This is what you said
'I believe if PC didn't amortize the low power CPU and the high end CPU, as well as various other IP blocks, R&D would be up. Let's call this $1 billion (we can debate this, obviously) total. '
this is what I said
'The annual cpu development costs would nowhere be near $1bn and not even $500m.'
If it takes ~$200m annually to keep a cpu pipeline going then double that if you want to include Atom as well as Core although in my view of the new server-only Intel Atom/Avoton would be dropped as they offer no performance/power advantage over LV Xeons. Either way my figure is right and yours is way off base. This is also an amount that is negligible in the face of DCG's multi-billion annual revenue so clearly the Core subsidizing effect on DCG is negligible when your article said it was pivotal. Once again you made a broad sweeping statement that had no basis in reality. It's just as well no-one really has to mark and pay your work on accuracy although it is turning into quite a job for me considering how sloppy you have been lately.
Face it - with INTC this repeatable strategy will make you more money than buying and holding. And the ex-dividend date is not even in the picture within the short to medium range.
. . . .
George Soros, "the man that broke the Bank of England," has had one of the longest and best track records of any big name investor of the past half-century. Founded in 1969, Soros Fund Management averaged a whopping 20% annual rate of return from the time of its founding in 1969 to its closing to outside investors in 2011. With a current net worth of $27 billion, the 83 year-old Soros is still active in investing and philanthropy, having given away more than $8 billion over his lifetime.
According to the most recent 13F filing, three interesting trades his fund made last quarter were buying shares of tech titan Intel (NASDAQ: INTC ) and oilfield services behemoth (and occasional public enemy #1) Halliburton (NYSE: HAL ) , while selling a big chunk of his shares in struggling retailer J.C. Penney (NYSE: JCP ) . What can we learn from this living legend's moves? Let's take a closer look.
. . . .
Sentiment: Strong Buy
'We bow down to all you amateur and professional financial analyst BS'ers'
Very unprofessional. I was enjoying our debate which was free of name calling and personal attacks, but then you had to punctuate it with this 'gem'.
This post from 'chipguy' is great and actually serves to prove my point. If Intel's DCG were a standalone entity, it would be working on Haswell-EP, Haswell-EX, Broadwell-DT (for Xeon E3), Broadwell SoC, Airmont/Goldmont core, Denverton SoC and probably its successor, and according to LinkedIn the Skylake server chip has been in the works since 2014, so I wouldn't be surprised if it were working on Cannonlake and the various SoCs that'll pop out of that.
All I'm saying is that the Haswell/Broadwell/Skylake/Cannonlake/Airmont/Goldmont core R&D would all fall under DCG whereas before it would not. My point that there would be SOME (likely non-trivial) increased R&D allocated to this new DCG versus the old one would be there.
My observation is valid in direction even if magnitude can be debated. Instead of following through this debate to a cordial agreement to disagree on the magnitude, you instead choose to insult me yet again.
I don't think so, but this is in-line with the rumors I have heard of Apple offering top Intel guys ~40% raises over their current posts.
'chipguy Monday, 03/10/14 04:36:25 PM
Re: mas post# 131298
Post # of 131299
You have worked this out before so I wonder if you could just remind me what in your considered opinion the annual cpu development costs of Itanium were at its peak and now ?
At the recent peak (Tukwila physical design concurrent with Poulson
uarch development) it was probably on the order of $200m-$300m per
year. Now it is probably less than $10m (Kittson now seems to be a
rebadged/speed bumped, hidden feature enabled Poulson) which is
basically manufacturing/test support. AFAIK FTC moved onto other
things about a year ago when HP refused to fund a 22 nm Kittson. '
But #$%$ do chipguy and I know, after all we have only had real jobs in the IT/computing industry producing real products for real customers for decades. We bow down to all you amateur and professional financial analyst BS'ers.
Core asp is closer to $100 and server asp is about 4 times that. You also could keep refitting the other parts of the same Fab for years as well as not have to keep on the 2 year process treadmill as die size is not so important for a server part. The whole capex/process roadmap would lighten compared to what it is now so the amount to be depreciated per year would change and not just on a pro rata revenue basis. Remember also the low volumes all the Risc companies survive on and IBM, Fujitsu, Dec have/had their own Fabs. The company net margins would not be radically different to what they are now.
Fundamentally I still think you got it wrong in your analysis and that it is the billions of revenue at around $400 asp that make DCG what it is and not the IP/depreciation contribution from Core. Intel server asp has been rising over the years as PC asp has been falling and that is because Intel server chips have been migrating further up the food chain from their humble single/dual socket beginnings.
One further comment. Haswell tablets are generally heavier & thicker than their BYT counterparts and I believe all Haswell tablets have fans. Unless one needs the CPU horsepower of an ultrabook, BYT tablets are easier to carry & use. The idea of the 64 bit BYT tablet with 4GB RAM is a really encouraging.
ljack360 ... thanks for explaining the subroutine use with the instruction sets (or lack of). I'm not in that field, but knew a little and thought it might be something like that. It was good to have someone spell it out for me. Thanks.
"You think only having 1 Fab (which in fact could just be the development one D1X used as production too) and not having 3 Fabs elsewhere won't mean a massive drop in capex ? GIGO. "
Not what I said. I am assuming that the depreciation applied to each operating segment is done weighted to the revenue contribution. I also believe that the Xeons benefit from running in factories that have already been depreciated by the Core chips before the Xeons ever hit the fab.
How many billions worth of Core chips will be run in the first 14nm factory before the very first Xeon E5/E7 on 14nm will be run?
You will have the ENTIRE 2014 and 2015 Broadwell/Skylake-DT/MB/ULT/ULX run happen in those factories before Broadwell-EN/EP/EX even begin their run. That fab is paid for.
So, let's do a thought exercise:
Say Intel need only build 1 fab, and let's say that it's $5 billion. If I run $20 billion worth of Core chips (@$150 ASP) through it (that's 132 million chips) over 2 years, and let's say that I've spread the depreciation of the $5 billion over those initial 132 million chips (~$37.5/chip), then how much depreciation does my shiny new Xeon on this new node cost?
Well, your wafer cost now includes minimal depreciation and, as an added bonus, the process is mature and yields very well...offset by the more complex/larger die.
So, no, I don't think Intel The Server Chip Company would be enough to keep the current ~47% DCG op margin.
Xeon Phi and Avoton are part of Other Architecture and new E5s are not released every year and are just a MP derivative version of E3 anyway. No, it's not in the same ballpark.
" Intel is still developing Itanium in its DCG business and that is not sticking out like a sore thumb is it ?"
I believe this was an NRE model similar to MSFT/SNE with the XB1/PS4 SoCs for AMD. HPQ paid for the development, AFAIK, but this isn't something I'd publish without further DD ;-)
How many chips does Intel release for servers in a year?
1. Xeon E3 (repurposed desktop chip) + chipset
2. Xeon E5 (several variants/chops) + chipset
3. Avoton/Denverton/etc. with variants such as Rangeley
4. Xeon Phi (which will apparently leverage Crystalwell/eDRAM)
Now, I don't know if BRCM's estimate included designing/validating the individual IPs, and I doubt BRCM's chips are as "mission critical" as the higher end Xeons...but yes, it's roughly in the same ballpark.