News Intel and TSMC agree to form chipmaking joint venture: Report

[Admin]

Intel and TSMC have reportedly reached a tentative agreement to form a joint venture for operating Intel's U.S. fabs.

Intel and TSMC agree to form chipmaking joint venture: Report : Read more

 

[Dustyboy1492]

So TSMC gets 20% of the new venture with just people and IP, no cash? They could just run it into the ground, intentionally hamstring the new company, and then profit from the monopoly afterward. This just makes no sense on so many levels. Is TSMC going to get Intel IP? Will the new company stop doing R&D and just farm that out to TSMC, forever trailing in process development? This really sounds like a TSMC monopoly on all leading edge nodes.

 

[Gururu]

This is better than the doomsday alternatives Intel was facing. My guess it will fall to 1 owner unless it does gangbusters.

 

[rluker5]

You know, that promised Chips Act funding would have been more helpful to support Intel's fabs with all of the money spent on building fabs to make chips in the US. Shame the US gov lost the money on the way to giving it to its intended recipients who may or may not have included that in their planning.

Edit: Isn't Intel about to take the lead on process node this year and then widen the gap when they start using high NA? Especially for tariff avoiding local markets. I'm feeling pretty suspicious about how good this is for Intel.

 

[KennyRedSocks]

They could just run it into the ground

That'd be a bit hard to do with only a 20% stake.

 

[LibertyWell]

Almost all big corporations already own each other equalling a giant monolith: https://old.bitchute.com/video/EARLAOYg47YQ

Letting them further merge together via moves like this is a bad ideal to put it mildly. Where's the FTC and the Sherman Antitrust Act?

 

[thestryker]

I've yet to see any argument behind this that makes any sense. I don't see how it benefits Intel let alone how it benefits TSMC.

The last manufacturing joint venture Intel had was IMFT which existed only to share storage R&D and fab costs. The deal between Intel and UMC seems similar to this, but in that case UMC brings the commercial node experience and Intel brings the fabs. This type of partnership doesn't make sense at all for leading edge silicon fabrication because it would undermine the business model entirely.

If it's not this type of partnership then it makes literally no sense for TSMC. What interest would they have in running competing fabs?

The only thing Intel would get out of this is some cost savings and potentially more people who have experience with contract fabrication. This doesn't really sound like any sort of win and certainly doesn't help the company long term.

 

[phead128]

I seriously doubt TSMC wants to touch Intel Foundry, which is deadweight, unless Trump admin is putting a gun to Taiwan's head and forcing it to.

 

[bit_user]

Shame the US gov lost the money on the way to giving it to its intended recipients who may or may not have included that in their planning.

What do you mean "lost the money"? Were any of the funds due for milestones that had been achieved not successfully delivered?

And yes, for projects that had been greenlit, I'm sure those funds were included in the planning.

 

[bit_user]

That'd be a bit hard to do with only a 20% stake.

TSMC is supposedly going to be the one running the plants, using their IP. So, it doesn't really matter that their ownership stake is only 20%. In fact, having a lower ownership stake means they have less to lose by it the Intel fabs not being very successful.

What I'd suspect is that TSMC will steer the Intel fabs towards less lucrative & cutting-edge nodes, while keeping the most desirable nodes for their own fabs.

I've yet to see any argument behind this that makes any sense. I don't see how it benefits Intel let alone how it benefits TSMC.

TSMC gets more US-based production volume in a timeframe well in advance of when their new buildouts would come online.

As for Intel, well, it's not really about what's good for Intel, is it? I think the motive for pushing this scheme is just to ramp up US-based production volume more quickly, in order to feed the demand for AI silicon. I'm sure Intel is either being given an offer that's too tempting for its investors to refuse, or else perhaps the Defense Production Act will be invoked to eliminate Intel's agency in the decision.

 

[thestryker]

TSMC gets more US-based production volume in a timeframe well in advance of when their new buildouts would come online.

You're assuming the fabs are interchangeable which I highly doubt they are. While it might be faster than building from scratch it would hardly be fast.

As for Intel, well, it's not really about what's good for Intel, is it? I think the motive for pushing this scheme is just to ramp up US-based production volume more quickly, in order to feed the demand for AI silicon. I'm sure Intel is either being given an offer that's too tempting for its investors to refuse, or else perhaps the Defense Production Act will be invoked to eliminate Intel's agency in the decision.

Intel loses all future CHIPs funding and likely would have billions clawed back. This scheme isn't going to be fast and it would be extremely stupid. None of what you said makes any logical sense, but I also wouldn't put it past the fools in charge.

 

[TerryLaze]

TSMC is supposedly going to be the one running the plants, using their IP. So, it doesn't really matter that their ownership stake is only 20%. In fact, having a lower ownership stake means they have less to lose by it the Intel fabs not being very successful.

What I'd suspect is that TSMC will steer the Intel fabs towards less lucrative & cutting-edge nodes, while keeping the most desirable nodes for their own fabs.


TSMC gets more US-based production volume in a timeframe well in advance of when their new buildouts would come online.

As for Intel, well, it's not really about what's good for Intel, is it? I think the motive for pushing this scheme is just to ramp up US-based production volume more quickly, in order to feed the demand for AI silicon. I'm sure Intel is either being given an offer that's too tempting for its investors to refuse, or else perhaps the Defense Production Act will be invoked to eliminate Intel's agency in the decision.

Did you read any article that said all of that?! Care to share? Because this article didn't touch on any of that.
For all we know TSMC is just going to be the token figure head, forced to that position if they want to do any business in the US, just so that other companies have less of an issue of using intel FABs.
Actually for all we know it's all based on rumors, it might be true but it might also not be.

 

[nogaard777]

The only good I can think of coming from this is Intel is good at designing chips, and making chips, but they have little experience in making chips for other firms. THAT'S what TSMC is good at and what Intel has to gain.

But I can't help thinking TSMC was more than happy to agree when all they have to do is supply some people and time, of which magically gives them a 20% stake in what Intel has tens of billions invested, but also they get a sneak peek into how Intel has leapfrogged them to have 2nm (18A) ready before they do.

Intel dominated TSMC for decades and they did it without the massive subsidies and tax breaks the Taiwanese government gave to TSMC. A 20% stake just for being in an advisory role seems a stupidly high price to pay while you show the competition your entire playbook?

This just screams of Intel being thrown under the bus so the body parts can be sold off to the highest bidder.

 

[bit_user]

You're assuming the fabs are interchangeable which I highly doubt they are. While it might be faster than building from scratch it would hardly be fast.

I'm certainly not assuming they're interchangeable, but I think it's reasonable to expect they could retrofit an existing Intel fab with a TSMC production line much faster than they could build new capacity from scratch.

For all we know TSMC is just going to be the token figure head, forced to that position if they want to do any business in the US, just so that other companies have less of an issue of using intel FABs.

It wouldn't make any sense to have TSMC involved, if they weren't converting Intel's fabs over to their production process.

Actually for all we know it's all based on rumors, it might be true but it might also not be.

Yes, this is all rumors and speculation, right now. Your guess is as good as mine.

 

[bit_user]

But I can't help thinking TSMC was more than happy to agree when all they have to do is supply some people and time, of which magically gives them a 20% stake in what Intel has tens of billions invested, but also they get a sneak peek into how Intel has leapfrogged them to have 2nm (18A) ready before they do.

The only good reason to get TSMC involved would be for their IP.

Intel dominated TSMC for decades and they did it without the massive subsidies and tax breaks the Taiwanese government gave to TSMC.

They dominated until about 2018 or so. The foundry business is a lot more difficult and expensive than it used to be. That's why Intel had to get outside customers. They can't keep funding their fabs and node R&D on the backs of their own products.

This just screams of Intel being thrown under the bus so the body parts can be sold off to the highest bidder.

Yes, it's a bad deal for Intel, long-term. Unfortunately, given the current state of the stock market, I expect a lot of investors will jump at the chance for a quick payout on their Intel shares.

 

[TerryLaze]

It wouldn't make any sense to have TSMC involved, if they weren't converting Intel's fabs over to their production process.

It wouldn't make sense for tsmc...

It wouldn't make sense for intel otherwise,
they would give up 20% of their fabs but would gain all of the customers that would be afraid to use intel fabs but would be ok with using tsmc fabs.

 

[thestryker]

I'm certainly not assuming they're interchangeable, but I think it's reasonable to expect they could retrofit an existing Intel fab with a TSMC production line much faster than they could build new capacity from scratch.

It's hard to say how much faster it would be due to the complex subflooring. If there is anything there that had to be changed out you'd likely be talking about moving EUV machines and there's no way that isn't both expensive and time consuming.

 

[TerryLaze]

That's why Intel had to get outside customers. They can't keep funding their fabs and node R&D on the backs of their own products.

That will never make any sense to me.
A FAB can not have more than 100% capacity, if you have 100% capacity with your own products you get to sell them at a premium, you are not just paid to manufacture them you are also getting paid the normal margin.

How will having outside customers that will only pay for manufacturing be more lucrative than getting the normal margin on top of that?!
If intel uses any extra capacity to pump out GPUs they will make more money than having all of the customers.

 

[bit_user]

That will never make any sense to me.
A FAB can not have more than 100% capacity, if you have 100% capacity with your own products you get to sell them at a premium, you are not just paid to manufacture them you are also getting paid the normal margin.

What you're not considering is the R&D costs of a node. Those are fixed, irrespective of how many customers or production volume you have. So, the more you can scale your production volume, the more you can amortize those up-front costs.

You're also assuming there are no economies to be gained by scaling up production further, which I think is also incorrect.

By your logic, IDM 2.0 makes no sense. Why would Intel have tried IDM twice, if it didn't make any sense, with even the new CEO talking up the IDM strategy in his first public announcements? Do you really think Intel is so clueless? And if IDM weren't essential to the business strategy, why wouldn't that have been tossed out, already?

You can't explain what Intel is doing (ignoring the current rumor), unless what I'm saying is right.

 

[rluker5]

That will never make any sense to me.
A FAB can not have more than 100% capacity, if you have 100% capacity with your own products you get to sell them at a premium, you are not just paid to manufacture them you are also getting paid the normal margin.

How will having outside customers that will only pay for manufacturing be more lucrative than getting the normal margin on top of that?!
If intel uses any extra capacity to pump out GPUs they will make more money than having all of the customers.

I think it might be a matter of scale. R&D costs relatively less the more you produce at each node. If Intel fabs more than they can sell of their products then it starts t make sense. What doesn't make sense is when they have their chips fabbed by someone else to fab someone else's chips.

 

[bit_user]

What doesn't make sense is when they have their chips fabbed by someone else to fab someone else's chips.

Which chips are you talking about? Their GPUs or the actual CPU tiles?

With the GPUs, I think I read somewhere that Intel 7 wasn't well-suited to GPUs. Not sure about why they didn't opt for Intel 3, but perhaps cost?

As for the CPU nodes, Intel said at the time of the Lunar Lake announcement that their designers were given the freedom to choose the best node and I guess it turned out that TSMC N3B was a better node than Intel 3 (with 20A coming too late). Then, because they already had their designs ported to N3B, it became an easy decision to cancel 20A for Arrow Lake and just stick with TSMC. The increased costs of using TSMC for Arrow Lake must've been less than what they saved by scrapping 20A and just focusing on 18A.

So, I think it really just has to do with timing, Intel still running behind TSMC, and the fact that TSMC had a wider variety of nodes, some of which are better-suited to things like GPUs. What doesn't make any sense is refusing to use a competitor's fab when it makes sense for the business, just out of pride. Especially if they're eventually planning on spinning off the fabs anyway, in which case they'd ultimately have the freedom to choose the best node for the product & delivery schedule.

The idea of letting the design team choose the most appropriate fab is consistent with the idea of trying to offer the best products possible, unlike what happened with CPUs like Ice Lake, where they had no choice but to use Intel's 10 nm node, and the product clearly suffered for it.

 

[TerryLaze]

What you're not considering is the R&D costs of a node. Those are fixed, irrespective of how many customers or production volume you have. So, the more you can scale your production volume, the more you can amortize those up-front costs.

You're also assuming there are no economies to be gained by scaling up production further, which I think is also incorrect.

By your logic, IDM 2.0 makes no sense. Why would Intel have tried IDM twice, if it didn't make any sense, with even the new CEO talking up the IDM strategy in his first public announcements? Do you really think Intel is so clueless? And if IDM weren't essential to the business strategy, why wouldn't that have been tossed out, already?

You can't explain what Intel is doing (ignoring the current rumor), unless what I'm saying is right.

What you are saying here makes sense, intel makes more FAB capacity then they need to, or can use by themselves, to make more money in the long run.
This is drastically different from what you said before which was
"They can't keep funding their fabs and node R&D on the backs of their own products."

 

[phead128]

That will never make any sense to me.
A FAB can not have more than 100% capacity, if you have 100% capacity with your own products you get to sell them at a premium, you are not just paid to manufacture them you are also getting paid the normal margin.

How will having outside customers that will only pay for manufacturing be more lucrative than getting the normal margin on top of that?!
If intel uses any extra capacity to pump out GPUs they will make more money than having all of the customers.

R&D costs into leading nodes need to be recouped. It's extremely expensive,

"They can't keep funding their fabs and node R&D on the backs of their own products."

What Bituser said makes complete sense. Intel does not have competitive AI GPU offerings, which is now the main money maker of semiconductor epoch. It doesn't have competitive mobile chip offering, 18A isn't really for mobile phones is it? They need 14A for mobile chips, Its own products for servers and laptops isn't going to carry Intel Foundry.

Anyways, TSMC is only doing this because Trump Admin is threatening tariffs on Taiwan and threatening to pull CHIPS ACT. Otherwise, it makes no sense whatsoever.

 

[jp7189]

This makes sense for TSMC and the US if a calamity is expected to hit Tawain. It might make sense for INTC investors if they believe TSMC expertise will fix Intel fab problems.

 

[abufrejoval]

Almost all big corporations already own each other equalling a giant monolith: https://old.bitchute.com/video/EARLAOYg47YQ

Letting them further merge together via moves like this is a bad ideal to put it mildly. Where's the FTC and the Sherman Antitrust Act?

And those are just US regulators and laws, which the current boss tends to replace with whatever he feels like today.

I doubt he'll take no for an answer by all those other anti monopoly bodies globally out there.