UMC Investing $3.6 billion on 28 nm Manufacturing Capabilities Amidst Worldwide Semiconductor Shortages

[Raevenlord]

UMC has announced plans to invest $3.6 billion in increasing output from its 28 nm manufacturing facilities. This move comes amidst a global semiconductor shortage, and isn't the first time a semiconductor manufacturer "dust off" their older manufacturing processes as a way to remove pressure from more modern silicon manufacturing capabilities. In this case, UMC will be increasing manufacturing output from its 300 mm Fab 12A facility in Tainan, Taiwan.

UMC has entered agreements with some of its clients, who will be paying upfront for expected chip rollout in the future. In exchange, clients will get the benefits of preset pricing (thus avoiding any potential increases arising from increased demand or general price fluctuation), as well as UMC's assurance of certain manufacturing volume allocation towards their needs. Fab 12A currently manufactures 90,000 300 mm wafers per month (wpm). An additional 10,000 wpm is being installed this year and phase six will add another 27,500 wpm to the mix. The mature 28 nm tools will be installed in floors that already feature support for future tooling upgrades to 14 nm. UMC expects to hire around 1,000 additional employees as part of this expansion effort.



View at TechPowerUp Main Site

 

[PerfectWave]

So we can buy back 28nm GPU Kappa

 

[xChoice]

28nm? can I get a new Nexus 5 then? (but with a bigger battery plz )

 

[Chaitanya]

So is this going to help automotive and other industrial sectors to alliviate the shortage of chips?

 

[xkm1948]

This can feed the auto industry so good.

 

[TheLostSwede]

So is this going to help automotive and other industrial sectors to alliviate the shortage of chips?

Most likely, yes, but it's two years out at best.

 

[mechtech]

Toss a coin to your foundry.......................

 

[TumbleGeorge]

From many years was written many articles that some manufacturers or trademarks has plans to invest certain amount billions in that and what...Where is proving that realize of this plans and with cost of this number of money which is declare?

 

[Frick]

From many years was written many articles that some manufacturers or trademarks has plans to invest certain amount billions in that and what...Where is proving that realize of this plans and with cost of this number of money which is declare?

As a rule lying to investors is known as "fraud".

 

[Wirko]

Most likely, yes, but it's two years out at best.

The building is already there. The machines need not be the most advanced ones, and UMC can probably buy some of them on the used market. Can the process of buying and installing everything take that long?
Also, it's not clear whether the current manufacturing can run uninterrupted while the new equipment is being installed or not. If not ... ouch.

 

[Tomgang]

Oh cool. Soon I can buy a gtx 660 ti again...

 

[Caring1]

Phhttt, funny comments here, most likely Intel ramping up CPU production, next they'll be telling us 28 nm is the CPU we need

 

[Tom Yum]

Phhttt, funny comments here, most likely Intel ramping up CPU production, next they'll be telling us 28 nm is the CPU we need

Intel: Meet Nuclear Lake, manufactured on our advanced and all new* Super Ultra Enhanced FinFET 28nm++ process, extreme SuperPI performance at only 115W TDP**

* Licensed from UMC
** 500W PL2.

 

[TheLostSwede]

The building is already there. The machines need not be the most advanced ones, and UMC can probably buy some of them on the used market. Can the process of buying and installing everything take that long?
Also, it's not clear whether the current manufacturing can run uninterrupted while the new equipment is being installed or not. If not ... ouch.

Q2 2023 according to the local news.

Taiwan chipmaker UMC to spend $3.6 billion to expand capacity | Taiwan News | Apr. 29, 2021 12:00

www.taiwannews.com.tw

 

[Dave65]

Going backwards is never a good thing.

 

[lexluthermiester]

Going backwards is never a good thing.

Rubbish. This move makes a lot of sense. Nothing needs 5nm/7nm/1nm/14nm to run perfectly.

 

[watzupken]

Its an odd move to invest in a very matured node. Instead of investing forward, I feel UMC is investing backward and will likely be throwaway money/ effort considering its 28nm that we are looking at here. While chip demand is very high now, it will eventually reach a point that demand for all these 28nm chips will stop/ go back to normal. Even now, besides budget SOCs and chips using 28nm, it is unlikely to attract more demand. Backporting from a more advance node is unlikely here as well since it will take time and $$$, often with less desirable results.

 

[lexluthermiester]

Its an odd move to invest in a very matured node. Instead of investing forward, I feel UMC is investing backward and will likely be throwaway money/ effort considering its 28nm that we are looking at here.

I prefer to look at it as an investment in something that works reusing existing technology and equipment. Again, nothing needs ultra small lith tech to work. 28nm will work perfectly for a great deal of use case scenario's. Even 32nm and 45nm would be solid for a lot of things. We don't need bleeding edge tech for things that are not bleeding edge.

 

[MIRTAZAPINE]

I prefer to look at it as an investment in something that works reusing existing technology and equipment. Again, nothing needs ultra small lith tech to work. 28nm will work perfectly for a great deal of use case scenario's. Even 32nm and 45nm would be solid for a lot of things. We don't need bleeding edge tech for things that are not bleeding edge.

True not everything need the bleeding edge. Hell 28nm would be good replacing 90nm tech that is still in use for devices that does require that advance processes for example your washing machine, toaster, fridge, automobile and industrial control systems etc.

 

[Wirko]

It's not just about manufacturing. Chip design costs have run out of control at finer nodes.

 

[user556]

It's not just about manufacturing. Chip design costs have run out of control at finer nodes.

That's a good read. It explains specifically why 28 nm. That's the smallest planar gate (pre-FinFET) technology. Explaining that FinFET has a distinct problem of not able to provide definable gate width.

 

[Why_Me]

Intel To Invest $3.5 Billion In New Mexico Fab Upgrade for Foveros Tech (Updated)

Intel's spending spree continues

www.tomshardware.com

 

[watzupken]

I prefer to look at it as an investment in something that works reusing existing technology and equipment. Again, nothing needs ultra small lith tech to work. 28nm will work perfectly for a great deal of use case scenario's. Even 32nm and 45nm would be solid for a lot of things. We don't need bleeding edge tech for things that are not bleeding edge.

I don't disagree with what you mentioned, but the problem is that the demand and money are going towards the finer/advnace nodes. Chips used in cars are still on more matured nodes, but the questions are for how long, and how profitable they are in the longer run. Granted the amount that UMC is investing into 28nm fab isn't great as compared to the likes of TSMC, Samsung and Intel, so I guess they would have projected good profit to cover the cost of investment.

 

[silentbogo]

Granted the amount that UMC is investing into 28nm fab isn't great as compared to the likes of TSMC, Samsung and Intel, so I guess they would have projected good profit to cover the cost of investment.

Most of it is pretty much "guaranteed orders".
Right before this announcement they bought lots of fab equipment from Samsung in order to manufacture CMOS sensors for Samsung.
I'm sure it's not just "projections" that drive this expansion. They are most likely booked to the brim already. 28nm is still widely used everywhere: not just automotive, industrial, embedded/IoT etc., but also parts for consumer electronics which you never think about, like OLED & LCD drivers, touch screen controllers, PMICs, SuperI/O chips, memory, flash storage and lots of other stuff.

 

[Wirko]

28nm is still widely used everywhere: not just automotive, industrial, embedded/IoT etc., but also parts for consumer electronics which you never think about, like OLED & LCD drivers, touch screen controllers, PMICs, SuperI/O chips, memory, flash storage and lots of other stuff.

Here's another way to look at that: according to Wikipedia, 32 and 28 nm processes were used to manufacture chips with 1 to 7 billion transistors, with few exceptions. The number gradually rose over time. I'd take this as a rough guide even several years later. If you have a design with up to 1B or 2B transistors, 28 nm is probably too costly; if you have something over 7B or 10B, you seek a finer node.