Question [WSJ] Intel in talks to buy GloFo

[jpiniero]

Wat. Says deal would be about $30B if the deal goes through.

WSJ News Exclusive | Intel Is in Talks to Buy GlobalFoundries for About $30 Billion

A deal for chip maker GlobalFoundries would be Intel’s largest acquisition. It comes as Intel is launching a major push to become a chip manufacturer for other companies.

www.wsj.com

 

[Roland00Address]

Interestingly, about half of TSMC's revenue is from legacy process nodes (16/20nm and older). This funds the development of the leading edge processes. The legacy nodes also have the majority of customers. Intel may be looking to build out this capability to make Intel Foundry Services a viable competitor.

View attachment 47421

HPC at the Heart of TSMC’s Growth

Just as in so many other areas of the world economy where real growth is being driven by those who can afford the bleeding edge, chipmakers are seeing the

www.nextplatform.com

Nods Intel used to use older foundries for things like chipsets, thunderbolt controllers, and other chips that did not need to be the latest geometries. I am wondering if that stuff is less needed with trying to move everything on die, and if it can't be moved on die use packaging to use older geometries.

Fabs need to be at full capacity for maximum profitability (likewise we want some spare capacity so we do not have shortages problems such as cars something that is messing up the inflation indexes right now for there are less cars and thus used car prices is so much larger in the basket of goods compared to previous years.)

 

[Doug S]

Fabs need to be at full capacity for maximum profitability (likewise we want some spare capacity so we do not have shortages problems such as cars something that is messing up the inflation indexes right now for there are less cars and thus used car prices is so much larger in the basket of goods compared to previous years.)

Spare capacity is something you want to see for the overall economy to be able to absorb increases in demand, but it isn't in the best interest of individual foundries as they'd rather sell all their capacity and run at 100%. Just another example of tragedy of the commons.

 

[blckgrffn]

Spare capacity is something you want to see for the overall economy to be able to absorb increases in demand, but it isn't in the best interest of individual foundries as they'd rather sell all their capacity and run at 100%. Just another example of tragedy of the commons.

Or natural disasters. Or trade wars. Or, I guess, pandemics (is that a natural disaster? ha). Or fires in subcomponent facilities.

It takes some special foresight to add capacity to the system that allows for shocks to be absorbed. It's too easy to just to take up all the slack with any extra capacity and leverage it for more money.

And even with excess capacity, with some chaotic system modeling (like highways) with enough load there will always be snarls and momentary bottlenecks. ¯\_(ツ)_/¯

 

[Doug S]

Or natural disasters. Or trade wars. Or, I guess, pandemics (is that a natural disaster? ha). Or fires in subcomponent facilities.

It takes some special foresight to add capacity to the system that allows for shocks to be absorbed. It's too easy to just to take up all the slack with any extra capacity and leverage it for more money.

And even with excess capacity, with some chaotic system modeling (like highways) with enough load there will always be snarls and momentary bottlenecks. ¯\_(ツ)_/¯

Yes like I said it is beneficial overall, but it is not beneficial for individual companies to maintain slack capacity. If you have that "foresight" and there is no shock, then you might lose your job as CEO if you have a half empty fab and lose money by the truckload because you spent billions expanding capacity that went unused.

This is why we have seen the DRAM market go through numerous feast and famine cycles over the years. Too much capacity lowers prices, so companies push back investments in additional capacity that don't make sense at current prices. As demand increases (or supply lowers due to a fire, earthquake, etc. taking out some productive capacity) that slack capacity goes away, prices increase, and everyone decides all at once to invest in more capacity, which once it all comes online gets them back to step 1.

The only way to provide slack capacity would be outside government intervention in the market, which would be fought mightily by the investors in the companies affected (as well as investors in other companies who fear they may be next for legislated slack capacity)

This is simply something capitalism cannot properly handle, at least not in markets like chips where it takes a year or two and costs billions to bring additional capacity online.

 

[Mopetar]

It sounds like market forces have handled the industry fine over the years, so it's silly to think they won't going forward. There may be points where demand is high and it seems like no one is in a great hurry to satisfy it, but if there's even higher demand for something else, investment will seek to fill that first.

Maybe it seems non ideal to anyone who really wants a GPU right now, but any investment into building more of those is money not invested into something else. To someone who cares nothing for a desktop GPU that just looks like wasted money.

 

[Thibsie]

Mmm if x% of the capacity is kept as spare and subsidized for security (or whatever reasons of the same kind) reasons, I don't see the problem. Of course this means government money and some will be up in arms but well, there aren't huge number of solutions.

 

[Roland00Address]

Spare capacity is something you want to see for the overall economy to be able to absorb increases in demand, but it isn't in the best interest of individual foundries as they'd rather sell all their capacity and run at 100%. Just another example of tragedy of the commons.

There are multiple ways we can align incentives such as "The Government" saying they will buy Wafers at a low price, to guarantee a minimum demand, but they are a customer of last resort* for X years, where other "Private" companies can buy the wafers at a higher price and they get priority.

Of course in such a situation "The Government" can get expeditated service for things like military by paying more. Likewise we as a society can actually build public foundries much like we have built public hospitals, public schools, etc. There are many solutions to "this problem," but hey this is not a new thing we been debating this stuff for at least 170 years ( actually far longer, but I am thinking of a specific Thomas Carlyle's phrasing of "this problem." )

It sounds like market forces have handled the industry fine over the years, so it's silly to think they won't going forward.

This is highly debated. I have an opinion on the matter but for simplicity sake I am just going to point to graphs.

Supplying Demand: The Chip Shortage in Macro Context

By Skanda Amarnath and Alex Williams Many thanks to Hassan Khan for helping with this piece

employamerica.medium.com

I am pointing at the graphs for you can argue if "the facts" are good or bad, but the boom and bust cycle was on one trajectory and then suddenly changed for almost 20 years after the the stock market crash and inconsistent demand and this in turn produce inconsistent supply leading to boom and bust. The ram market and the other market for silicon does not like this uncertainty, it is wasteful, and it does not lead to lower prices.

Maybe it seems non ideal to anyone who really wants a GPU right now, but any investment into building more of those is money not invested into something else. To someone who cares nothing for a desktop GPU that just looks like wasted money.

This last bit is flat out untrue, yet it feels like a truism. Humans deal with "waste" all the time, I may grow, buy, and plan to eat the strawberries but not eat them fast enough before the mold gets to them. Likewise we have also another form of waste where we never invested in the first place due to uncertainty. Both of them are not "excellent" they are an excess of doing too much and an excess of doing too little. No one can predict the future, not even the market.

The market is a reactive force, not a proactive one, and it is very good at being reactive for certain types of problems but not others. (and that is okay, markets can still be good and useful even if they have limits.) It is just important to know the limits the pros and cons of the situation.

 

[coercitiv]

@Vattila just one problem with all these takes on the possible acquisition of GF, they don't talk much about splitting up design and manufacturing. In fact, becoming a supplier for the U.S. Department of Defense may provide Intel with the wildcard they need to evade such a requirement from US government. I think all options are still on the table, and Intel will try very hard to have their cake and eat it too.

whoops...

Intel Foundry Services Drafted By US Department of Defense For Next-Gen Fab Needs

In short, the DoD wants to ensure it will be able to have its chips (and other necessary commercial hardware) fabbed within the United States on a leading-edge commercial manufacturing node, and it is tapping a consortium of companies lead by Intel to develop the necessary foundry ecosystem.

 

[moinmoin]

"a consortium of companies lead by Intel to develop the necessary foundry ecosystem"
🤔
"Along with Intel, the consortium also includes IBM, Cadence, Synopsys and other companies, all of whom will be providing their relevant expertise and technologies to the project."

And others? OK. Honestly think AT's summary is not really good there, the linked page is somewhat more specific, but doesn't spell out who exactly were awarded the programs aside Qualcomm and Intel:

"The Department of Defense (DoD) currently has no on-shore access to foundry technology capable of meeting the long-term leading edge (<7nm Digital Complementary Metal-Oxide Semiconductor (CMOS) node) microelectronics fabrication needs for DoD specific designs or commercial-off-the-shelf (COTS) components."

"There is currently no commercially viable leading edge pure-play foundry option on-shore that will keep the U.S. industry on the leading edge of microelectronic integrated circuit fabrication needs."

Rapid Assured Microelectronics Prototypes - Commercial (RAMP-C) - NSTXL %

S²MARTS Project No. 21-07 This opportunity has been awarded to Qualcomm Technologies and Intel Technologies

nstxl.org

 

[scineram]

Well the TSMC N5 fab would fit that role.

 

[Ajay]

Well the TSMC N5 fab would fit that role.

True, but Gelsinger has done an excellent job of creating a ‘consortium' of Semicon businesses and used their joint economic impacts to get Congress critters and the administration involved in increasing US Semicon capabilities. This is becoming large domestic initiative that more and more in politics are getting on board with. None of them wants to look like they are opposed to strengthening US domestic manufacturing. Plus, there are lots of $$s that will go into many congressional districts.

With the military making this a defacto national security issue - TSMC will be at the back of the bus.

Like it or not, this is what is happening. This is often how the US goes about saving or promoting high technology infrastructure.

 

[Doug S]

TSMC's N5 fab in the US would still be used to fab a lot of defense related stuff. While plenty of parts defense contractors use are custom made and use the Trusted Foundry program, there are tons of "off the shelf" parts. Procurement at the Pentagon requires some percentage of parts (or rather their value) be "made in the USA" so having a US based TSMC fab means many of those ancillary chips not covered under Trusted Foundry can be acquired there to help meet the "made in the USA" content requirement.

 

[Roland00Address]

Like it or not, this is what is happening. This is often how the US goes about saving or promoting high technology infrastructure.

1) This happens all the time in other industries besides silicon production,
2) This happens not just in the US but in many other countries.

We do not think of #2 for the US is often in the leading edge or close to it, and thus we assume myths that are not true. While it was not a government intervention the reason Microsoft is MS with DOS and later Windows is that IBM insisted on having at least two suppliers of chips, and Columbia Data Products and later Compaq reversed engineered and created a similar but legally different BIOS and thus you could not do a hardware lockdown and thus Microsoft was the most innovate layer that collected 20 to 30% of the profits of the entire software industry. The government wants a consortium much like IBM wanted at least two players with silicon. It makes the military less "brittle" to not have a single part supplier.

And the logic applies to the greater market as a whole, the whole industry both hardware, software, and people who use those services are better off if we have at least 2 leading fabs, and right now it is only TSMC (with the lead), and Samsung plus Intel (who are slightly behind.)

 

[Schmide]

With the military making this a defacto national security issue - TSMC will be at the back of the bus.

Not completely. Military often cares about multi-sourced procurement such that mission critical parts can have redundancy.

(and I started typing this and the above post came up. Just read that)

 

[DrMrLordX]

With the military making this a defacto national security issue - TSMC will be at the back of the bus.

It's a shame. It would have been a heck of a foreign policy coup to entreat TSMC to move more of their operations overseas as a hedge against regional aggression. Instead Congress is shoveling money on a foundry that is more of a floundry.

 

[Ajay]

It's a shame. It would have been a heck of a foreign policy coup to entreat TSMC to move more of their operations overseas as a hedge against regional aggression. Instead Congress is shoveling money on a foundry that is more of a floundry.

Yes, but I think some team Intel has much more lobbying power. Maybe the final outcome will be different. There was some language about the need for a pure play foundry - so I hope, at the very least, that Intel splits off (though probably still owns) it’s fabs into a separate division with it’s own C-suite etc.

 

[Roland00Address]

Yes, but I think some team Intel has much more lobbying power. Maybe the final outcome will be different. There was some language about the need for a pure play foundry - so I hope, at the very least, that Intel splits off (though probably still owns) it’s fabs into a separate division with it’s own C-suite etc.

This was announced in March of 2021, on Day 35 of the new CEO being on the job…

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

www.anandtech.com

and it is also one of the rules Congress and the Pentagon is insisting on, here is anandtech from two days ago.

AnandTech Forums: Technology, Hardware, Software, and Deals

Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

www.anandtech.com

The 2nd link was always in the know where government contracts would have insisted on separate accounting identities for the play foundry and the rest of intel’s business. Governments including the US do not just hand out money they often do action with said contracts to shape the markets in ways they think are more pro-competitive for the greater economy as a whole. If Intel wanted subsidies or a very profitable contract said rules would also be designed to make this a less of bottleneck in the future. Especially since Intel is having a chip shortage announced in May where they can not get some of the raw substrates slowing down their production, likewise some of their partners can not finish their computers even if Intel is meeting delivery for other components such as WiFi chips and other things on the motherboard are also having shortages.
————

Every since the Truman Era we have various agencies such as the CEA to identify economic bottlenecks, to create supply redundancy and stockpiles. Note this is inherently political and we usually see small actions and not big actions where there is a large incumbent that is domestic, and lots of small incumbents that are domestic for large company or small company complains you are picking sides even if it helps the remaining 95% (or whatever number) part of the economy. Steel may love it one moment all the investment, and hate it when there is a steel stockpile which makes it harder to raise prices which increases their profits but causes inflation down the component ladder. Silicon is the new steel or oil or whatever resource you want to call it.

But when the US industry is behind we get less complaints from the incumbents. It is the way of the world for this similar logic also applies to other countries as well. The US is just one of the larger economy hubs that is all.

 

[Doug S]

It's a shame. It would have been a heck of a foreign policy coup to entreat TSMC to move more of their operations overseas as a hedge against regional aggression. Instead Congress is shoveling money on a foundry that is more of a floundry.

Taiwan would not sit still and allow one of their biggest (if not THE biggest) company on their soil to move major portions of their operations overseas, any more than we'd sit still if another country was trying to convince Apple or Microsoft or Amazon to do so.

 

[DrMrLordX]

any more than we'd sit still if another country was trying to convince Apple or Microsoft or Amazon to do so.

. . . seriously? Apple has virtually no manufacturing in the United States. We've had expatriation of corporate assets for decades. Trick would be to get TSMC to build more fabs overseas, that's it. Not to get them to close their existing fabs in Taiwan (that would be foolish).

 

[Ajay]

This was announced in March of 2021, on Day 35 of the new CEO being on the job…

That’s not how I think they should be setup. All fabs should be part of IFS, except , perhaps, for a prototype fab (so Intel can stay on the cutting edge in x86). It should be a separate company, with it’s own executives. The new company could offer new shares to generate some funds an have something to offer execs, etc. Intel would own a super majority of the stock, perhaps by corporate bylaws. Something like that would be the best investment I’d public dollars and present a new face to other semiconductor design customers. Then IFS would truly be a US based pure play Fab.

Just my opinion.

 

[NostaSeronx]

Since, GlobalFoundries is in the thread.

Starting October 2021, 12LP&12LP+ will start the process of being removed from Malta. The products that are replacing FF will be 90nm Skywater/MIT FDSOI, 45nm GF FDSOI, 28nm GF Bulk(ULP+?), and 22nm GF FDX. There might be plans for getting 12FDX out faster at Malta, but not sure. Marketing going forward is advised to focus on FDSOI/Bulk and not on FF at Malta.

 

[Doug S]

. . . seriously? Apple has virtually no manufacturing in the United States. We've had expatriation of corporate assets for decades. Trick would be to get TSMC to build more fabs overseas, that's it. Not to get them to close their existing fabs in Taiwan (that would be foolish).

Manufacturing chips is WAY different than "manfacturing" (assembly) of phones. One requires a lot of people with advanced degrees, tons of support people (for all the super complex machines, environmental systems, etc.) with advanced degrees, lots of high value employees.

The other requires people with minimal skill and no education, lots of low value employees. Low enough value that if Apple was somehow forced to bring iPhone assembly to the US, it would be via a robotic assembly line that would employ a very small number of high value employees.

The vast majority of Apple's high value employees are based in the US. "Designed in California, assembled in China".

 

[Roland00Address]

That’s not how I think they should be setup. All fabs should be part of IFS, except , perhaps, for a prototype fab (so Intel can stay on the cutting edge in x86). It should be a separate company, with it’s own executives. The new company could offer new shares to generate some funds an have something to offer execs, etc. Intel would own a super majority of the stock, perhaps by corporate bylaws. Something like that would be the best investment I’d public dollars and present a new face to other semiconductor design customers. Then IFS would truly be a US based pure play Fab.

Just my opinion.

I get that, but wall street would hate it for precisely the reason you like it. If things go south that part of the new company stands alone, likewise you can price the individual two companies as I think X has more of a future than Y, or I think X and Y or linked, or I think Y has more of a future. This uncertainty Wall Street hates. Likewise other people see this as the story of myth, the myth of being self made and you proved your destiny via your success.

And it is not illogical for Wall Street to hate all of this, they can easily borrow money cheaper than issuing new stock (they are in fact doing stock buybacks), Wall Street likes Intel currently even if tech nerds do not. Legal arbitrage where since someone is willing to loan you money cheaper than your stock price appreciating you do that. Much like if you can get a bank loan cheaper than X assets appreciation you use that bank loan to purchase another thing that appreciates.

-----

What Intel wants right now is not money to build more fabs, for Wall Street is willing to give them money. Intel wants guaranteed promises that those fabs will be full and guaranteed wafer purchases since they are "investing" and a fab takes years to pay off. Liquid Capital to Illiquid Capital with a high rate of return.

I say we give them this (this is my opinion) for shortages of raw components causes greater economic lost downstream in the supply chain. For example 6 days ago Toyota gave investor guidance that they want to make 900k cars but can only make 540k cars since they still need chips and there is no point with building those extra 360k cars without the parts. Let me repeat that, Toyota is running their Car Foundry at 60% capacity since it can't get what it needs from dozens of Chip foundries (not just Intel)

 

[jpiniero]

I say we give them this (this is my opinion) for shortages of raw components causes greater economic lost downstream in the supply chain. For example 6 days ago Toyota gave investor guidance that they want to make 900k cars but can only make 540k cars since they still need chips and there is no point with building those extra 360k cars without the parts. Let me repeat that, Toyota is running their Car Foundry at 60% capacity since it can't get what it needs from dozens of Chip foundries (not just Intel)

That's because all of them gave away their chip production slots figuring that sales would collapse and there's no opportunity to gain slots back because every child in the US now needs a Chromebook.

I don't know who the hell is buying a car right other than the rental car companies rebuilding their fleet but I guess demand is picking up.

 

[Roland00Address]

That's because all of them gave away their chip production slots figuring that sales would collapse and there's no opportunity to gain slots back because every child in the US now needs a Chromebook.

I don't know who the hell is buying a car right other than the rental car companies rebuilding their fleet but I guess demand is picking up.

My father bought a used car 45 days ago. He also is driving another car that has over 300k miles on it.

But that is just examples, there are more things that use computer chips than cars and computers since chips are in everything. Of course most of those other things I am talking are using older foundry geometries like 28nm, 14nm and so on. But if you are Intel you want to build the best or close to best foundries you can make, even if we also need to simultaneously (we here being the greater economy and not intel or the militarily specifically) want people to also increase the amount of foundry space at leading edge geometries but also the mid geometries.