[DigiTimes] TSMC 10 nm trial production in 2015, mass production in 2016

[Fjodor2001]

Unsustainable? Even with their billions of losses in mobile, Intel is still a 65% gross margin company. By the end of 2015, Intel will have its mobile business up to snuff from the entry smartphone to the high-end tablet market, with both Core and Atom, compatible with both Android and Windows. Their huge transistor lead (inclusive cost per transistor lead) will do the rest.

Yes, theoretically the rest of Intel could fund the losses in mobile. But if it goes on eventually the shareholders will say stop. Because it's simply bad business, and if they ditch the mobile section the profit will be higher.

Intel's market share of chips for smart phones and tablets is ridiculously small. ARM rules that segment. Intel has been saying for years that next year, things will change. But it never does. Intel's process tech lead (which is shrinking BTW), does not matter if they're not selling many chips in that segment anyway.

 

[Arachnotronic]

Yes, theoretically the rest of Intel could fund the losses in mobile. But if it goes on eventually the shareholders will say stop. Because it's simply bad business, and if they ditch the mobile section the profit will be higher.

Intel's market share of chips for smart phones and tablets is ridiculously small. ARM rules that segment. Intel has been saying for years that next year, things will change. But it never does. Intel's process tech lead (which is shrinking BTW), does not matter if they're not selling many chips in that segment anyway.

Even with the large mobile losses, Intel still makes substantially more in net profit than do Qualcomm and TSMC individually.

Moreover, without a credible mobile strategy, Intel will become irrelevant in the microprocessor space over the next 10-15 years. Intel management is making the correct decision to invest in mobile now while the getting is good in its other businesses.

 

[Fjodor2001]

Even with the large mobile losses, Intel still makes substantially more in net profit than do Qualcomm and TSMC individually.

Qualcomm net income 2013 (fiscal): $6.85 billion
Intel net income 2013: $9.6 billion

So the difference is not that big. And then I did not even add TSMC's part of the profit from the devices sold.

Also, note that those are the total numbers for the companies, and Intel covers a much wider range of segments than Qualcomm. So how do you think the profit numbers look in the mobile phone / tablet chip segment for Qualcomm vs Intel?

Moreover, without a credible mobile strategy, Intel will become irrelevant in the microprocessor space over the next 10-15 years.

I think that's a bit too dramatic. People will still be buying desktops, servers and laptops 10-15 years from now. But the scenery and volumes may change.

Also, signs are that even tablet sales are not increasing as much anymore. So eventually the tablet / smart phone segment will be "good enough" and saturated too. It's no safe haven for eternity either.

Intel management is making the correct decision to invest in mobile now while the getting is good in its other businesses.

Yes, it's a good decision if they eventually will become a major player in that segment. But so far they have not succeeded, despite trying for years. At some point you have to agree it makes sense for them to give up, right? Having a business segment that makes billions of dollars in losses every year is not sustainable in the long run from a shareholder perspective.

 

[rgallant]

If you look at it like that, chances are it's Intel that will be priced out of the market. Eventually there could be just one foundry having all the world's collected semiconductor volumes outside Intel, competing against Intel. Guess who will have the largest volumes?

As for volumes on mobile, Intel better start signing some major contracts. So far they have no really large contracts, and their mobile business is losing billions of dollars. It's unsustainable. I think it's more likely that Qualcomm, Samsung and Apple will continue to rule that market segment. Intel will still be king of the desktop, laptop and server segments though, which really is no bad thing at all.

just a noob but I think intel could build better arm chips two years before anyone, also IF they wanted to , do they have the fabs yes or no ?could they ship apple chips on 14nm at the end of the year ? if they had the contracts .

 

[witeken]

Qualcomm net income 2013 (fiscal): $6.85 billion
Intel net income 2013: $9.6 billion

So the difference is not that big. And then I did not even add TSMC's part of the profit from the devices sold.

Also, note that those are the total numbers for the companies, and Intel covers a much wider range of segments than Qualcomm. So how do you think the profit numbers look in the mobile phone / tablet chip segment for Qualcomm vs Intel?

That's his whole point. In the numbers you posted, Intel has still almost 50% more income, despite the billions of mobile investments and contra-revenue.

I think that's a bit too dramatic. People will still be buying desktops, servers and laptops 10-15 years from now. But the scenery and volumes may change.

I think Brian has a good explanation for going into mobile. There's a lot of innovation going on in the mobile space. And despite being the biggest semiconductor company, Intel is not represented in that market.

Yes, it's a good decision if they eventually will become a major player in that segment. But so far they have not succeeded, despite trying for years. At some point you have to agree it makes sense for them to give up, right? Having a business segment that makes billions of dollars in losses every year is not sustainable in the long run from a shareholder perspective.

Intel hasn't been trying for years. Saltwell was a joke and Intel didn't even have a leading edge SoC manufacturing process.

Their real competitive architecture was scheduled to launch on their 22nm SoC process in 2013 which it did. And instead of waiting another 5-6 years, Atom has been promoted to Intel's Tick-Tock cadence, so Airmont will be on the newest node from the start.

 

[mrmt]

I think that's a bit too dramatic. People will still be buying desktops, servers and laptops 10-15 years from now. But the scenery and volumes may change.

Also, signs are that even tablet sales are not increasing as much anymore. So eventually the tablet / smart phone segment will be "good enough" and saturated too. It's no safe haven for eternity either.

That's not a bit too dramatic. Every IC generation costs more to develop and every foundry nodes needs more volume to pay off. When Intel17 says that Intel will become irrelevant if it doesn't have a mobile strategy, he is not saying that the desktop, server and laptop will end, but that Intel won't be able to sustain their current business model (high volume, bleeding edge node, bleeding edge IP) with laptop, desktop and server. They basically become the next AMD but this time with fabs.

Intel basically lost the chance to dominate the mobile market as we know today. What they are trying to do is to establish a solid foothold there, because the next trend is towards wearables and embedded, which will leverage a lot on mobile.

What if Intel doesn't get this solid foothold? Then their business model as a vertical, integrated IDM will have to change, and that means goodbye to their top notch ROI rates.

 

[krumme]

That's not a bit too dramatic. Every IC generation costs more to develop and every foundry nodes needs more volume to pay off. When Intel17 says that Intel will become irrelevant if it doesn't have a mobile strategy, he is not saying that the desktop, server and laptop will end, but that Intel won't be able to sustain their current business model (high volume, bleeding edge node, bleeding edge IP) with laptop, desktop and server. They basically become the next AMD but this time with fabs.

Intel basically lost the chance to dominate the mobile market as we know today. What they are trying to do is to establish a solid foothold there, because the next trend is towards wearables and embedded, which will leverage a lot on mobile.

What if Intel doesn't get this solid foothold? Then their business model as a vertical, integrated IDM will have to change, and that means goodbye to their top notch ROI rates.

I just bought a samsung gen3 10.1 pad for my 5 year old with clovertrail+ (an awfull processor in real use btw app loading times is bogus - but what does she care? ). It was cheap as dirt. no cover needed.
In 1.5 year i can get a atomgen2 whatever or a15 a7 octo for dirt cheap.
I still fail to see where the money is.
Have been saying this for 2 years and it just gets worse, and shows in contra revenue and losses imho.
Still plenty of money in x86 servers.
Boring but reality.

 

[Arachnotronic]

I just bought a samsung gen3 10.1 pad for my 5 year old with clovertrail+ (an awfull processor in real use btw app loading times is bogus - but what does she care? ). It was cheap as dirt. no cover needed.
In 1.5 year i can get a atomgen2 whatever or a15 a7 octo for dirt cheap.
I still fail to see where the money is.
Have been saying this for 2 years and it just gets worse, and shows in contra revenue and losses imho.
Still plenty of money in x86 servers.
Boring but reality.

Look at Qualcomm's QCT financials. The average selling price per MSM is in the low $20 range and yet Qualcomm's chip division makes nearly as much money as Intel's server chip division (though on a higher revenue base).

There's plenty of money to be made on dirt cheap chips if you're selling them at scale.

 

[Ajay]

Qualcomm net income 2013 (fiscal): $6.85 billion
Intel net income 2013: $9.6 billion

So the difference is not that big. And then I did not even add TSMC's part of the profit from the devices sold.

Also, note that those are the total numbers for the companies, and Intel covers a much wider range of segments than Qualcomm. So how do you think the profit numbers look in the mobile phone / tablet chip segment for Qualcomm vs Intel?

Well, QCOM won't have any problem meeting the challenge of escalating design cost! That makes three companies for sure (apple & intel). I'm not sure about Samsung - since that depends on internal politics, though the new president of Samsung expects the LSI division to really up it's game (which it's already done visa vi process tech).

I guess Intel will really need to increase it's lead in process technology to hurt its competitors in the ultra-mobile space. This is going to be an interesting battle!

 

[TrulyUncouth]

I just bought a samsung gen3 10.1 pad for my 5 year old with clovertrail+ (an awfull processor in real use btw app loading times is bogus - but what does she care? ). It was cheap as dirt. no cover needed.
In 1.5 year i can get a atomgen2 whatever or a15 a7 octo for dirt cheap.
I still fail to see where the money is.
Have been saying this for 2 years and it just gets worse, and shows in contra revenue and losses imho.
Still plenty of money in x86 servers.
Boring but reality.

While I see your point, I just got an asus T100 and it has gotten me excited for cherry trail like no other. I am playing real PC games and absolutely loving it.

I think there is definitely room for intel in the Win8 tablet space. In android I think they are going into a pretty much already commodotized market and will end up having a very small market share.

I can only hope they stick it out work on getting huge performance gains so I can buy a better windows tablet each year.

 

[Homeles]

While I see your point, I just got an asus T100 and it has gotten me excited for cherry trail like no other. I am playing real PC games and absolutely loving it.

I think there is definitely room for intel in the Win8 tablet space. In android I think they are going into a pretty much already commodotized market and will end up having a very small market share.

I can only hope they stick it out work on getting huge performance gains so I can buy a better windows tablet each year.

There's room for Intel to compete in the Android space as well. There's no reason why they can't get their costs down, and that's exactly what they're doing.

 

[krumme]

Intel gets their cost down. Adapts capex and solutions. But while they are doing so the user benefit of faster cpu in mobile is going the same way as desktop and laptops. Its shows in crazy stuff like dual a9 beeing replaced by quad a7 in cheap tablets. How much is a a7 ? Isnt it like 0.5mm2 ! Lol. Next move is probably octo core miniscule arm a7 replacement at 1.2GHz made at 20nm where each core is 0.25mm2

 

[TrulyUncouth]

There's room for Intel to compete in the Android space as well. There's no reason why they can't get their costs down, and that's exactly what they're doing.

My point is more along the lines with what krumme just posted- I think they only can differentiate using their x86 abilities. I use to be a constant phone upgrader going through 6 phones in 5 years. Now I am on the same phone for 2 years and due for an upgrade but I am not even looking at flagships because performance has kind of stopped mattering. I notice no lag in my browsing on my phone, no games have lagged or look bad enough for me to notice, and my battery life is enough to get me through a day.

Basically, I fear in the mobile android space SOC's have become/are quickly becoming commodities. Thats why my opinion is that Intel should do everything possible to get win8 out there. They have a de facto monopoly on the space and if people get hooked on wintablets like I am quickly becoming then intel can benefit hugely.

Perhaps if they need higher volume to fill their foundry pipes they can open up foundry space to non-competitive companies? Intel will not be selling a $4 SOC like mediatek or allwinner does for a very long time. Let them eat up your extra capacity if you can't move enough chips.

 

[Khato]

Intel will not be selling a $4 SOC like mediatek or allwinner does for a very long time.

That's what they signed up Rockchip for

 

[witeken]

Performance will stay a marketing argument. Companies will keep buying the best SoC they can get for their money. If that happens to be an SoC from Intel, they will move on, away from Qualcomm.

 

[Fjodor2001]

Performance will stay a marketing argument. Companies will keep buying the best SoC they can get for their money. If that happens to be an SoC from Intel, they will move on, away from Qualcomm.

The problem for Intel is that Android effectively is ARM. Yes, you can build Android for x86 too, and Android Java is CPU arch independent. But Android APK:s can have native ARM code in them anyway, and several do.

Also, note that it's not just CPU performance that matters. It's also mobile radio performance, GPU performance, power consumption, price, etc.

Finally, you also need customers to sign up for buying your chips. I'm not so sure Samsung and Apple which dominate this market want Intel to enter it. Let's assume some smaller competitor signs up for using Intel chips, but then it has to be vastly superior to the alternatives for customers to consider it to be worth considering switching to. And I don't think Intel is there actually.

 

[mrmt]

Intel gets their cost down. Adapts capex and solutions. But while they are doing so the user benefit of faster cpu in mobile is going the same way as desktop and laptops. Its shows in crazy stuff like dual a9 beeing replaced by quad a7 in cheap tablets. How much is a a7 ? Isnt it like 0.5mm2 ! Lol. Next move is probably octo core miniscule arm a7 replacement at 1.2GHz made at 20nm where each core is 0.25mm2

Don't jump on Hans de Vries bandwagon that quickly. Core count alone doesn't mean anything on the SoC context. ARM SoCs with much smaller cores are getting similar SoC size to Atom or even Core SKUs. Will this 8C A7 have much smaller manufacturing costs than bigger core designs, or is it just because of the simple design that makes everything else easier to validate?

 

[Idontcare]

If you look at it like that, chances are it's Intel that will be priced out of the market. Eventually there could be just one foundry having all the world's collected semiconductor volumes outside Intel, competing against Intel. Guess who will have the largest volumes?

As for volumes on mobile, Intel better start signing some major contracts. So far they have no really large contracts, and their mobile business is losing billions of dollars. It's unsustainable. I think it's more likely that Qualcomm, Samsung and Apple will continue to rule that market segment. Intel will still be king of the desktop, laptop and server segments though, which really is no bad thing at all.

The tipping point will come when we see which company is the first to HVM with a 450mm fab.

Will it be pure-play foundry (TSMC/UMC/GF) will it be IDM (Intel, Samsung, Micron, Toshiba, etc).

The reason I say tipping point is because no company is going to invest $20B into a HVM 450mm fab without having considerable confidence that they will have the wafer volumes 6 yrs out to support such an investment decision.

When we see the 450mm fab's getting tooled up, that's when you know which businesses (be they foundry or IDM) has the volumes (2-4 yrs down the road) to support their respective business models for the next decade.

 

[TrulyUncouth]

That's what they signed up Rockchip for

Fair enough, if it gets more chips into intel fabs so they can share process costs- but I thought rockchip was going to be using intel IP but producing at some place like TSMC?

Performance will stay a marketing argument. Companies will keep buying the best SoC they can get for their money. If that happens to be an SoC from Intel, they will move on, away from Qualcomm.

My point is that there are a ton of customers out there like me who always bought the highest performance but don't need to anymore. We've reached the "good enough" level like we did with computers where people will be replacing less often and buying cheaper when they do.

Another piece of anecdotal evidence but for instance me and my brother's family share an ATT plan and now we have moved to a plan where we don't get subsidized phones. Use to be there was an incentive to go for the most expensive phone so you didn't "waste" your subsidy. Now I pay full price for the phone so if I can't tell a difference in performance there is no incentive to go for top of the line. Also this increases the drive to buy used since older phones are more usable than the older phones of yester-year.

Between all these forces its no surprise Samsung has seen dips in flagship sales and mobile profitability and the entire market seems to be in a rough patch. I just think intel would be foolish to go for the performance segment since it looks like that segment will get smaller as we go forward.

I say go whole hog on windows, and try their damnedest to talk microsoft into some unholy win8/9 phone that can be an all-in-one and run full windows stuff.

 

[Madpacket]

Fair enough, if it gets more chips into intel fabs so they can share process costs- but I thought rockchip was going to be using intel IP but producing at some place like TSMC?



My point is that there are a ton of customers out there like me who always bought the highest performance but don't need to anymore. We've reached the "good enough" level like we did with computers where people will be replacing less often and buying cheaper when they do.

Another piece of anecdotal evidence but for instance me and my brother's family share an ATT plan and now we have moved to a plan where we don't get subsidized phones. Use to be there was an incentive to go for the most expensive phone so you didn't "waste" your subsidy. Now I pay full price for the phone so if I can't tell a difference in performance there is no incentive to go for top of the line. Also this increases the drive to buy used since older phones are more usable than the older phones of yester-year.

Between all these forces its no surprise Samsung has seen dips in flagship sales and mobile profitability and the entire market seems to be in a rough patch. I just think intel would be foolish to go for the performance segment since it looks like that segment will get smaller as we go forward.

I say go whole hog on windows, and try their damnedest to talk microsoft into some unholy win8/9 phone that can be an all-in-one and run full windows stuff.

+1. Have to keep in mind subsidized phones is mostly a north american phenomenon as well. We've reached fast enough now with mobile SoC's but battery life and more flexible form factors will be driving the need for improved SoC's from here on out, increase in performance is now a side benefit.

I said this years ago but Intel shouldn't bother with competing with ARM on Android as x86 cannot offer a big enough advantage there. Unfortunately they're stuck due to Microsoft's failure in mobile where their CPU'S could make a difference and the longer they lag behind the more irrelevant the platform becomes.

If both Microsoft and Intel don't turn things around soon it'll be pretty interesting for them.

 

[mikk]

TSMC 10nm Volume production is planned for 2017 they told recently. With their usual delays we should add 3 years to this timeframe. 16nm volume is planned for 2016.

Currently, we are qualifying customer info products with 16 nanometer technology and it will be ready for volume ramp next year, 2016.

Our 10 nanometer technology development is progressing and our qualification schedule at the end of 2015 [..] We are now working with customers for their product tape outs. We expect this volume production in 2017.

http://www.computerbase.de/2015-01/tsmc-16-nm-in-grossserie-erst-2016-10-nm-bereits-2017/




While Intel claims their 10nm is ready for market at the end of 2015. I guess this basically means they could build first ES silicon on 10nm, until we can buy first products it will take 1 year I guess. If they suffer from the same issues like 14nm much longer than that. At least they learned from their 14nm issues.

We are timing on 10-nanometers. We are not going to come out with -- we’ll be introducing a 10-nanometer to the marketplace in general, probably until the end of this year. So we will give, as we go through the year, probably by the investor meeting in November, we’ll give you an outlook on how and what timing is for 10-nanometers.

http://www.computerbase.de/2015-01/intels-10-nm-fertigung-soll-ende-2015-marktreif-sein/

 

[witeken]

It seems I was right with some thread that I created but got locked because the title would have been "wrong", but now it seems there indeed won't much if any 14nm products in 2015, except the incidental iPhone or Galaxy Note. In fact, if TSMC only ramps 20nm+FinFET in 2016, that would be a massive 2 years after 20nm without FinFET or a year later than promised. You can already take the 2017 HVM of their 10nm node with a grain of salt. Also because it isn't comparable at all (at least in terms of performance and power; transistor feature size still to be determined) to Intel's 10nm, so my educated guess (same as ShintaiDK) of EOY2018 for their 2nd node FinFET might still be true (unless EUV miraculously helps them with yields and stuff).

@Mikk: That quote seems wrong. Intel is this year paying the cost of the 10nm factory start-up (primarily in last Q4 and H1), to ramp probably somewhere around January '16. Mark Bohr said at IEDM that 10nm wouldn't have the same issues as 14nm in terms of yields, while Intel has reiterated their 2 year Tick-Tock model a TON in 2014 and this most recent earnings call. Intel currently seems to be the only company that is able to pull this (almost) 2 year cadence off.

 

[krumme]

Ss process with gf on the sideline seems to win the 14nm and A9 race.
http://www.digitimes.com/news/a20150115PD212.html
Lots of capacity.
Probably steems fra stealing sa stuff. Lol.

 

[Idontcare]

Ss process with gf on the sideline seems to win the 14nm and A9 race.
http://www.digitimes.com/news/a20150115PD212.html
Lots of capacity.
Probably steems fra stealing sa stuff. Lol.

That is indeed interesting.

 

[Vesku]

Good news for AMD, being tied to GF looks like they will finally get in ahead of Nvidia on finfet. Although going of the digitimes numbers they will now play second fiddle to Apple for wafer priority beyond the first ~10,000 a month. That's got to sting after being made to keep GF 32nm and now 28nm going.

In regards to TSMC 10nm, does the reported delay in 16nm likely have any impact on their 10nm development?