Gamers Nexus: The Future of Intel
- Thread starter XYang2023
- Start date
hist78
Well-known member
This is a great video. According to one of the managers interviewed in this video, the Intel Fab 52 and 62 in Arizona are still 3 to 5 years to complete. Then where can Intel HVM the upcoming 18A in 2025? It seems the Oregon fab is the only option. But does it have enough capacity to support both Intel and external customers' needs?
Last edited:
hist78
Well-known member
Intel broke ground on its Arizona Fabs 52 and 62 in September 2021, while TSMC began construction of its Arizona fab in June 2021. The Intel and TSMC Arizona sites are about an hour's drive apart in the Greater Phoenix area. TSMC is expected to begin high-volume manufacturing (HVM) in the first half of 2025, while Intel is expected to complete at least one of its two fabs by 2027.
This means it will take TSMC about 4 years to build a fab and start HVM, while Intel will need 6+ years to achieve the same goal.
It’s an interesting subject to explore the reasons behind this 2-year (or more) difference.
This means it will take TSMC about 4 years to build a fab and start HVM, while Intel will need 6+ years to achieve the same goal.
It’s an interesting subject to explore the reasons behind this 2-year (or more) difference.
siliconbruh999
Well-known member
Won't the process that is to be manufactured affect it considering TSMC is using 4 years older process Nd intel wilt be the latest 18/14A
hist78
Well-known member
Yes, I agree. But Intel has said it plans to do 18A HVM in 2025, which means Intel likely knows what it’s doing. However, where will Intel execute 18A HVM? I guess that leaves Intel Oregon as the only option, meaning the capacity will be relatively limited.
This has several implications:
- 1. For the next 3 to 5 years, Intel must collaborate with TSMC to secure enough manufacturing capacity to support various Intel product deliveries. This also allows Intel to reserve some meaningful Intel Foundry (IF) capacity to begin engaging with new IF customers.
- 2. This could be an additional reason why Intel didn't win the large Sony PlayStation 6 deal. The known reasons so far are that the profit margin was too slim and AMD has a stronger position in providing backward compatibility. I suspect that Intel's limited manufacturing capacity may also be a major factor.
This video reportedly was filmed a year ago.
hist78
Well-known member
Good to know. Where did you see this info? Thanks.
I seached arizona new fabs news online, they were planning to begin production in at least one of these two AZ new fabs by the end of 2024, which is clearly unlikely at this point. However, in terms of delay, even an Intel-unfriendly media digitimes claimed it will be sometime in 2025.
Regarding your comment on Intel/Sony deal, I don't think it has anything to do with the lack of fab capacity, because these chips are not supposed to be high end, and are not going to be manufactured on the leading edge nodes. That piece of news actually happened a year ago, and the details were likely leaked by someone who wanted to drive $intc price lower in early Sept, when there were a flood of "bad news" driving $intc to as low as $18.xx.
johnson.z.wu
New member
Can the Fab start operating before "completion"? From the video, it seems that the FAB is already running.
That makes sense. If they want to be ready in 2025 for HVM, should Intel do building and fitting in parallel?
johnson.z.wu
New member
I dont know the chip fab. It is quite normal for other industry.
Artificer60
Well-known member
Back in April The Register reported:
"The two facilities in Chandler, Arizona – Intel's Fab 52 and Fab 62 – have made the most progress since Gelsinger announced them in early 2021. As of December 2023, Intel revealed work on the fab's concrete superstructure had been completed. Construction crews are now working to install the automated material handling system, which Intel describes as an "automated highway" used to transport wafers.
The fabs are expected to come online later this year or early in 2025, and are slated to produce chips based on Intel's next-gen Angstrom era process tech – including its mass market 18A node."
I have not found anything more recent to indicate that the timeline has changed. Assuming that the Oregon site starts 18A production in early 2025, I would expect Intel's Copy Exactly methodology to have the Arizona fabs up and running on 18A 6 to 8 months later as that seems to be the typical time frame for a process to be transferred from their development fab to ramp.
"The two facilities in Chandler, Arizona – Intel's Fab 52 and Fab 62 – have made the most progress since Gelsinger announced them in early 2021. As of December 2023, Intel revealed work on the fab's concrete superstructure had been completed. Construction crews are now working to install the automated material handling system, which Intel describes as an "automated highway" used to transport wafers.
The fabs are expected to come online later this year or early in 2025, and are slated to produce chips based on Intel's next-gen Angstrom era process tech – including its mass market 18A node."
I have not found anything more recent to indicate that the timeline has changed. Assuming that the Oregon site starts 18A production in early 2025, I would expect Intel's Copy Exactly methodology to have the Arizona fabs up and running on 18A 6 to 8 months later as that seems to be the typical time frame for a process to be transferred from their development fab to ramp.
Artificer60
Well-known member
The clean room has to be fully enclosed and operational. It is not unusual to start production before all the tooling is moved in, but all the facilities and the fab building itself must be completed first.
hist78
Well-known member
I watched the video again, and the Intel construction managers said it would take 3 to 5 years to complete from the September 2021 starting date. This puts one of the two fabs (52 & 62) into completion between 2024, 2025, or 2026—earlier than I originally thought. We'll see if that's the case next year.
Regarding Intel losing the PlayStation 6 deal, the interim release of the PlayStation 5 Pro (coming out November 7, 2024) is believed to use TSMC N4 already. That means the PlayStation 6 will likely use TSMC’s N3 or N4 at a minimum. According to some reports, the PlayStation 6 is expected to hit the market between 2027 and 2028, although I believe it could arrive earlier, perhaps in 2026. In that timeframe, Intel's Intel 3 and Intel 4 (likely from the Intel Ireland fab) would be Intel's best option for Sony. Older nodes like Intel 7 are no longer competitive in terms of cost and performance.
Although Intel is eager to break into the foundry business and secure a "whale" customer, why didn't it win the Sony PlayStation 6 deal, which involves producing up to a billion units? By any standards, that’s a huge order.
According to the Reuters report, it’s because of Intel's minimum profit margin requirements, and AMD's superior backward compatibility. While those reasons are likely true, I believe there’s more to the story. For instance, Sony’s one billion units over 10 years translates to about 100 million units per year. That's a significant order that would greatly improve Intel Foundry’s utilization. So why can’t Intel make enough money from this contract, while TSMC+AMD can? There’s no way AMD and TSMC would take a loss on such a large contract—because even a $1 loss per unit would mean a $1 billion total loss!
Why can't Intel's "superior" IDM model compete against AMD(fabless) + TSMC (Pure play Foundry)?
Is Intel 3 or Intel 4 too expensive to produce, or is Intel Ireland’s fab lacking the capacity to handle such a massive order?
Intel is looking for a whale customer, but when such an opportunity presents itself, can Intel rise to the challenge? If Intel can’t compete with TSMC + AMD in terms of cost and profitability on a deal this size, what’s the point of pursuing the foundry business?
As for the possibility of market manipulation, in my opinion, it's always a possibility, but it's unlikely in this case—especially considering Reuters' internal controls and legal issues. A more logical source of the information could be Intel itself. You might notice that several positive Intel-related news pieces were released around the same time. It’s typical of Intel to use good news to offset bad news, or to minimize the impact by preemptively revealing negative information.
Note that this deal negotiation happened more than one year ago. Intel had a different assessment of itself than today. It was unlikely Intel would think it will not have capacity in 2027/2028 timeframe.
Also, you are painting this deal as quite attractive, a whale for IFS. It is anything but. All these sony or xbox deals don't make much money, and are not viewed as "high end" when they hit the market. A desirable whale is a high volume, cutting edge, 18A customer that can help fund the fab and establish its credibility. This deal was not, because it only satisfies the high volume criteria.
Also, you are painting this deal as quite attractive, a whale for IFS. It is anything but. All these sony or xbox deals don't make much money, and are not viewed as "high end" when they hit the market. A desirable whale is a high volume, cutting edge, 18A customer that can help fund the fab and establish its credibility. This deal was not, because it only satisfies the high volume criteria.
Ideally, customers that offer both high volume and cutting-edge demands are the most desirable. However, it's doubtful that there are many such opportunities for Intel. If Intel is serious about its foundry services, it needs to build trust quickly. Supporting customers with either high volume or cutting-edge needs could still help enhance Intel's reputation.