Intel lost $1bn on mobile in Q3 '14

[Abwx]

Your "80%" estimate surely does not leave much -- if any -- room for COGS alone, nevermind R&D.

That leaves 200m/quarter, it s not unreasonable to assume that Intel mobile dpts RDs expense drain as much as AMD s whole RD, it s not like their mobile offering is not reusing IP from other CPUs, particularly the GPU wich is derived from the main CPU line.

As said if we are to follow the usual "specialists" estimations even secretaries and sweepers are millionaires at Intel...

 

[pw257008]

I'd think BT R&D is split between PC and mobile divisions as well

 

[Enigmoid]

Tell me where the money s going in that case, that s all good to contradict but all i m reading as answers is things like , " no, it s wrong", "this is not possible" and the likes, i told you, ignorance is used as a counter argument..

For sure, but still an anwser like the ones i quoted, since you like hints i d tell you that RD is mainly enginering ressources consuming, your PCs, or rather mainframes, do not think and all computing ressources are shared, thoses who works on BT use the same HPC servers as the crews dedicated to Skylake for instance.

Intel published their RD budget, it s 2.8bn for last quarter and this is for all divisions including the processes RDs expenses, yet we have people saying that BT and a few ICs dedicated RDs are dispatched 1bn/quarter, as much as AMD whole RD for a full year, AMD s engineers are apparently capable of working 48h/day according to the RD costs estimated by "accounting experts" that hang by there.

I'm not making a claim so burden of proof is not on me.

I'm saying that if you want to make a claim then support it with facts or alternatively offer a hypothesis and back that up with adequate and reasonable logic.

You made a statement without any qualifier (ie not "...likely 80%..."), then tried to attribute the spending completely to payroll.

 

[Abwx]

I'm not making a claim so burden of proof is not on me.

I'm saying that if you want to make a claim then support it with facts or alternatively offer a hypothesis and back that up with adequate and reasonable logic.

You made a statement without any qualifier (ie not "...likely 80%..."), then tried to attribute the spending completely to payroll.

And yet i see no estimation that could contradict my sayings, because if one start to do the calculations he will see that i m still too optimist when estimating the workforce cost, but still if you dont agree tell me how much cost an engineer knowing that all are not paid up to US standards.

So rather than perpetualy saying "no", "you have no proof", take a litlle more risks and give us your estimation of an average EE engineer gross salary, i m all ears but just dont come with numbers like 250K-500K/year, yet you ll have to get to those absurd amounts to try to handle the RD house card argument.

Edit : A hint to help, Intel s whole RD is 2.8bn/quarter.

I'd think BT R&D is split between PC and mobile divisions as well

It is part of the 2.8bn/quarter, according to some people here BT and generics mobile ICs eat at least 30% of Intel s whole RD...

 

[Idontcare]

...estimation of an average EE engineer gross salary, i m all ears but just dont come with numbers like 250K-500K/year, yet you ll have to get to those absurd amounts to try to handle the RD house card argument.

If Intel's average EE doesn't punch above $250K/annum in total compensation footprint (impact to Intel's bottom-line) then I'd be completely aghast.

In fact, I'd estimate the compensation footprint for an EE at Intel is going to run just slightly north of $300K.

General rule of thumb in the USA is that compensation footprint will be roughly double the annual base salary. A base salary of $150K for an EE is about bang-on for Intel EE's on average. Starting EE's (PhD mind you, as we are talking R&D here and not standard fab engineers for manufacturing or test) are probably looking at anywhere between $110K-$120K base salary.

So $220K-$240K compensation footprint once you've added up all the bonuses, 401K, taxation (including that lovely 12% SSN+Medicare employer tax), and that is just for your starting PhD EE.

There is a reason employers add headcount as slowly and as minimally as possible, they cost a fricken ton of money even though they all (myself included!) plead poverty at the drop of a hat

 

[CHADBOGA]

I don't believe Intel is directing their contra campaign against AMD, but I do believe AMD is caught in the crossfire between Intel and Mobile.

Yep, In Intel taking the fight to the ARMy, AMD will end up as collateral damage.

 

[Erenhardt]

In my country it is illegal to sell stuff cheaper than the cost to manufacture. Are there any similar laws elsewhere in the world?

 

[III-V]

In my country it is illegal to sell stuff cheaper than the cost to manufacture. Are there any similar laws elsewhere in the world?

For perhaps the 100th time, Intel is not selling less than it costs to manufacture. Good grief.

 

[ShintaiDK]

In my country it is illegal to sell stuff cheaper than the cost to manufacture. Are there any similar laws elsewhere in the world?

No its not. And you should know this just looking at argiculture.

 

[AtenRa]

For perhaps the 100th time, Intel is not selling less than it costs to manufacture. Good grief.

BayTrail die = ~100mm2

There are roughly 600 dies per 300mm wafer

At 80% yields that translates to ~480 dies per wafer

Let just say that each 22nm wafer cost $3000 today

$3000/480 = $6,25 per die

Even if wafer cost is at $2500 which i dont believe Intel's 22nm process on US Fabs is capable, but even then each die costs , just to manufacture the die is $5,2.

Now, we all know Intel sells BayTrail for tablets at lower than that after Contra Revenue is calculated.

 

[AtenRa]

Also,

Mobile and Communications Group Q3 2014
Net Revenue = 1M
Operating loss = 1043M

Lets say they spend 250M for Contra Revenue per Quarter

1043 - 250 = 793M

Mobile and Communications Group doesnt only sell ATOM SoCs, so lets assume they made 100M of revenue from the rest.

793-100 = 693M

Now, they sold 15M SoCs in Q3 2014

693M - 15M SoCs = ~$46

That means they have to sell 15M SoCs per quarter (60M per year) at $46 to come even.

Or

They have to sell 30M SoCs per Quarter (120M per Year) at ~$20 to come even.

Anyone here believes Intel's Mobile and Communications Group will ever come to be profitable ???

 

[Abwx]

No its not. And you should know this just looking at argiculture.

We live both in Europe and you know that it s forbidden at this level, agriculture is officialy subsided and is a special case that in no way extend to all the rest of the business, in the EU it apply to retailers only, although in France it s fordbidden for anybody.

 

[Phynaz]

In the mean time I'm still waiting for that citation.

 

[Idontcare]

Anyone here believes Intel's Mobile and Communications Group will ever come to be profitable ???

Does it need to? Is that the strategic objective for Intel?

Or are they OK with this being something that holds back their gross margins by 3% overall but keeps the competition at bay for another decade while they run their process node teams faster than TSMC or Samsung can such that in a decade's time there are no other economically viable options but Intel's latest?

Finfet's were a 10 year endeavor by Intel, that is an eye for the long view. Maybe mobile comm is no different?

 

[Arachnotronic]

Does it need to? Is that the strategic objective for Intel?

Or are they OK with this being something that holds back their gross margins by 3% overall but keeps the competition at bay for another decade while they run their process node teams faster than TSMC or Samsung can such that in a decade's time there are no other economically viable options but Intel's latest?

Finfet's were a 10 year endeavor by Intel, that is an eye for the long view. Maybe mobile comm is no different?

Idontcare

Based on what Intel has stated repeatedly in its investor calls (so, all usual caveats apply), its goal is to eventually run this business at a profit.

They expect to "significantly shrink the loss" during 2015, if that's worth anything.

 

[Qwertilot]

Even if it doesn't ever run a notable profit, if it lets them keep building/filling their fabs - and so making the huge profits they do on their other stuff, it'll be very worthwhile. Must be quite a big part of it all.

 

[III-V]

Anyone here believes Intel's Mobile and Communications Group will ever come to be profitable ???

Absolutely. Do you not know what products are in the pipeline, or something?

Intel's XMM 7260 is in Samsung's Galaxy Note 4 and Galaxy Alpha, and will continue to be adopted in other products over the course of this year and next year. In addition, they have an unnamed successor to the XMM 7260 due next year as well.

Cherry Trail, Broxton, SoFIA LTE and SoFIA 3G are due next year. Broxton might slip into 2016, but the other 3 are a certainty for 2015.

I don't think they'll be profitable in 2015 quite yet, but they'll certainly cut their losses significantly.

 

[Arachnotronic]

In addition, they have an unnamed successor to the XMM 7260 due next year as well.

It's not unnamed; the next one should be XMM 7360 (shocker, I know), then the 14-nanometer follow-on is XMM 7460.

 

[teejee]

Absolutely. Do you not know what products are in the pipeline, or something?

Intel's XMM 7260 is in Samsung's Galaxy Note 4 and Galaxy Alpha, and will continue to be adopted in other products over the course of this year and next year. In addition, they have an unnamed successor to the XMM 7260 due next year as well.

Cherry Trail, Broxton, SoFIA LTE and SoFIA 3G are due next year. Broxton might slip into 2016, but the other 3 are a certainty for 2015.

I don't think they'll be profitable in 2015 quite yet, but they'll certainly cut their losses significantly.

It doesn't seem to be an Intel modem in Note 4. http://technews.co/2014/10/13/cost-of-samsung-galaxy-note-4-higher-than-iphone-6-plus-report/

 

[AtenRa]

Does it need to? Is that the strategic objective for Intel?

Or are they OK with this being something that holds back their gross margins by 3% overall but keeps the competition at bay for another decade while they run their process node teams faster than TSMC or Samsung can such that in a decade's time there are no other economically viable options but Intel's latest?

Finfet's were a 10 year endeavor by Intel, that is an eye for the long view. Maybe mobile comm is no different?

Very nice observation, I believe they both dont like to loose money from that group and they would like to make it profitable eventually. And secondly and perhaps most importantly all this about mobile (Tablet and Phone) is about production volumes and future Fab economics.

You will know much better than anyone else here about R&D cost and how high volumes any Fab will need to produce in order to introduce a new process every two years like before(for products like Intel is selling not high margin custom ICs). If fabs on new highly advanced, high cost FinFet nodes will not have the necessary volume production to depreciate each node within a given time they will have to prolong the previous node for 6-12 or even more months before they will start the next.

Intel knows that only desktop, laptop and server x86 market is not getting higher volumes quick enough to sustain this two year tick-tock strategy any longer. They need to find a new high volume market, aka mobile (tablet & Phone). They need to get high market share in order to have high volumes, they dont care to make profit directly out of it NOW, they will want to make all the money they lost later when they will dominate that market.

Thats how i see it, could be wrong.

 

[III-V]

It doesn't seem to be an Intel modem in Note 4. http://technews.co/2014/10/13/cost-of-samsung-galaxy-note-4-higher-than-iphone-6-plus-report/

Fun fact: smartphones come in different flavors designed for different countries.

It's not unnamed; the next one should be XMM 7360 (shocker, I know), then the 14-nanometer follow-on is XMM 7460.

Ah, thanks for that. Was looking at an older roadmap, it seems. XMM 7460 should be incredible... I really want to see FinFET's apparent superiority over planar for RF in action.

 

[Arachnotronic]

It doesn't seem to be an Intel modem in Note 4. http://technews.co/2014/10/13/cost-of-samsung-galaxy-note-4-higher-than-iphone-6-plus-report/

XMM 7260 will be in the Note 4 for select regions. For example, the US-based version is Qualcomm based (AFAIK), but some Int'l markets will see an Exynos + XMM 7260 version.

I don't think the Note 4/Galaxy Alpha volumes will be super duper huge for Intel, but given the $1 million quarter that they just did in MCG, every little bit helps.

 

[teejee]

XMM 7260 will be in the Note 4 for select regions. For example, the US-based version is Qualcomm based (AFAIK), but some Int'l markets will see an Exynos + XMM 7260 version.

I don't think the Note 4/Galaxy Alpha volumes will be super duper huge for Intel, but given the $1 million quarter that they just did in MCG, every little bit helps.

The teardown is for an Exynos Note 4 though. I can't find any convincing evidence of XMM 7260 in any Note 4 version.
It is very easy to find convincing evidence for 7260 in Samsung Alpha.

 

[Arachnotronic]

The teardown is for an Exynos Note 4 though. I can't find any convincing evidence of XMM 7260 in any Note 4 version.
It is very easy to find convincing evidence for 7260 in Samsung Alpha.

You're right; this Taiwan-oriented version of the Note 4 has a Samsung 303 LTE-A Cat. 6 baseband.

I think also Intel may be making a false claim asserting that it and Qualcomm are the only two shipping category 6 LTE. Apparently Samsung is, and I know HiSilicon's Kirin 920 also has a category 6 LTE-A baseband built in.

Anyway, Intel's XMM 7260 may be in Note 4 models destined for other regions, though.

Do you know if any carriers in Taiwan actually support category 6 speeds?

 

[Idontcare]

You're right; this Taiwan-oriented version of the Note 4 has a Samsung 303 LTE-A Cat. 6 baseband.

I think also Intel may be making a false claim asserting that it and Qualcomm are the only two shipping category 6 LTE. Apparently Samsung is, and I know HiSilicon's Kirin 920 also has a category 6 LTE-A baseband built in.

Anyway, Intel's XMM 7260 may be in Note 4 models destined for other regions, though.

Do you know if any carriers in Taiwan actually support category 6 speeds?

I'm here (or there, however you look at it ) so I can ask around. Problem is the language barrier and the technology knowledge gap is even more striking here versus the USA.

I could ask 1,000 different people about Cat 6 LTE-A, in Mandarin or English, and I'd get nothing but blank stares looking back at me. But I'll check around and see if I get lucky. First place to start is in the Samsung stores, obviously. Unless someone else beats me to it.