000 Shenzhen, China
TJ 200 mm 90 nm - 350 nm 50,000 Tianjin, China
LF 200 mm 90 nm - 180 nm 50,000 Avezzano, Italy

SMIC's latest progress slots in well with China's ambitious “Made in China 2025” plan. Under the plan, government planners want to achieve a 70% chip self-sufficiency by 2025, which having a leading-edge fab will help with. However there's doubt among analysts doubt that it is possible. Most of the ICs produced in China by 2025 will be made by companies based outside of the country.

Past 14nm, SMIC is already at work on its 10nm and 7nm processes as well, as previously confirmed by the company in 2018. Both processes are extremely costly to design, but since the semiconductor industry is growing in general and because of generous funding from the Chinese government (and various affiliated parties), SMIC has enough money for the necessary R&D. Working towards that goal, last year SMIC acquired an EUV step-and-scan system from ASML for $120 million, which is expected to be delivered early in 2019 for use in 7 nm process development and eventually mass production.

Related Reading:

Sources: CN Beta, China Securities, DigiTimes, EETimes, IC Insights, InstantFlashNews, SemiEngineering, TechNews

POST A COMMENT

20 Comments

View All Comments

  • Gondalf - Monday, February 11, 2019 - link

    Smic want to torpedo Tsmc. Intel is not a foundry and never will be. Reply
  • ABR - Saturday, February 09, 2019 - link

    "SMIC is getting $10 billion to build capacity..." - where from? Gov't grant? Loan? That is a chunk of change. Reply
  • PeachNCream - Saturday, February 09, 2019 - link

    Most likely from Alphabet/Google. Reply
  • FullmetalTitan - Saturday, February 09, 2019 - link

    I'm gonna call bullshit on "in house developed", and definitely on hitting 95% yield before starting mass production of their first 14nm FinFET design.

    Good money says we will see a design that is very familiar, and that 95% is either the defect or parametric limited yield, but not the closed yield, unless they have absurdly lax process margin windows.
    Reply
  • name99 - Sunday, February 10, 2019 - link

    "Under the plan, government planners want to achieve a 70% chip self-sufficiency by 2025, which having a leading-edge fab will help with. However there's doubt among analysts doubt that it is possible. "

    That's the difference between China and the US in one sentence:
    China reckons it's worth trying for a stretch goal, and that the results will be valuable even if the precise goal is not achieved (if they hit eg 60% rather than 70%). Meanwhile the "analysts" think it's not worth trying anything that might be difficult, and that might possibly fail.

    One of these attitudes is likely to lead to better longterm outcomes than the other...
    Reply
  • zepi - Sunday, February 10, 2019 - link

    In ”communist” China government makes bets with taxpayer money. Maybe they get it right and SMCI benefits. Maybe they end up paying a fortune and failing miserably with 10nm or 7nm. Or arriving late to market unable to compete. They might waste billions (which is common for all countries and governmental projects) and get nothing in return.

    Personally I don’t think it is right that governments makes these bets with taxpayer money. Give the money to the poor (or don’t tax it in the first place) and let people themselves decide if they rather invest in the schooling of their kids, healthcare or if they want to gamble in semiconductor market.
    Reply
  • HollyDOL - Monday, February 11, 2019 - link

    Looking on how high demand is for fabs, it seems they could have gone much worse with their money, unless some very big game changer comes they at least won't be on a losing side.

    But you are right ofc, investment that would increase the knowledge level of their population would have much higher impact long term.

    Note I intentionally say 'knowledge level'... paying money to print more uni diploma without actually increasing skills of your people would be wasting those funds even more than this 'bet'.
    Reply
  • Spunjji - Tuesday, February 12, 2019 - link

    Making bets with other people's money is a cornerstone of capitalist society, too - it's just done by corporations that are effectively run as small dictatorships / oligarchies, and instead of "tax" they use profits garnered from things that people can't really avoid buying and/or have no real competitive choice in where they buy it from.

    The point I'm making isn't "capitalism bad", though - it's just that you're criticising something (making bets with other people's money) and attributing it to one type of economy, when it's a nontrivial problem that we have in all existing economies.

    Ideally we'd build economies and social systems where we can all have good schools, good healthcare and solid development in semiconductor technology.
    Reply
  • jjj - Monday, February 11, 2019 - link

    This is likely about as fake news as it can get.
    Even if they start in H1, it's gonna be risk production.
    Reply
  • Spunjji - Tuesday, February 12, 2019 - link

    Source? Reply

Log in

Don't have an account? Sign up now