Globalfoundries Investing $1.4 billion in Fabrication Capacity Expansion, Anticipate IPO

[Raevenlord]

Globalfoundries has announced they are in the early stages of a massive $1.4 billion investment in their manufacturing capabilities, which aims to increase overall production of semiconductor chips. One third of this development investment will be pre-paid by Globalfoundries' customers, who by investing this way, are also pre-allocating Globalfoundries' future supply - it's a way for the company to receive funds for not-yet-produced wafers, enabling it to proceed with these expansion plans. The company usually reserves $700 million yearly for capacity expansions and technology improvements, so this $1.4 billion figure essentially doubles that.

That doubling comes at a time where existing capacity throughout the semiconductor industry is showing not to be sufficient for global demand. The plans will see Globalfoundries increase its wafer production capability by 13% this year, and 20% in 2022. The increased funding for developmental expenses will be allocated equally between the company's three manufacturing plants already installed in Dresden, Malta (New York) and Singapore. Globalfoundries' 2020 revenue ended up at $5.7 billion, a cutback from 2017's $6.176 billion. The company, however, projects its revenue to increase 9-10% in 2021 due to the current unprecedented demand for its fabrication technologies. The company is also looking to capitalize on this demand in another way: by bringing its IPO forward. Where before the company planned to go public in 2022 or even 2023, the increased current demand and prospective YoY growth places the company in a good place for such a move.



View at TechPowerUp Main Site

 

[henok.gk]

Doubt they'd cook anything competitive to TSMC but the semiconductor business could use everyones help at this point.

 

[TumbleGeorge]

Something new from GF for next their nanometers?

 

[W1zzard]

Trying to ride the silicon shortage hype to make $$ with an IPO?

 

[Spartoz]

GF still exists? I tought they where dead

 

[Frick]

Doubt they'd cook anything competitive to TSMC but the semiconductor business could use everyones help at this point.

GF still exists? I tought they where dead

They just stepped out of the bleeding edge game, which kinda made sense IMO. They were already lagging behind in that regard, and focusing on refining their 12/14nm process is not bad. Plenty of business to be had even if they use older nodes. Not everything makes sense on bleeding edge processes.

 

[Wirko]

GF still has an alliance with Samsung (and IBM, too?). They might be planning to introduce the 7nm process at a later time, in 2-3 years, with Samsung's help and by buying their used machines. An IPO would be a sensible choice, GF does not have many options to raise several billion $ for that.

 

[Raevenlord]

Trying to ride the silicon shortage hype to make $$ with an IPO?

I'd say so.

 

[xSneak]

How does their business compare to united microelectronics corporation in Taiwan?


Is the IPO just ownership trying to cash out, or will it provide funding for factory and node upgrades?

 

[nfineon]

Since they got taken over by money interests from the middle east, the owners didn't want to invest any more into leading edge tech and so they continue to fall way behind only getting small contracts on much older nodes. They were already on a downward trajectory and will be even moreso as time moves on because getting on the 7nm process is actually CHEAPER than using the old 14nm tech, each wafer is more expensive yes but you get many many more die per wafer on the latest nodes so it's the best of both worlds scenario for customers which is why everyone and their mom is trying to get on TSMC/Samsung. They are basically trying to IPO and cash out asap given they aren't a real player in the marketplace except when TSMC/SIMC/Samsung just don't have any fab capacity. Sad really, they could be in the hot money seat at the moment if they would have leased Samsung's 8nm nodes but the investors just want their money back without spending any more pennies on massive upgrades. They would have had customers banging down their doors right now but their desire for immediate profits prevented them from making the logical decision.

 

[Raevenlord]

Since they got taken over by money interests from the middle east, the owners didn't want to invest any more into leading edge tech and so they continue to fall way behind only getting small contracts on much older nodes. They were already on a downward trajectory and will be even moreso as time moves on because getting on the 7nm process is actually CHEAPER than using the old 14nm tech, each wafer is more expensive yes but you get many many more die per wafer on the latest nodes so it's the best of both worlds scenario for customers which is why everyone and their mom is trying to get on TSMC/Samsung. They are basically trying to IPO and cash out asap given they aren't a real player in the marketplace except when TSMC/SIMC/Samsung just don't have any fab capacity. Sad really, they could be in the hot money seat at the moment if they would have leased Samsung's 8nm nodes but the investors just want their money back without spending any more pennies on massive upgrades. They would have had customers banging down their doors right now but their desire for immediate profits prevented them from making the logical decision.

I think there are some fundamental incorrections on what you stated, particularly on it being cheaper to manufacture or acquire products on the 7 nm node.

 

[dragontamer5788]

14nm is still being used in automotive and other industries. GF probably got a bright future: not as a leading-edge technology, but as a maker of "mundane" chips: like power transistors or older automotive grade (or space-grade) designs.

Note: the space-grade Perseverance rover that landed on Mars a few weeks ago uses space-grade RAD750, built from a 150nm node.

RAD750 - Wikipedia

en.wikipedia.org

There's PLENTY of business at the lower-tech side of things. Just because GloFo doesn't have any timeline for 7nm or lower doesn't mean that they're dead. It just means that they're probably picking up designs like the RAD750.

 

[Chrispy_]

Does anyone know what GF's roadmap is for processes newer than 2019's 12LP+?

More capacity of 12LP+ is obviously great for everyone right now, since it's still a high-end, relatively modern node - but we cannot afford for GF to give up on competing with Samsung, Intel, and TSMC. Intel doesn't share which would matter if Intel could make anything 10nm right now, and the Samsung TSMC duopoly isn't working out so great at the moment....

 

[dragontamer5788]

Does anyone know what GF's roadmap is for processes newer than 2019's 12LP+?

More capacity of 12LP+ is obviously great for everyone right now, since it's still a high-end, relatively modern node - but we cannot afford for GF to give up on competing with Samsung, Intel, and TSMC. Intel doesn't share which would matter if Intel could make anything 10nm right now, and the Samsung TSMC duopoly isn't working out so great at the moment....

Given GF's corporate situation, I find it hard to believe that they'll raise money for a more advanced node any time soon.

GF needs more money. Maybe if they IPO successfully, they will have the money they need to start the R&D process once more. But short term / near term, it doesn't seem like them competing vs TSMC / Samsung is going to lead to profits of any kind.

Current plans are for GF to IPO in 2022 or so. Giant investments (like supporting a newer advanced node) would probably wait until then...

 

[RandallFlagg]

Honestly I think just increasing 12/14nm capacity at GloFlo would drive benefits in other areas.

It's worth noting that their 12/14nm is significantly better \ more dense than all of the other so-called 12/14/16nm nodes, except Intel 14nm, where they are close to parity. It's a perfectly suitable node for all kinds of SoCs, GPUs, chipsets and so on.

To put this in comparison, 20x0 Nvidia cards were made on TSMC 12nm with a density of 24.7MT/mm2. GloFlo's 12LP is about 36MT/mm2, or about 50% more dense.

 

[Chrispy_]

Honestly I think just increasing 12/14nm capacity at GloFlo would drive benefits in other areas.

It's worth noting that their 12/14nm is significantly better \ more dense than all of the other so-called 12/14/16nm nodes, except Intel 14nm, where they are close to parity. It's a perfectly suitable node for all kinds of SoCs, GPUs, chipsets and so on.

To put this in comparison, 20x0 Nvidia cards were made on TSMC 12nm with a density of 24.7MT/mm2. GloFlo's 12LP is about 36MT/mm2, or about 50% more dense.

I don't think 12LP clocks very high though.

I know that both Vega and the RX590 were both tuned to run on the 12LP+ process and the chip design tweaked specifically to get higher clocks, both with disappointing results.

GF's 12LP+ basically craps out beyond 1.5GHz no matter how much the design is tuned for it.

Like you say though, there are plenty of mainstream products outside of flagships that could certainly be made to work extremely well on 12LP+