Intel 14nm capacity issue
- Thread starter Fred Chen
- Start date
Checking the list of Intel fab sites: List of Intel manufacturing sites - Wikipedia, it is hard to see why there would be shortage of 14nm capacity.
Fab 24 Ireland and Fab D1D Oregon are dedicated solely to 14nm, Fab 32 Arizona and Fab D1C Oregon are 22/14nm.
10nm does not have any dedicated fabs, only Fab D1X Oregon (14/10nm) and Fab 28 Israel (22/10nm).
Fab 24 Ireland and Fab D1D Oregon are dedicated solely to 14nm, Fab 32 Arizona and Fab D1C Oregon are 22/14nm.
10nm does not have any dedicated fabs, only Fab D1X Oregon (14/10nm) and Fab 28 Israel (22/10nm).
FYI:
Intel's interim CEO, Bob Swan, says the chipmaker has enough supply to meet its full-year revenue outlook.
"We continue to believe we will have at least the supply to meet the full-year revenue outlook we announced in July, which was $4.5 billion higher than our January expectations," he wrote in a post entitled "Supply Update"on the chipmaker's website Friday.
Intel shares closed up 3.1 percent Friday.
On July 26, Intel gave 2018 full year non-GAAP sales guidance of $69.5 billion.
Swan said the company is "making progress" in its transition to the 10-nanometer manufacturing process and expects "volume production" in 2019.
Intel interim CEO says chipmaker has supply to meet its full-year revenue outlook
Intel's interim CEO, Bob Swan, says the chipmaker has enough supply to meet its full-year revenue outlook.
"We continue to believe we will have at least the supply to meet the full-year revenue outlook we announced in July, which was $4.5 billion higher than our January expectations," he wrote in a post entitled "Supply Update"on the chipmaker's website Friday.
Intel shares closed up 3.1 percent Friday.
On July 26, Intel gave 2018 full year non-GAAP sales guidance of $69.5 billion.
Swan said the company is "making progress" in its transition to the 10-nanometer manufacturing process and expects "volume production" in 2019.
Intel interim CEO says chipmaker has supply to meet its full-year revenue outlook
There is a factual shortage already so Swans words are just for the stage. Pricing of Intel i-series chips in the Netherlands accross all retail have jumped a massive 30% on average the last three weeks. The 8700 even jumped a staggering 50% on both lowest and average offerings: Intel Core i7-8700 Boxed - Prijzen - Tweakers
Its not a coincidence this coincides with the launch of the iphone XS/XS+ and the XR also in volume production. The modems must have been in volume production for a few months now and inventory of chips at normal prices has probably dried up by now.
It's pretty clear to me Intel was banking on Apple's modem deal to fill excess 14nm capacity once the CPU's would start moving to 10nm.
Since this hasnt happened, and Intel probably didnt secure capacity for the Intel modem at TSMC at 10nm (which would be comparable enough to their own 14nm process to be acceptable for Apple) they are now faced with both their own CPU's still on 14nm, some chipsets now also on 14nm, and the massive amount of iphone modems they are contractually obligated to deliver on the same node.
What baffles the mind is it seems they are only acting on this shortage now, while they could have seen this coming for at least 6 months.
Its not a coincidence this coincides with the launch of the iphone XS/XS+ and the XR also in volume production. The modems must have been in volume production for a few months now and inventory of chips at normal prices has probably dried up by now.
It's pretty clear to me Intel was banking on Apple's modem deal to fill excess 14nm capacity once the CPU's would start moving to 10nm.
Since this hasnt happened, and Intel probably didnt secure capacity for the Intel modem at TSMC at 10nm (which would be comparable enough to their own 14nm process to be acceptable for Apple) they are now faced with both their own CPU's still on 14nm, some chipsets now also on 14nm, and the massive amount of iphone modems they are contractually obligated to deliver on the same node.
What baffles the mind is it seems they are only acting on this shortage now, while they could have seen this coming for at least 6 months.
Last edited:
Also, its not just the low-end, its also not just consumer CPU's, its not even just the smaller OEM's experiencing shortages
My new Zbook for work ordered about a month ago is being pushed back further and further with no clear delivery date in sight. This is the top of the line mobile Core i7 from the largest OEM. Bob Swans words have been disproven in less then two months. Something is seriously going wrong at Intel right now and its going to cost them customers.
Consumer demand for Core is reportedly plummeting in Europe due to the massive price hikes according to a large German retailer.
My new Zbook for work ordered about a month ago is being pushed back further and further with no clear delivery date in sight. This is the top of the line mobile Core i7 from the largest OEM. Bob Swans words have been disproven in less then two months. Something is seriously going wrong at Intel right now and its going to cost them customers.
Consumer demand for Core is reportedly plummeting in Europe due to the massive price hikes according to a large German retailer.
Last edited:
Another example of a poorly managed company. The thing with ramping yield is that sometimes the tweaks help but sometimes they hurt. I'm guessing that Intel was overly optimistic on 10nm yield ramp this year and held off on 14nm expansion as long as possible.
I had the impression that Intel was manufacturing their modems at TSMC. So are you saying that they are making them in their own factories or that they were counting on moving the production from TSMC to their own 14nm once they moved most of their CPUs to 10nm?
hist78
Well-known member
The cause of the supply shortage might be due to the market fluctuation, inaccurate forecast, poor management, tight financial strength, limited capex, profitability consideration, resource constraints, and technical difficulties. At the same time Intel might be hedging their future with multiple product lines, one manufactured in-house and one made by TSMC. The current Intel/TSMC outsourcing deal might be used as a preparation just in case future Intel's manufacturing capability (including 10nm) can't satisfy the market demand.
After several years of delay, Intel is running out of reasons about why Intel's 10nm hasn't come to the market. In case of any additional and serious 10nm difficulties, Intel's leadership will probably choose to protect Intel's revenue and profit over must-manufacture-100%-in-house ego. I think for those Intel's clients, such as HP, DELL, Leveno, Asus, Acer, or even Apple, they don't really care who are "making" Intel's chips. They will be very happy as long as Intel offers them products with good performance/features, great price, and abundant supply.
After several years of delay, Intel is running out of reasons about why Intel's 10nm hasn't come to the market. In case of any additional and serious 10nm difficulties, Intel's leadership will probably choose to protect Intel's revenue and profit over must-manufacture-100%-in-house ego. I think for those Intel's clients, such as HP, DELL, Leveno, Asus, Acer, or even Apple, they don't really care who are "making" Intel's chips. They will be very happy as long as Intel offers them products with good performance/features, great price, and abundant supply.
Last edited:
Intel customers might not care who fabs Intel chips, but Intel sure does -- the core principle of the whole company for the last 20 years has been "our CPUs are the best in the world because we also have the world's best process technology".
Admitting that this is no longer true would not only be one of the biggest losses of face ever for a management team, it would signal the end of Intel's fabs as volumes of existing 14nm products fall away (even assuming 10nm, or "12nm" or whatever it's called finally makes it to market) and new ones all go to TSMC.
Admitting that this is no longer true would not only be one of the biggest losses of face ever for a management team, it would signal the end of Intel's fabs as volumes of existing 14nm products fall away (even assuming 10nm, or "12nm" or whatever it's called finally makes it to market) and new ones all go to TSMC.
The Intel modems were originally done in TSMC 28nm by other companies that were acquired by Intel. I believe the current Intel modem chips, the ones Apple are using, are done on internal 14nm.
hist78
Well-known member
Intel is not choosing between two options: 100% made-in-house or 10% outsourcing to TSMC. They can use either one or both with various allocations depending on several factors:
1. Product lines.
2. Cost, profit margin, and volume.
2. Time to market and resource availability.
3. Scale of economy.
4. Yield and various technical aspects.
5. Customers' demands.
6. Market and technology opportunities.
7. Corporate image and ego.
Personally, I think the #7 might be a major consideration for many people especially those who are involved in engineering and manufacturing. But to the Intel's senior leadership, they may have different perspectives.
I can imagine the hard choice Intel is facing today. For example, Intel won the new iPhone XS and XS Max modem supply contract. It's really a good news for Intel but at what kind of cost and consequence? Apple sold 215 million units of iPhone in 2017. So we can assume Intel will provide similar amount of modems to Apple in 2018 and use Intel's in-house 14nm manufacturing resources.
Think about there're about 262 million units of PC sold in 2017 and Intel provided more than 90% of CPU for them. 200 million units of Intel made iPhone modems is not a small thing for Intel, even considering a modem chip takes smaller footprint on a wafer.
According to IHS Markit, Intel takes in about $17 for each of those iPhone modems. In Intel's product offerings, $17 a piece is probably in the low end category. Is it really a perfect arrangement for Intel while DELL, HP, Lenovo, Asus, and Acer are all yelling at Intel for inadequate supply and lost opportunities? And the only reason Intel can offer is that they already maxed out its 14nm manufacturing capacity (and it's because Apple)?
Last edited:
This is what I heard as well:
Intel jumps on report it could speed up production of its 10-nanometer chips (INTL) | Markets Insider
"Intel's second-half production levels suggest upside to analyst revenue estimates for the fourth quarter and first quarter of 2019," Steve Mullane, analyst BlueFin Research Partners wrote in a note sent out to clients on Tuesday, according to Bloomberg. The note said suppliers think production of the processors could be pulled forward from the June 2019 timeline by four to six weeks.
Intel jumps on report it could speed up production of its 10-nanometer chips (INTL) | Markets Insider
"Intel's second-half production levels suggest upside to analyst revenue estimates for the fourth quarter and first quarter of 2019," Steve Mullane, analyst BlueFin Research Partners wrote in a note sent out to clients on Tuesday, according to Bloomberg. The note said suppliers think production of the processors could be pulled forward from the June 2019 timeline by four to six weeks.
Looks too much like following Globalfoundries.
But which 10nm chips exactly? The ones Intel desperately want 10nm for are the big high-margin server CPUs starting with IceLake, which has now been pushed out to mid-2020, and the gap filled in the meantime with yet another 14nm refresh which will be completely uncompetitive with AMD Rome on both power, speed and price. TSMC 7nm has been running products since the start of this year and is ahead of their planned yield curve, so AMD should be able to hammer Intel in this market for at least 18 months -- assuming they can get enough wafers after Apple have slurped most of them up.
Another paper launch like Intel's existing pathetic 2-core GPU-free offering won't mean anything except to dumb investors and analysts. My guess is that their first real 10nm CPUs in mid-2019 will be low-power CPU/GPU for laptops (to keep chip size down and yield up), then bigger desktop chips at the end of 2019, then the massive server chips in mid 2020 -- this will let them follow the yield curve down on their 12nm (sorry, "improved 10nm") process.
All guesswork, but given the changes Intel are having to make to both process and chip designs I'd be surprised if they can deliver 10nm products faster than this.
Last edited:
Intel had an open letter almost a week ago: Supply Update | Intel Newsroom
$15 billion capex, including $1 billion for 14nm capacity expansion. 10nm yields reported as improving.
$15 billion capex, including $1 billion for 14nm capacity expansion. 10nm yields reported as improving.
Intel has lost credibility so all their PR talk on 10nm HVM is just useless. imo the real test is volume ramp of ICL in 2019. I think Intel is still not out of the woods completely on 10nm problems. My expectation is the 10nm chips launch in Q4 2019 with low power tablet chips ICL-Y followed by ICL-S desktop and ICL-H notebook CPUs in mid 2020 followed by ICL-SP for server chips in late 2020. We only need to look back to 14nm to see that it took 18 months from the first 14nm chips to server chips. Intel now wants to say that they will have shrunk that delay to 6 months. I am calling it a bluff. The fact that Cooper Lake exists is a dead giveaway that Intel ICL-SP is not a 2020 product. Cascade Lake is slated for a Q1 2019 launch and CooperLake for Q1 2020 atleast for high volume and not for just early ship to Google. This puts ICL-SP as a early 2021 product. Meanwhile TSMC N7 HPC is going to ramp first with Vega 7nm in late 2018 and then with AMD Rome by mid 2019 and the N7 yield learning is going to help N7 HPC customers too. Rome is expected to deliver 2x perf/watt of first gen EPYC with a 250w TDP. The performance of Rome looks to be outstanding and I doubt Intel will be able to match Rome before they move to a true chiplet based design with Sapphire Rapids and EMIB. Lisa Su said Zen 2 improves on Zen in multiple dimensions and given that Moore's law is now dependent on advanced packaging to deliver big generational improvements I think Rome is a 8 CPU die + 1 I/O die design on silicon interposer using TSMC's CoWOS. AMD's CES 2019 keynote cannot get here any sooner. I am excited that for once Intel will be outclassed on both manufacturing and CPU and system architecture at the same time.
Last edited:
They would say that. Reading between the lines...
"We're making progress with 10nm. Yields are improving and we continue to expect volume production in 2019"
If they weren't making progress and improving yields (from terrible) something would be wrong. Volume production of what, and when in 2019?
"We are taking a customer-first approach. We're working with your teams to align demand with available supply"
We have to supply Apple with shedloads of 14nm modems or they'll sue us out of business, everyone else get to to the back of the queue. BTW, an i7-8700K will now cost you 50% more than a month ago because we can't make enough CPUs to keep up with demand, because 10nm isn't ready as planned. Suck it up, we don't care. we're raking it in...
Yeah, that statement itself was kind of bland. Usually, it is preferred for the company to say, we have identified these problems and addressed them already. Seems it is not the case yet.
Last edited:
Did Steve's notes contain any specifics as to WHAT was sped up? Is Cobalt or SAQP yielding better? Why? Did they relax their pitches to 40nm? What exactly changed? That is more believable than some supplier rumors. Separately, isn't the new $1B capex for 14nm a red flag enough? When can that investment result in chips; next summer? Isn't that close to their supposed 10nm ramp? If 10nm looks like a big deal that investment makes no sense. It only looks logical if the are using it as insurance against a poor 10nm ramp at the same time.