News Broadcom disappointed with Intel 18A process technology — says it's not currently viable for high-volume production

[Admin]

Broadcom believes that Intel 18A is not viable for mass production, but the evaluation continues.

Broadcom disappointed with Intel 18A process technology — says it's not currently viable for high-volume production : Read more

 

[Lucky_SLS]

Correction needed : chaplets > chiplets

"Intel needs to be able to make chaplets of this scale with good yields"

 

[magbarn]

I guess Intel went full GM/Boeing in the last 5-10 years. Don't understand how the company could allow itself to screw up in so many levels. Did they replace all their management with useless MBA's in the last decade?

 

[TheSecondPower]

I don't know what to make of this article. What can I compare it to? When else has foundry satisfaction been in news? When Samsung and TSMC were both making iPhone chips? I guess more recently when Nvidia 3000 chips were made on the Samsung 8nm node. Does Broadcom make a lot of chips? Are they unhappy with the rate of progress, the tooling, the defect rate, or the performance of the defect-free parts? Or is piling on Intel just what'd popular today? What's the purpose in saying anything at all, since having Intel as an alternative is no worse than having no alternative. Publicly giving a bad report about Intel shouldn't help Broadcom.

 

[Marlin1975]

I don't know what to make of this article. What can I compare it to? When else has foundry satisfaction been in news? When Samsung and TSMC were both making iPhone chips? I guess more recently when Nvidia 3000 chips were made on the Samsung 8nm node. Does Broadcom make a lot of chips? Are they unhappy with the rate of progress, the tooling, the defect rate, or the performance of the defect-free parts? Or is piling on Intel just what'd popular today? What's the purpose in saying anything at all, since having Intel as an alternative is no worse than having no alternative. Publicly giving a bad report about Intel shouldn't help Broadcom.

You have it backwards, Intel needs broadcom not the other way.

Intel is desperate for cash flow and having a 3rd party say your "company saving" node is not good is a bad sign. Broadcom can go else where like you said. Intel needs money so it has few options.

 

[Lucky_SLS]

I guess it depends on how the wafer contract is made. If Intel is willing to absorb the defective wafer cost, it can be a good deal and might attract customers from TSMC.

 

[TheSecondPower]

You have it backwards, Intel needs broadcom not the other way.

Intel is desperate for cash flow and having a 3rd party say your "company saving" node is not good is a bad sign. Broadcom can go else where like you said. Intel needs money so it has few options.

My point is simply that Broadcam has nothing to gain by giving a bad report about Intel. Since there are only 3 foundry providers outside of China it's unwise to alienate one. Whoever gave this report to Reuters is rogue or stupid.

And my larger point is that without more context, this bad report about Intel is meaningless. Many things could explain the dissatisfaction, and Intel is still new to catering to foundry customers.

 

[bit_user]

I guess it depends on how the wafer contract is made. If Intel is willing to absorb the defective wafer cost, it can be a good deal and might attract customers from TSMC.

Intel is not currently in a position to absorb more losses, though. They need customers, but they also need to recoup the upfront investment in these nodes. Poor yields put them between a rock and a hard place.

 

[ThomasKinsley]

I guess Intel went full GM/Boeing in the last 5-10 years. Don't understand how the company could allow itself to screw up in so many levels. Did they replace all their management with useless MBA's in the last decade?

Definitely 10 years. The Downfall, Spectre and Meltdown bugs crippled Intel chips by 30%-50%. Since Intel was making only incremental gains in performance every year, these exploits set Intel back by years. They also struggled for years to get beyond 10nm capability - something even China managed to do using DUV machines. It's the same old story. They were so far ahead of AMD that they became apathetic. I'm convinced now that the Core 2 Duo and Quad series were an accident. Before that Intel was running Pentium 4 chips that were hot and slow.

My point is simply that Broadcam has nothing to gain by giving a bad report about Intel. Since there are only 3 foundry providers outside of China it's unwise to alienate one. Whoever gave this report to Reuters is rogue or stupid.

Call it the BlackBerry effect. Bad press reduces prospects, which generates further bad press, which reduces further prospects, etc. Intel can get out of this viscous cycle if they produce a good node.

 

[Pierce2623]

I guess it depends on how the wafer contract is made. If Intel is willing to absorb the defective wafer cost, it can be a good deal and might attract customers from TSMC.

Intel is no longer sitting on the necessary cash pile for deals like that. They’re hemorrhaging over a billion a quarter right now.

 

[why_wolf]

I guess it depends on how the wafer contract is made. If Intel is willing to absorb the defective wafer cost, it can be a good deal and might attract customers from TSMC.

Yeah it all depends on how low of a price they can offer, while still turning a profit. Intel has historically shown they have a problem with charging anything less than 1st tier pricing even when the product isn't worth that. Which results in the product under selling and eventually being cancelled. Wasting time and money.

Of course if they have to offer a price lower than production cost than that isn't sustainable long term unless they are sure they can reduce the waste quickly and turn the contract profitable in the longer term.

Acquiring customers for the foundry business is beyond critical for Intel. Even if their node was pumping out wafers with .1 defect density they still need to offer pricing lower than TSMC to win customer contracts.

Intel is for all intents and purposes the new guy on the block that nobody knows or trusts. They have to offer discounts until they earn a reputation for doing a good job.

 

[Gururu]

I don't know what to make of this article. What can I compare it to? When else has foundry satisfaction been in news? When Samsung and TSMC were both making iPhone chips? I guess more recently when Nvidia 3000 chips were made on the Samsung 8nm node. Does Broadcom make a lot of chips? Are they unhappy with the rate of progress, the tooling, the defect rate, or the performance of the defect-free parts? Or is piling on Intel just what'd popular today? What's the purpose in saying anything at all, since having Intel as an alternative is no worse than having no alternative. Publicly giving a bad report about Intel shouldn't help Broadcom.

Agreed. The Reuters blurb itself is thin in information and cites three undisclosed sources "familiar" with the situation. This is not coming from an official source and needs to be taken without context.

 

[TheHerald]

I don't know what to make of this article. What can I compare it to? When else has foundry satisfaction been in news? When Samsung and TSMC were both making iPhone chips? I guess more recently when Nvidia 3000 chips were made on the Samsung 8nm node. Does Broadcom make a lot of chips? Are they unhappy with the rate of progress, the tooling, the defect rate, or the performance of the defect-free parts? Or is piling on Intel just what'd popular today? What's the purpose in saying anything at all, since having Intel as an alternative is no worse than having no alternative. Publicly giving a bad report about Intel shouldn't help Broadcom.

Public pressure to get a better deal.

Of course it's unethical but in the corporate business that's the normal I guess.

 

[bit_user]

Public pressure to get a better deal.

Of course it's unethical but in the corporate business that's the normal I guess.

I don't see why Intel would give them a better deal, just because they say bad things about the node.

Fab customers aren't like you or me. They make decisions based on hard data. Badmouthing or speaking highly of a fab is unlikely to affect someone else's decision about whether to use it.

As for investors, any impact from comments like these will be very short-lived. Ultimately, what they care about are the profits and losses.

 

[thestryker]

I guess Intel went full GM/Boeing in the last 5-10 years. Don't understand how the company could allow itself to screw up in so many levels. Did they replace all their management with useless MBA's in the last decade?

After Grove stepped down in 1998 Barrett (always forget about him for some reason) stepped down in 2005 until Gelsinger was hired in 2021 the CEOs were all financial types. You can imagine what happened next.

 

[vanadiel007]

I still believe they should put up a joint venture with AMD, rather than compete against AMD.
Both companies could benefit from that, while at the same time provide balance against Nvidia in the server and emerging AI market.

 

[ikjadoon]

My point is simply that Broadcam has nothing to gain by giving a bad report about Intel. Since there are only 3 foundry providers outside of China it's unwise to alienate one. Whoever gave this report to Reuters is rogue or stupid.

And my larger point is that without more context, this bad report about Intel is meaningless. Many things could explain the dissatisfaction, and Intel is still new to catering to foundry customers.

What makes you think Broadcom leaked this to Reuters?

I might expect this to be an Intel leak. Intel is letting go 15,000 employees & with Intel stock as it is, I imagine some Intel (ex) employees are neither that loyal nor that protective about Intel any more.

//

"Meaningless" is a little strong. We wouldn't expect Reuters to dive deep into wafer manufacturing to detail what precisely allegedly went wrong with Broadcom's 18A wafers.

Qualcomm also allegedly rejected Intel 20A / 18A two years ago; it was not necessarily quality, but timelines (assuming contracts are arranged as X fabrication quality at Y volume at Z date). Here, the WSJ also was relatively vague: I suspected partly because WSJ wouldn't be able to explain the details to their target audience and because the end result (deal or no deal) is more important than the details.

In early 2022, Intel’s foundry arm sent a delegation to Qualcomm’s San Diego headquarters, where they met with CEO Cristiano Amon. Then Intel missed a June performance milestone toward producing those chips commercially. It missed another in December.

Qualcomm executives concluded Intel would struggle making the kind of cellphone chips they wanted, even if it succeeded in making high-performance processors. Qualcomm told Intel it was pausing work while it waits for Intel to show progress, according to people involved in the discussions.

Source: https://www.wsj.com/articles/intel-gelsinger-nvidia-turnaround-30febac6

 

[Jame5]

Publicly giving a bad report about Intel shouldn't help Broadcom.

It definitely should.

By putting public pressure on Intel, even if they are happy with the node, they can get capacity at reduced cost as Intel chases their business. It's a very standard business power play.

Is it dumb? Sure. Does it work? Also yes.

*Edit: Exactly as someone said above me. Terrible first post on my part.

 

[Gururu]

Broadcom full of bad news today. I guess it would help to offload blame to Intel.