Discussion Intel's past, present and future

[jdubs03]

No one. The whole idea of the CHIPS act was "we need domestic silicon production", which was an allergic reaction to various chip shortages during covid etc. Which was hilarious since it arrived late and since Intel was probably not the best domestic semiconductor company to handle that problem. So in the spirit of just handing money to Intel under the assumption that hey, give money to our last best foundry, just . . . give them the money and call it a day. CHIPS was never a good idea in the first place, and sticking a few hundred post-docs up their butts isn't going to make it any less of a boondoggle.

If we are talking domestic semiconductor foundries Intel was exactly the right one. The only other option would’ve been global foundries and they aren’t doing sub 3 nm processes.

 

[DrMrLordX]

If we are talking domestic semiconductor foundries Intel was exactly the right one. The only other option would’ve been global foundries and they aren’t doing sub 3 nm processes.

Oh yeah let's give money to the company that systematically eliminated the entire American semiconductor industry. Nah, hard pass, but they didn't ask me. Meanwhile GF has active contracts with the DoD, an actual foundry business, low-power/voltage nodes suitable for things like automotive, etc.

 

[jdubs03]

Oh yeah let's give money to the company that systematically eliminated the entire American semiconductor industry. Nah, hard pass, but they didn't ask me. Meanwhile GF has active contracts with the DoD, an actual foundry business, low-power/voltage nodes suitable for things like automotive, etc.

Aren’t they at like 12nm? They’re nowhere near cutting edge. Now maybe for F35s that’s okay, but for the products that we all follow news on, it isn’t.

Obviously they’ve had problems particularly with getting 10nm right, but they’ve hit the bottom and there is some light at the end of the tunnel.

 

[DrMrLordX]

Aren’t they at like 12nm? They’re nowhere near cutting edge.

Yes, and CHIPS wasn't originally just about cutting edge. It was initially brought about to address the need for automotive chips since that's where the shortages were felt the most by domestic manufacturers.

 

[jdubs03]

Yes, and CHIPS wasn't originally just about cutting edge. It was initially brought about to address the need for automotive chips since that's where the shortages were felt the most by domestic manufacturers.

Indeed it wasn’t. But you seem to be advocating for no funding for Intel whereas everyone else gets some. Seems pretty harsh.

 

[DrMrLordX]

Indeed it wasn’t.

AI and the CHIPS Act Could Reshape the Auto… | FBT Gibbons

It's a familiar problem by now: a would-be buyer goes to a car lot or online vendor and can't find the make or model they had in mind.

frostbrowntodd.com

People were calling for the Feds to do something about automotive chip shortages. Some of our politicians realized that American semiconductor dominance had given way to something else entirely, though their idea of how to fix the problem was warped.

But you seem to be advocating for no funding for Intel whereas everyone else gets some. Seems pretty harsh.

My idea is: cut Intel out of the picture unless they're willing to undo some of the damage they did to American semiconductor manufacturing. Also Intel still doesn't have their foundry ducks in a row so why funnel cash to them when they're busy taking wafers from TSMC for their designs? The amount of money they spent on N3 is greater than what they stood to gain from the CHIPS act. But, you know what? They didn't listen to me. There are three possible outcomes: administer a large CHIPS grant to Intel overseen by ~500 federal workers, administer a large CHIPS grant to Intel administered by no one in particular, or effectively cancel the grant because hey let's fire some federal workers! Looks like option #3 is the one on the table for now.

Interestingly enough, if the Intel portion of CHIPS goes unawarded due to cuts in the federal workforce, the money slated for TSMC and GF may go unawarded as well.

 

[Joe NYC]

After all of the speculation, fake news stories of TSMC bailing out Intel fabs, buying them, managing them (and everything in between), Intel came up with exactly ZERO out of the TSMC proposal and press conference with Trump and C.C. Wei.

TSMC is going to:
- accelerate currently planed 3 fabs
- add additional 3 fabs
- add 2 packaging plants
- add R+D center
- $100 billion investment in addition to existing plan of $65 billion for total of $165 billion

And TSMC will do it with no US government subsidy. Maybe the drastic cuts in personnel administering of Chips Act was a prelude to phasing out of Chips Act.

C.C. Wei also made a slight dig against Intel, saying that chips being produced at the Arizona fab are the most advanced chips being produced in the US.

After this press conference, Intel is left with zero leverage. Trump does not seem to even consider Intel to be a player in the market, by saying TSMC is a monopoly.

If the plan proceeds, TSMC will, in a couple of years, surpass Intel in manufacturing in the US.

 

[dacostafilipe]

No one. The whole idea of the CHIPS act was "we need domestic silicon production", which was an allergic reaction to various chip shortages during covid etc. Which was hilarious since it arrived late and since Intel was probably not the best domestic semiconductor company to handle that problem. So in the spirit of just handing money to Intel under the assumption that hey, give money to our last best foundry, just . . . give them the money and call it a day. CHIPS was never a good idea in the first place, and sticking a few hundred post-docs up their butts isn't going to make it any less of a boondoggle.

I could agree about not spending that money all, instead of regulation how that money is spend, but "just give them the money", that's something else.

Well, let's just agree to disagree on that one 😉

 

[DavidC1]

After this press conference, Intel is left with zero leverage. Trump does not seem to even consider Intel to be a player in the market, by saying TSMC is a monopoly.

If the plan proceeds, TSMC will, in a couple of years, surpass Intel in manufacturing in the US.

Aside from the delusional folk that believe that WW3 will somehow be a net positive for Intel, it was always more logical to give the contracts to the company that is already doing as they say they are.

Intel was in such a mess that it would need Pat's vision perfectly being executed... and then some, and we know that wasn't the case, even if 18A turns out to be good.

 

[DrMrLordX]

If the plan proceeds, TSMC will, in a couple of years, surpass Intel in manufacturing in the US.

There's ways for Intel to fight back, but that means swallowing their pride and collaborating with other domestic also-ran foundries. Which honestly they needed to start yesterday.

I could agree about not spending that money all, instead of regulation how that money is spend, but "just give them the money", that's something else.

Well, let's just agree to disagree on that one 😉

Hey I didn't want them to get that CHIPS funding either, but I'm trying to be realistic. There's billions of dollars awarded with very little oversight all across the country in the form of grants and tax breaks without about as much oversight. If Congress were hell-bent on handing free money to Intel, it was always going to be a waste regardless of how many minders there were.

 

[Joe NYC]

There's ways for Intel to fight back, but that means swallowing their pride and collaborating with other domestic also-ran foundries. Which honestly they needed to start yesterday.

Intel made a mistake in not buying Global Foundries a few years ago, before their IPO.

Intel could have already been in foundry business, with customer relationships. Which something that is still missing.

 

[igor_kavinski]

Intel made a mistake in not buying Global Foundries a few years ago, before their IPO.

Or they could've ruined them completely. Intel has messed up their acquisitions far worse than they have messed themselves up. At least Intel stands, for now. Where are the companies they acquired? Softmachines VISC? Buried most likely because Intel considered them a threat (or most likely IDC because they had the potential to surpass the P-core designs in single threaded performance). Bought out VIA's x86 team too and then buried their processor designs. Even the Skymont team would've been disbanded had Intel not needed them really badly. Politics reign supreme inside Intel. The dirtiest kind.

 

[Joe NYC]

Or they could've ruined them completely. Intel has messed up their acquisitions far worse than they have messed themselves up.

Good point.

 

[desrever]

Most likely TSMC will just find a reason not to spend that money at some point. Its just to placate Trump for now.

 

[jdubs03]

If TSMC is making N3E while they’re on N2P/A16 in Taiwan, I wouldn’t be throwing a party.

You’ll still have a domestic manufacturer at a far superior process. And I know you don’t like it, but that’s still Intel.

 

[Joe NYC]

Most likely TSMC will just find a reason not to spend that money at some point. Its just to placate Trump for now.

Not this time. This time, there is an enforcement mechanism, and it is taking place in less than 2 months in new administration. 3 years 10 months to see this through.

And while TSMC was very lukewarm about Arizona at first, it seems like they are now more enthusiastic about it.

 

[igor_kavinski]

And while TSMC was very lukewarm about Arizona at first, it seems like they are now more enthusiastic about it.

They must know something about Intel that we don't. Like how long they have till they are broken apart piece by piece.

 

[Markfw]

You know, it was just about a year ago or so, all I could hear is "Intel is too big to fail".

 

[igor_kavinski]

You know, it was just about a year ago or so, all I could hear is "Intel is too big to fail".

Things have really taken a serious turn after Pat's firing. Even those not worried before are pretty concerned now. When you have current Intel employees publicly posting, hey, don't break Intel up! It means, they are almost sure that they will lose their jobs so why not try one last ditch effort to save their company?

 

[DavidC1]

Bought out VIA's x86 team too and then buried their processor designs. Even the Skymont team would've been disbanded had Intel not needed them really badly. Politics reign supreme inside Intel. The dirtiest kind.

Both E core and the Via's team were located in Austin. Even before acquisitions the E core team likely was influenced by them. From what I know lots of the team members are from them.

"Via" is just the company that owned Centaur Technologies, the company that did amazing things with a shoestring human and financial budget. It was thought to be "impossible" when they claimed x86 chip could be made with just $10 million USD and 2 years.

I think the integration accelerated after they bought out the CPU division team a few years ago. The description on the exact nature of the deal was murky. It probably had to do with getting employees and the knowledge of how they operated so efficiently.

 

[Joe NYC]

They must know something about Intel that we don't. Like how long they have till they are broken apart piece by piece.

It's not a secret to anyone (never mind TSMC) that AMD is beating Intel N3 based products using N4 node.

TSMC also knows that N2 node is going to be more competitive than Intel 18A. Where wil that leave Intel Products?

Dell, for one, is running for exits:

AMD Exec: Dell Commercial PC Deal To Cover Broad Customer Base

The head of AMD’s client business unit, Rahul Tikoo, revealed to CRN in an interview more details about its landmark commercial PC deal with Dell Technologies and outlined how AMD plans to win in this year’s Windows 11 PC refresh.

www.crn.com

 

[igor_kavinski]

Crap! If Dell goes then what's left for Intel??? Asian markets???

 

[Joe NYC]

Crap! If Dell goes then what's left for Intel??? Asian markets???

Once fully phased in (say by the end of 2025), AMD could gain 5-6% of global client PC market share, just trough Dell alone.

 

[DrMrLordX]

Crap! If Dell goes then what's left for Intel??? Asian markets???

Intel would still have volume on their side for certain products. They can ink deals with other OEMs. The terms won't be as favorable.

 

[Gideon]

A rather interesting article:

https://www.investing.com/news/transcripts/intel-at-morgan-stanley-conference-strategic-focus-on-foundry-and-ai-93CH-3913018

reddit discussion:

https://www.reddit.com/r/hardware/comments/1j5frht/intel_at_morgan_stanley_conference_strategic

Some interesting highlights:

• 30% of wafers outsourced to TSMC currently, "probably a high water mark for [Intel]"
• Panther Lake/18a have better yields than Meteor Lake at same stage in development cycle, "I want to be very clear. Panther Lake is on track to launch in the second half of this year. That launch date has not changed."
• Goal is to be "second largest external foundry by 2030," need roughly 15 billion in external foundry revenue to achieve that, aiming for 10% of the external foundry market.
• Intel 18a competes with both N2 and N3
• First external tapeout on 18a in first half of 2025.
• Panther Lake launching second half of 2025, "real volume" in 2026, won't help substantially with margins until then.
• Hybrid bonding / advanced packaging issues with Clearwater Forest are "fixed," 18a "looks good," delay "not a wafer issue."
• ASP per wafer will go up "three times faster than cost per wafer" on the road to Intel Foundry breaking even.
• CapEx guidance down to $20 billion from $20-23 billion for the year, "we've got north of $50 billion in what we call construction in progress or assets under construction."
• Intel Products is "a little bit of a first among equals because we know we have to get that right."
• Intel Datacenter strategy is to "optimize for share over margin" and to be "a little bit more price aggressive where we think we need to be in order to do that because we think it’s important to defend the share."
• E-core server market (i.e. Sierra Forest) is "probably a little bit smaller than we originally thought," server market share is "really going to be dependent upon how we execute on the P core more than the [E core]."
• Making progress on Altera stake sale/IPO
• Mobileye "not core to the business," leadership will "be pragmatic about balancing capital needs."
• 18a looks more interesting to high performance compute foundry customers compared to mobile customers.

I particularly liked this comment:

> E-core server market (i.e. Sierra Forest) is "probably a little bit smaller than we originally thought," server market share is "really going to be dependent upon how we execute on the P core more than the [E core]."

This is very telling. ARM has taken over the small core server market. Where x86 is still in demand is very high performance server needs where AMD leads.