Intel Q4 results. 13.8B$ revenue. 52.7B$ year.

[mrmt]

Pretty sure AMD's Q3 was up more compared to Intels YoY, wasn't it?

As usual, you are wrong. Intel PC client group revenues was down 3.5% YoY while AMD Computing solutions was down 15% YoY. But keep trying, I promise I'll try to keep track of the "gems" you drop here on the forum.

 

[AtenRa]

You also need SoCs and product wins. If only a few devices get a 20nm SoC in 2014, Intel would have H2 for itself at 14nm vs. 28nm. For example, Nvidia also wants 20nm for its Maxwell GPUs.

TSMC said 20nm in 2014 will be 10% of its total 2014 wafer sales. And since ARM SoCs are small dies at or less than 100mm2 then we are talking about a lot of 20nm products in 2014.

http://www.digitimes.com/news/a20140117VL200.html

Also at the investors meeting, TSMC co-CEO CC Wei disclosed that shipments of 20nm process technology will reach about 10% of the company's total wafer sales in 2014. The foundry has moved the node to volume production in the first quarter.

 

[witeken]

It would be available even earlier:
http://www.newelectronics.co.uk/electronics-news/tsmc-ramps-up-16nm-finfet-production/58416/

At which yield cost? Starting production of 2 new process nodes in 1 years seems too good to be true.

 

[bullzz]

@atenra - I dont understand what you mean by "2014 has 12 months, also Intel 14nm products will be released in H2 and most probable in Q4 2014."

companies put out a product once every 12 months. and 2014 products qcomm and nvidia (intel's competetors) are 28nm. now apple and samsung can put out 20nm products but they are in a league of their own

 

[AtenRa]

At which yield cost? Starting production of 2 new process nodes in 1 years seems too good to be true.

http://hexus.net/business/news/components/65057-tsmc-says-16nm-enter-volume-production-year/

Currently, the 16FinFET SRAM yield is already close to that of the 20SoC process." Liu added that there are over 20 16FinFET products scheduled for a 2014 tape out including SoCs, CPUs and graphics chips.

 

[mrmt]

This period has been lasting for at least 4 years: http://seekingalpha.com/article/1951261-intels-mobile-money-losers
Heavy losses have been lasting for two years and given the contra-revenue Intel said it would use, I'm not sure the division will be profitable this year (unless contra-revenue expenses are accounted for somewhere else?).

What you are seeing today is the work of 4 years inside the R&D pipeline, and that costs a lot of money that Intel had to pay in R&D. Intel changed the Atom business model, from a small, low volume product to a bleeding edge product. Intel had to pay the costs of the development of a bleeding edge line up while having only lagging edge products to foot the bill, that is the cause of the losses you are seeing here.

But in 2014 the product they have already developed on this business model is on the market. They need to get the right volumes here, they need to get the scale otherwise the entire R&D pipeline and the viability of their foundry business is toast. This is why BK is so keen to get the right volumes on Atom, and this is why he must do that until Q214.

That said, division will probably not be profitable, or at least not too profitable. Intel is saying that they will buy market share, and that means either lower gross margins or higher OPEX. Also Intel executive team is saying that they need to grow that business on volume, as it doesn't have the required footprint to be sustainable.

Intel will have a nice advantage when they roll out 14nm, but they won't be able to capitalize on it if Bay Trail doesn't give Cherry Trail the right starting point. This is what we should be watching at Intel right now, not mixing financial metrics and pondering whether bay trail is a failure or not.

 

[AtenRa]

@atenra - I dont understand what you mean by "2014 has 12 months, also Intel 14nm products will be released in H2 and most probable in Q4 2014."

companies put out a product once every 12 months. and 2014 products qcomm and nvidia (intel's competetors) are 28nm. now apple and samsung can put out 20nm products but they are in a league of their own

NV is not the only company making ARM products. I dont know if NV will make a 20nm product in 2014 but others will do.

 

[Exophase]

At which yield cost? Starting production of 2 new process nodes in 1 years seems too good to be true.

Because it's not like a traditional node advancement. I didn't see anyone bat an eyelid when Intel introduced both to 22nm geometry and finFETs in the same node transition and delivered it more or less on time..

 

[Homeles]

Because it's not like a traditional node advancement. I didn't see anyone bat an eyelid when Intel introduced both to 22nm geometry and finFETs in the same node transition and delivered it more or less on time..

Right. While it's probably very aggressive from an economic standpoint, the technical aspect of delivering 16/14nm FinFETs should be relatively sound.

 

[witeken]

Because it's not like a traditional node advancement. I didn't see anyone bat an eyelid when Intel introduced both to 22nm geometry and finFETs in the same node transition and delivered it more or less on time..

Intel's HKMG was also one process node before TSMC's, for example (~3 to 4 years earlier).

 

[ShintaiDK]

TSMC and GloFos 14/16nm are 20nm with Finfets. Its really slowed down drasticly from those 2 companies. As we directly "enjoy" on the GPU side.

 

[Khato]

http://hexus.net/business/news/components/65057-tsmc-says-16nm-enter-volume-production-year/

Currently, the 16FinFET SRAM yield is already close to that of the 20SoC process." Liu added that there are over 20 16FinFET products scheduled for a 2014 tape out including SoCs, CPUs and graphics chips.

Good for them, they can make the simplest useful structure with good yields on their 16FinFET process. Now if they'd said that their 16FenFET logic yield is already close to that of the 20SoC process that'd be impressive and a good sign for it coming out on or ahead of schedule. Obtaining decent yields on SRAM though? Pretty inconsequential.

No, Intel's quite safe from an onslaught of 16FinFET products until well into 2015 at the earliest. And that's assuming that TSMC isn't hiding any dirty details that they hope to figure out in the interim.

 

[pablo87]

Decent #'s but capex far exceeded depreciation once again, this time by $1.3B. If revenues were growing that might make sense but they are not. There's potentially a tax benefit angle as well since depreciation for tax > depreciation for GAAP but that's a sidebar (thank you Intel for paying taxes - don't think AMD pays any and probably won't for a very long time).

 

[Exophase]

Intel's HKMG was also one process node before TSMC's, for example (~3 to 4 years earlier).

Yeah, and they still transitioned to that too roughly in parallel with a shrink, so what's your point exactly?

 

[pablo87]

Regarding Intel SOC strategy, I can tell you unequivocally that performance matters as my old Nexus 7 is almost unusable now that I have the new one. Apple is still the King not that the hardware is so good but the OS simplicity and ease of use >> (my son loves Android, I find it annoying).

With continued improvement in switching functionality on and off (or downclocking more aggressively) in smaller and smaller blocks more and more instantaneously (you tell me), bigger die might be a viable strategy which would greatly benefit from Intel's Fab lead and capacity.

Meanwhile not sure what prevents Intel from cross subsidizing SOC with Server (ASP's up 7%). Nobody believes OEM's won't take the price bait, do they? They have for 20 years...

 

[VirtualLarry]

They're not doing anything illegal.

Should be 2015, with 2014 comprising a significant recoup of their investment.

Selling product below cost, with a specific intent to gain market share and harm competitors, is called "dumping", and is illegal.

I guess the question is, is Intel "dumping" on the mobile market to gain market share? Are they selling the chips below their true cost (including R&D and marketing)?

 

[ShintaiDK]

Selling product below cost, with a specific intent to gain market share and harm competitors, is called "dumping", and is illegal.

I guess the question is, is Intel "dumping" on the mobile market to gain market share? Are they selling the chips below their true cost (including R&D and marketing)?

Its similar to this:
http://vr-zone.com/articles/microso...huawei-make-windows-phone-handsets/69558.html

Microsoft will pay Samsung, Sony and Huawei to launch one Windows Phone handset in 2014.
For Microsoft, these handsets come at a hefty price. Samsung is set to receive $1.2 billion, Sony $500 million and Huawei $600 million. The money is for designing and launching one Windows Phone handset this year. This cash influx would be a huge incentive for manufacturers like Sony and Huawei to launch Windows Phone handsets.

 

[witeken]

Yeah, and they still transitioned to that too roughly in parallel with a shrink, so what's your point exactly?

My point is that it isn't surprising that Intel, or actually any company, released Tri-Gate (or another new technology) together with a die shrink. It's an interesting move from TSMC and Global Foundries to do it else and introduce finFET in a separate node.

 

[ShintaiDK]

My point is that it isn't surprising that Intel, or actually any company, released Tri-Gate (or another new technology) together with a die shrink. It's an interesting move from TSMC and Global Foundries to do it else and introduce finFET in a separate node.

TSMC and GloFo is not ready with finfets. And they now gonna use it as a second stage before the actually reach 14nm that they again now have to call 10nm. We had a long 3 year dry period to get 20nm. Seems the next waiting time for an actual new node is 4 years.

 

[witeken]

Selling product below cost, with a specific intent to gain market share and harm competitors, is called "dumping", and is illegal.

I guess the question is, is Intel "dumping" on the mobile market to gain market share? Are they selling the chips below their true cost (including R&D and marketing)?

Did you see my response?

http://forums.anandtech.com/showpost.php?p=35965544&postcount=41

The BoM is still quite high apparently, so they're willing to pay the BoM difference themselves for partners who want to use Bay Trail. That isn't illegal (also Intel is open about this and I haven't seen any site claiming that this isn't legal).

 

[Torn Mind]

Selling product below cost, with a specific intent to gain market share and harm competitors, is called "dumping", and is illegal.

I guess the question is, is Intel "dumping" on the mobile market to gain market share? Are they selling the chips below their true cost (including R&D and marketing)?

"Dumping" is a term typically used in international trade contexts and typically involves the domestic country trying to stave off the various negative effects dumping on the domestic economy. The Antidumping Act of 1921 in one sort of law that deals with this. Unless Intel is selling its processors cheaper than the domestic country's own Intel or x86 processors(which obviously don't exist), Intel can't be said to be dumping.

 

[AtenRa]

Intel to reduce global workforce by five percent in 2014

Intel Corp plans to reduce its global workforce of 107,000 by about 5 percent this year as the chipmaker, struggling with falling personal-computer sales, shifts focus to faster-growing areas, a company spokesman said on Friday.
The announcement, equivalent to over 5,000 positions, comes a day after Intel posted a fourth-quarter earnings report that did little to dispel concerns about a slowing PC industry.
"This is part of aligning our human resources to meet business needs," spokesman Chris Kraeuter told Reuters on Friday.

 

[SiliconWars]

Not surprising, that's always been their first move when things got rough. Damn that's a lot of people though.

 

[OCGuy]

Smart business doesn't keep welfare positions unfortunately. Shedding a little fat is a good idea.

 

[Exophase]

We had a long 3 year dry period to get 20nm.

No we didn't. TSMC 20nm is entering mass productive about 27 months after 28nm. That's not three years. If I called the time between Intel's first 32nm and first 22nm product three years people would flip out.