Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
How will the ASP drop to $52 affect product already in the channel? Will Intel have to make refunds to the channel inventory holders?
That's the way it worked in the electronics distribution business (back in the 90's anyway), the distributors get price protection. That's why the OEM's who do a buy/sell get pissed about bigger than expected price drops, they have to mark inventory to market. Makes sense if you think about it, otherwise distribution would be too risky to be a viable market. You'd be sitting on inventory depreciating at 10-20%/quarter, and subject to dramatic price drops. The manufacturers need an inventory buffer so you don't end up "backordered to the fab", disrupting your planning processes. The components are too specicalized (and change too often) to have a viable futures market.
In the past AMD didn't recognize revenue from distribution until it sold through to an end customer. Intc was apparently more aggressive on revenue recognition. Of course, Intc could have done a "deal you can't refuse".
Anyone seen any hard data on FBDIMM power draw and cost? I've seen numbers for the AMB anywhere from 6 to 12 watts, but most references are from last year. Micron weaseled out of it and put TBD on their data sheets as far as I can tell. I'm trying to figure out the BOM cost and power requirements for "equivalent" servers, say 2S & 16 or 32 GB of RAM.
Lastly are you sure that the cost structure of Spansion did not in fact increase permanently because of the separation?
I'm pretty sure Spansion didn't need the SDC in Sunnyvale and they probably didn't need Fab 25 in Austin. I believe AMD did some strategic asset shedding on this one. Spansion probably could have run all their profitable product through the JV fabs in Japan, now they have to fill the fabs with marginally profitable products. Assuming this is true, they have alot of potential for improved mix (ASP) if ORNAND or the integrated RAM/Flash takes off.
You have to remember that part of the mission of the government labs is develop new technologies. The labs are the leading edge in supercomputing, obviously they saw some potential in the Opteron that wasn't available on other platforms. Remember, at one point the transistor was proprietary technology.
http://www.sandia.gov/mission/ste/index.html
Realizing that the best science, technology, and engineering cannot be developed alone, Sandia collaborates and partners with a variety of like-minded industry, government, and academic institutions to build capabilities and accomplish our missions.</i?
Not all supercomputers are the same. The basic problem they're trying to solve is the interconnect, something the high bandwidth, low latency hypertransport bus is ideally suited for.
http://www.hoise.com/primeur/02/articles/monthly/CL-PR-02-02-1.html
He concluded by saying: "There are two principal types of supercomputers, adapted to different niches. These niches are important. Picking the wrong type of supercomputer wastes money because you are paying for unused transistors or connections. Reducing Overhead saves time in Type T systems. Reducing Latency is the preserve of Type C systems. The cost in Type C systems is in wires (connections) not in processors."
note: type C is clustered, type T is tightly coupled.
As you already know, one of Dell's arguments against using AMD in their lineup is that AMD is unable to supply them in the volume they require.
I think "volume" is Dell-speak for preferential access (and pricing). Their business model depends on it.
Hard to explain the rival between TU and A&M
It's basically the Texas version of the red state - blue state phenomenon. tu was the originally the liberal arts / lawyer crowd and A&M the engineering / agriculture. A&M was an all male military school until the 60's.
Just pointing out that your view is completely polarized, and plenty of educated people who have a fully sized brain can come to the conclusion that they want a Dell XPS with Pentium Extreme Edition.
I have 2 friend that did just that against my recommendations, both seriously regret their decisions. One bought a high end media center pc and the noise renders it virtually useless as far as he's concerned. The other says his pc sounds like a vacuum cleaner, and the home office is 80 degrees.
We're not talking about marginal differences here. I don't see Dell selling any media centers if Best Buy etc. starts putting the in their sound demo rooms where you can actually hear the noise, instead of out on the floor. Brand image is powerful, but it doesn't negate reality (other than Apple, but that's a cult).
Intc Buyback & Shares outstanding
Re: http://download.intel.com/intel/finance/presentations/IntelQ32005final.pdf (pg 10).
The question is the motivation, they're showing a significant drop in shares outstanding. Is this being done to keep up the appearances of being a growth stock and maintain the high P/E? Here's an interesting paper: What Drives Companies to Repurchase Their Stock?
http://gsbwww.uchicago.edu/news/capideas/oct04/repurchase.html
An interesting conclusion from this paper: "We also find that executives are more likely to undertake repurchases when earnings fall short of levels necessary to sustain prior growth rates in eps".
In the days of regulated utilities (high cash flow, stable business), they could have turned themselves into "growth stocks" by repurchasing shares instead of paying dividends.
Regarding INTC shares, as a corporation they're generating huge profits, the question is who are the beneficiaries. The analysts should be beating up management.
Accounting gurus may chime in, but stock repurchase doesn't hit income other than changing diluted eps. IIRC the repurchase in excess of issue price hits retained earnings. They also state they want to limit dilution to 2% / year.
As far as the scale of the repurchase activities, From the July '05 10-Q (Q3 isn't up yet) $4.3 B of the $6.7 B operating cash flow in 1H05 went to stock repurchase, while net income was $4.2 billion and dividends were $990 million.
From the 2004 annual report, Of the $13.1 billion in operating cash flow, about $7.5 billion went to stock repurchase. Net income for 04 was right at $7.5 billion, so the hit was significant and dividends were $1 billion . Note 4, pg 57 shows $7.5 billion spent to repurchace 301 million shares, while the shares outstanding only dropped 127 million.
If you look at the 2004 report(pg 56, note 3), you'll see that ~350 million option shares were excluded from the outstanding shares since they're underwater. P65 shows the number of options outstanding by range of exercise price. The 10-Q's don't show the same detail.
Some details on stock repurchase accounting http://www.nysscpa.org/cpajournal/2000/0800/dept/d85600a.htm
25 years ago the drinking age was 18 in Texas, Houston to Dallas was considered a 1 6-pack drive, Houston to El Paso was at least a case (drinking beer while driving was not illegal, only drinking too much while driving).
WRT A&M, all those liberal arts and education majors have to come from somewhere. My freshman Electrical Engineering class at A&M was about 1200 students, 75 graduated, about 1/2 of those in 4 years with no summer school.
The number of shares listed in the table are "diluted" shares, which means that all outstanding options and grants are already counted in that number.
I believe the diluted shares only counted the options with a strike price less than or equal to the market price. That can leave a pretty big overhang of underwater options. I haven't read the 10Q's in detail in a year or so, you have to read the footnotes for the specific treatment.
I fully agree that Itanic is a dog, Intc's Edsel. It's a big enough fiasco that it would have taken down a company with fewer resources and less market influence. I'm sure they wish they could get a do-over on the decision. But talking about the "overall" profitability isn't meaningful in a business context. I still think it was a good decision (corner the mid-high end processor market) but bad outcome/execution. They probably should have cut their losses earlier as well.
That said, the story isn't done yet. It's hard to say how they might adapt to their "learning experience". The key to Silicon Valley is learning from your failures. Itanic has to be like a 2x4 to the head, they'll execute the guilty and some of the innocent, revamp the organization and move forward. They'll be the same tenacious (rapacious?) competitor they've always been.
BTW, I am most definitely not an Intc cheerleader, I'm WAY overexposed on AMD and actually worked for them years ago. Personally, I hope Intc keeps flogging the Itanic and gets bogged down in CYA and blame shifting, but I wouldn't count on it.
- Why do you only try to estimate current development costs when forming an opinion about 'Itanium profitability'? Shouldn't all dev costs so far be included when discussing this? You can't just discard it like it never happened.
That's the sunk cost fallacy. It's business 101 (ok, maybe 201), you don't consider past costs when evaluating future investment opportunities. It doesn't matter whether the cash was spent on Itanium development or swiss cheese futures. The investment decision TODAY is whether Intc can get a decent return on the incremental resources invested, treating the past development as a fixed cost. There's almost certainly some strategic valuation still in the mix.
You also have to make a distinction between a bad outcome and a bad decision. Given the info that Intc had at the time, they thought they could lock up the 64 bit market with a proprietary solution. That would have been worth giga bucks. The decision was made in the AMD K5/K6 era, who would have thought AMD could come up and deliver something like the Opteron? A posteriori you have to call it a good (aggressive) decision with a bad outcome (or perhaps bad implementation, Intc should have had a better X86-64 contingency).
Fujitsu Introduces the LifeBook S2000 Thin and Light Notebook with Next-Generation AMD Turion 64 Mobile Technology
Notebook Ideal for Mobile Professionals and Students On-the-Go
From the data sheets:
Turion: 4 hrs on main battery, up to 7 hrs with both modular bay and main battery.
PentiumM: 3.75 hrs on main battery, up to 6.5 hrs with both modular bay and main battery.
Yes, that's with the same battery. If you take a look at Fujitsu's eCommerce site, the S2000 is $200 less at each of the 3 configuration levels.
Tom McCoy is Executive VP, not a legal hired gun.
If this thing really gets going and AMD does subpoena customers we should be in for some fireworks. Intc may have trained their folks not to leave an antitrust "smoking gun", but there should be ample evidence of a corpse on the customer side.
It could well be payback time for all these years of threatening and intimidating the OEM's. I can see the call between Intc &IBM CEO's "But Paul, we're under subpoena. I can't really do anything about it, that would be obstruction ...". Meanwhile, the entire IBM staff is digging through old emails: "Hey AMD, what about THIS one, maybe you should subpoena...."
BTW, you don't really think Intc puts their sales guys through antitrust training so they know the rules, it's so they know what NOT to put in writing.
HP to use AMD chip in business notebooks
TURION 64 BESTS INTEL'S CENTRINO
HP will use the Sunnyvale company's Turion 64 microprocessors in its ``Compaq nx'' series of business computers, including a ``thin and light'' $999 nx6125 model with a 15-inch screen that makes its debut today. The company said it will give business customers an alternative to Intel-only computers offered by competitor Dell.
http://www.siliconvalley.com/mld/siliconvalley/11956580.htm
I hope that when AMD's Fab36 and Chartered capacity comes online, that they don't 'just dogmatically' drop prices.
...as they will "then" be in a position to set good (slightly higher) prices and be able to supply the volume.
My bet is that they'll settle into a duopoly mode, but Intc may play hard ball and try to starve AMD of capital for the next generation. Some analysts have projected that the semi world will consolidate to 2-3 virtual consortia around fab technology since Intc about the only company left that can afford the cost and risk of dropping $4 billion on a fab. AMD-IBM-Chartered seems to be converging on a common fab technology.
AMD's virtual gorilla strategy may prove the key to long run success. Intc's focus on platform strategies has to concern the Taiwanese companies, 5 years out they have no market if AMD is gone. I can imagine the state of technology if Intc's driving the rate of innovation (we'd all be on Rambus by now). Putting Fab 30 & 36 in Dresden was another great move, never underestimate Europe's own sense of superiority and willingness to engage in protectionism.
A friend of mine ( Dr. in Marketing ) used to say "There's only one thing that motivates in business, and that's 'making money' "
The continual downward price-spiral in the PC business caused effectively by the price-bands remaining constant whilst new product are introduced, means that "holding products in stock" is a dangerous business.
The OEM's are willing participants in this exercise. Since the K6 generation they've been playing the "INTC = business = high margin" vs "AMD = home = low margin" game. The last thing they want is the collapsing margins of the home market. For the last 5 years or so you could (almost) always get a higher performing pc at the same price point using AMD cpu's (more $ for memory / disk / video).
I think it's a bit like game theory's Prisoner's Dilemma. Everyone is better off if everyone cooperates, but the first one to defect generates outsized gains. The difference here is the implicit threat of punishment of the first defector by the dominant supplier. Other than SUN (who wasn't really in the X86 market), we're just getting to the capacity/technology point where AMD can change the game.
more JFTC details on 'huge rebate system'
predatory pricing
An anti-competitive measure employed by a dominant company to protect market share from new or existing competitors. Predatory pricing involves temporarily pricing a product low enough to end a competitive threat.
http://www.advfn.com/money-words_term_3770_predatory_pricing.html
IIRC you have to show a company sells below average variable cost to prove predation. With high fixed fab, chip design & process R&D costs the variable cost is insignificant. IMHO the law doesn't adequately address the modern tech market. Under traditional accounting the variable cost of producing Microsoft Windows is probably 10 cents (how much does it cost to press a CD these days?).
The dual-core force is with AMD
During AMD's XChange presentation, Gary Bixler, director of the company's North American system builder channel, led a demonstration of upgrading a single-core, dual-processor Opteron system to a dual-core system.
The process, which included a BIOS upgrade, took about 15 minutes, and the demo ended with a 64-bit enabled Microsoft Windows operating system recognising four CPUs from the two dual-core processors in a single workstation.
.........
AMD contends its dual-core technology outpaces Intel's with its internal I/O architecture, which it said prevents processing bottlenecks. Bixler also noted Microsoft has begun using AMD-based systems to manage its US$60 billion in assets.
"It was taking them 40 hours to process [data for] all of these assets, just on an ongoing basis," Bixler said. By using AMD-based technology, Microsoft saw a 325 percent improvement in efficiency in its asset management applications, he said.
http://www.itnews.com.au/newsstory.aspx?CIaNID=18792
Opteron's architecture looks watered down compared to the original POWER4 interconnect- and memory-architecture
Sort of like Qualcomm watered down the military CDMA implementation. "Simplyfying the implementation while retaining key functionality" would be my phrasing. The essence of great engineering is elegantly simplicity.
Some more tidbits from the APC site, APC sells enclosures, so they don't have a vested interest in vendors. You'd have a hard time justifying the 40% power increase going with Dell instead of HP.
Configuring InfraStruXure Systems for:
IBM Blade Center Servers (consumes up to 714W/U)
ftp://www.apcmedia.com/salestools/RHAN-64VRGJ_R0_EN.pdf
Dell PowerEdge 1855 (consumes up to 757 W/U)
ftp://www.apcmedia.com/salestools/RHAN-6846DM_R0_EN.pdf
HP Proliant p-class (consumes up to 536W/U)
ftp://www.apcmedia.com/salestools/RHAN-66FQR4_R0_EN.pdf
From the weekly Computerworld Mailer. For those of you that don't realize the problems with data center power & cooling.
Some tidbits from the exec summaries
- While vendors are designing blade servers requiring up to 20KW per rack, most data centers designed to cool no more than 2kW per rack.
- From infrastructure cost standpoint, the benefits of increasing density are small, the benefits of reducing power consumption is high. A 50% reduction in space gives a 4% improvement in TCO, a 50% reduction in power gives 35% reduction in TCO.
- Has some graphs on actual capabilities of data centers. 90% of data centers had capability of <3kW/rack, 60W/ft^2 cooling.
From: Power and Cooling for High Density Racks & Blade Servers
ftp://www.apcmedia.com/salestools/SADE-5TNRK6_R3_EN.pdf
Are you struggling to maintain a data center packed with blades?
APC will guide you through the solution! This white paper will show you a ten-step approach on how to increase the cooling efficiency, the cooling capacity, and the power density of your existing data center! Download "Ten Steps to Solving Cooling Problems Caused by High-Density Server Deployment" right now.
http://zones.computerworld.com/apc?lpid0218059701142196idlp
Processor Watch: Who Really Deserves Credit for the New AMD?
Editorial by K^2. Nothing really new but a decent overview of K6 - K8 evolution
Despite all the praise heaped on the present AMD management, however, some past contributions may be underappreciated.
For one, AMD would be unable to deliver competitive processor systems without the help of third-party chip-set suppliers. During platform transitions, AMD itself has delivered chip sets that provide little more than basic functions. Of the third-party chip-set vendors, VIA and Nvidia have probably been the most influential in the success of the K6, Athlon, and Opteron processors. VIA has supported AMD with chip sets for many years; Nvidia was the right partner to help AMD break into the corporate PC market.
http://www.mdronline.com/watch/watch_Issue.asp?Volname=Issue+%23179&on=1#item1
...why should Microsoft screw themselves over going into the major earnings section of the year when AMD can only produce 5% of the 64 bit market...
Anyone know if Microsoft capitalizes their development costs? If they do, it may make sense to hold back the release until there's a potential for more volume.
Personally I think they're choking on SP2 & NX. They've been sloppy for so long they're having a problem getting out of the hole they dug for themselves. Anyone wanna bet they have internal code dependencies on the existence of buffer overruns? Much like the Catholic "Original Sin" doctrine, they've all been corrupted by the "Original Kludge" and Saint Bill the Architect can't bail them out.
Re: Die Bank
On die vs dice, I concur that the term die I used for both singular and plural forms. The term die bank pertains specifically to the semifinished goods inventory, which I believe was the origin of the terminology debate.
Much of the industry uses a push-pull planning approach. The lead time of the fab may be 12-16 weeks while assembly is 1-2 weeks. The fab, typically onshore, produces wafers (hopefully somewhat generic wafers) to a forecast (push) into die bank at the assembly plant, typically in Asia. Test / assembly pulls from die bank based on actual orders (pull), or at least in response to a near term demand forecast. You may even scrap entire lots if the sale price isn't worth the cost of assbly & test.
I don't know if anyone does build to order thru fab, although I have heard of alternative downstream processing. I've been out of the industry since the late 90's (worked at AMD for 5 years), but I don't think the planning has changed all that much.
Die bank as semi finished goods inventory:
http://bmrc.berkeley.edu/courseware/ICMfg92/text/fab-6
Decent overview of planning alternatives:
http://www.public.asu.edu/~ysun5/Chapter2.pdf
SAP does it:
http://www.sap.com/businessmaps/30BB776CAA1A44578F29AD46D1A302FF.htm
EPROM Still has legs. Apparently ATML's still making these on 0.5 micron. These things are going to be like the alligator was to the dinosaur. 20 years from now we'll have optical cpu's, but there'll still be EPROM's.
Improved Prospects Lure Suppliers Back Into EPROMs
Just a few years ago, the consensus was that the almost two-decades-old EPROM market was finally headed for extinction. Emerging flash memory devices threatened to displace EPROMs in every application from modems to PC BIOS. Established vendors such as Fujitsu Ltd., Intel Corp., and Philips Electronics N.V. were exiting the market. And even suppliers willing to hold on were scaling back EPROM production to open up capacit y for more profitable flash chips and microprocessors.
But predictions of the EPROM's demise remain premature. In the last year a number of vendors that had previously talked about rolling back production have changed their minds.
....
AMD is also offering higher-speed EPROMs. Although a major player in flash, the company remains committed to EPROMs. Its offerings range from 64-Kbit to 4-Mbit devices.
http://www.my-esm.com/techfocus/1048main.html
I've always heard it referred to as die bank. See under: Stage 2: Wafer Probe and Sorting at:
http://bmrc.berkeley.edu/courseware/ICMfg92/
The product isn't necessarily fully committed at this point. IIRC, my old employer made 4 & 8 port ethernet controllers from the same die in different packages. It's cheaper than scheduling different wafers thru the fab.
I've always heard it referred to as die bank. See under: Stage 2: Wafer Probe and Sorting at:
http://bmrc.berkeley.edu/courseware/ICMfg92/
IIRC, my old employer made 4 & 8 port ethernet controllers from the same die in different packages. It's cheaper than scheduling different wafers thru the fab.
I wouldn't plan on anything out of Dell at least until the 90nm (if ever). As before, they'll just use the design as negotiating leverage with INTC. INTC will throw them more money in "volume rebates" or co-marketing funds so they don't "lower" ASP's, and they can bury the costs in MG&A or the Communication Group. Opteron would have to get to 20-30% of the market before it becomes too costly for INTC to buy the Dell account.
In the near term the downside to INTC is higher than the AMD upside. INTC loses pricing power, even though AMD can't supply the entire Market. INTC has to deal with every potential defector and they have 98% of the server market. They also can't afford for AMD to make $$$$$ to pay off debt and buy fabs. I think you'll see radical downsizing and price reductions at INTC before you see AMD at $50.
I think the upside plays will be companies with innovative products that can leverage the low cost 64 bit performance, like Verari (audio workstations) or DCS (digital video projectors), maybe with SUNW as a dark horse.
http://www.verari.com/workstationcenter.asp
http://www.digitalcinemasolutions.net/
Most interesting comment from the eWeek article (makes sense why Sun is onboard):
Still, AMD is making inroads to the enterprise. VeriSign Inc. is moving several applications and databases off IBM, HP and Sun Unix servers and onto two- and four-way systems from Newisys Inc. running Opteron chips. These applications—which include the Mountain View, Calif., company's Domain Name System—require a lot of memory, and VeriSign was looking for a way to port them to a 64-bit platform that was less expensive than Unix systems, said Ari Balaugh, senior vice president for operations and infrastructure at the company.
After a six-month trial, Balaugh liked the price/performance of Opteron servers, he said. A Newisys system with four Opteron chips is 1.8 times faster than an eight-way Unix server, which for large applications requiring 16GB to 32GB of memory can translate into a fourfold to sixfold cost difference, he said.
"We were a little leery of [the Opteron] initially," Balaugh said. "But you get good enterprise features from Newisys, and the Opteron is just fast.".
http://www.eweek.com/article2/0,1759,1569994,00.asp
RE: anandtech A64 preview
Forget the techie BS, here's the real eye openers:
http://www.anandtech.com/cpu/showdoc.html?i=1856&p=8
To satisfy curiosity, we also compared performance of the Workstation nVidia Quadro FX2000 video card on both the dual Xeon Intel 875 platform and the single-CPU Opteron platform. You would expect that 2 Xeon 3.06 CPUs with 1MB of cache would be the clear winner of this comparison. The results, however, are quite surprising: The results are basically even, which is amazing considering we are comparing a single 2.0 GHz Opteron to Dual 3.06 Xeon with 1Mb cache.
Also, from the final summary
http://www.anandtech.com/cpu/showdoc.html?i=1856&p=11
Gaming is one area where our tests show Opteron at 2.0GHZ an amazing performer. When you find game benchmarks 10% to 20% higher, you are genuinely impressed. However, in some of the very latest DX9 benchmarks, Athlon64/Opteron was 40% to 50% faster.
.....
One particularly noteworthy area was the performance of the A64 level Opteron compared to an 875 Dual Xeon 3.06 system. We really expected the Xeon dually to trounce our single Opteron, but instead, found a virtual dead-heat.
Gotta wonder about the price/performance ratio, INTC may go the way of Sun in the workstation market.
Aquisition isn't a bad thing, in the startup world successful "equity events" are IPO or acquisition. Often acquisition is the better option because you instantly gain the acquiring firm's sales channels and economies of scale (engineering support, vendor leverage & manufacturing scalability...). In the networking world the best thing going was to be acquired by Cisco. Self funded growth is a slow process, the IP is worth alot more to a company that can ramp quickly before other companies enter the market.
With the RIF's, life in a startup is like that, there's a huge focus on managing cash flow. Hit a major milestone then RIF a good chunk of your tech staff, just keep the architects and visionaries. On the marketing front, RIF the people after the plan has been designed and contracts cut. You either make yourself indispensible or assume your "full time" position is just another contract gig.