Ted some more follow up on a post you did this a.m.
JC's post
"JCholewa
New Member posted 04-11-2000 13:28
--------------------------------------------------------------------------------
Hmmm. You know, I get the strongest impression that people here are avoiding the real core issues wrt (apologies to the acronym patrol) AMD. Making sweeping, blind statements such as "Athlon will keep AMD going" and "AMD will ALWAYS screw up" is frankly childish in nature. I apologize for the insult though, I mean no true disrespect, just wanted to jolt y'all a little.
I've only begun observing AMD closely in the past two years, so I will not claim to have experienced knowledge in all aspects of this company; however, I have strongly gone in depth, I believe, with AMD's fundamental actions over the past decades and I believe that I can lend some assistance with AMD's future.
Okay, first of all, blindly predicting that earnings warning on September this year is irresponsible. It's a guess -- a "WAG", as some investors say it. And it has no backbone, nothing to justify itself.
My specialty in this matter (I hope I do not appear too abrasive or arrogant, I apologize if I come off that way) is with the technological aspect. Way back in late 1998 it became obvious that AMD would benefit very strongly off this "Athlon" design. The reason for this is that it is a superpipelined microprocessor, and in fact the instruction pipeline is a bit smoother than that on Intel's P6 core. The end result of this is that the Athlon is able to reach higher binsplits than the Pentium III. What I mean by that is that the Athlon on an equivalent manufacturing technology is able to reach a higher clock speed range, on average, than the Coppermine (Intel's current PIII).
The whole key to profitability in the microprocessor market is binsplits. He who makes the most megahertz wins.
AMD's last serious bout with success and profitability was back in the pre-Pentium days. Intel introduced the 486, and AMD did their usual second sourcing thing. But what AMD did beyond this was to release their 486 parts at *higher frequencies* than Intel's offerings. AMD's grabbed nearly 30% of the market, partly due to this tactic. And they did decently, money-wise.
Then, Intel released their Pentium series. After a couple false starts, they kicked up some major process improvements and brought this part to beyond-486 levels, both in performance and frequency. This is where the troubles exhibited by AMD, Cyrix, IBM, and all other x86 "cloners" truly began. Basically, IBM and a couple others dropped out totally at that point (IBM and SGS, I think, later went back and cloned the 6x86). Cyrix and AMD tried to squeeze out the rest of the 486 series as they had to -- for the first time ever -- produce their own x86 compatible designs.
The first independent x86 generations from AMD and Cyrix were rush jobs. AMD's, in particular, was a very superscalar but very sluggishly ramping part (K5). Cyrix's 6x86 was a bit slower per clock but could reach better frequencies (it eventually made it, in fact, to 300MHz and perhaps past that), so it grabbed a small amount of success. But the essential situation was that the Pentium was much more rampable than the K5, which meant that Intel took the entirety of the high end profits.
The "next generation" parts from Intel and AMD were a very similar situation. The K6 was about as rampable as the Pentium, but had better performance (this was a huge change from the K5 -- AMD's operating frequencies essentially doubled). Nonetheless, Intel's P6 had a superpipelined core, which essentially means that it was able to hit tremendously high frequencies effortlessly. The K6, having a shallow pipeline, had to be pushed to (and beyond) the breaking point in order to even hint at competing with the P6. AMD succeeded in ramping the K6 core to seemingly impossible speeds -- they frankly have an amazing fab team, and it's really pathetic that the fab guys took total blame for the K6's design-related ramping problems in late 1997 and early 1999. But I digress ...
K6 increased AMD's sales over K5 dramatically. In fact, K6 would have been a profitable venture had AMD kept their expenditures to their normal levels. However, AMD's management (in a move that was, in total polar opposition to most analysts' impressions of these guys' capabilities, frankly brilliant) decided to ramp up expenditures to insane levels in order to push more advanced designs and additional manufacturing facilities. A combination of these insanely high spending levels and the tentative rampability of the K6 led to very large losses. Essentially, though, AMD was "sacrificing the present to save the future", to coin a phrase.
Okay ... now, this whole "Athlon" thing, and why it's different, and how it's not different:
Athlon is based on the K7 core, which is the first x86 core ever designed with higher rampability than the most rampable offering offered by Intel. The K7, like the P6, is a superpipelined core, but it is a bit better balanced, which is why Intel's P6 core is having problems achieving frequency parity with the K7 (P6 is about a speed grade behind, and there are hints that AMD may be holding back further due to marketing issues), despite Intel having a process with superior effective gate sizes and transistor switching speeds. Basically, as long as Intel's fastest core is the P6, they will be struggling to keep up with AMD in the Megahertz race. Current estimates have AMD achieving volume at 1GHz by as early as late May or June, while Intel will not reach this mark until middle to late Q3 (three months later, that is).
Many analysts seem to be suggesting that Intel can, at any time, simply depress prices and wipe AMD out of existence. This is a flawed statement. When AMD was constrained to the low end, they were barely profitable. Intel, on the other hand, had high end revenues to offset weak low end prices. The sole reason why Intel has been able to pursue so-called "predatory" pricing actions is because they have been able to keep their high end prices high while all other competitors only have low end prices which would be supressed by such an action. AMD currently has the same advantage as Intel, though, with revenues from the high end that can absorb incredible drops at the low end. Until Intel can reestablish this speed grade dominance, they will be totally unable to attempt this sort of tactic.
The question that you should really be asking yourselves is "when will Intel be able to achieve speed grade dominance?". This question should take precedence over trends showing AMD's recent rise -- this past action has little-to-no bearing on future performance. This question should take precedence over pithy opinions of the CEO of the company -- if anything, his actions over the past two years have been exactly what they should have been given the company's situation (imho, the only truly bad things Jerry has done are *1* deciding to make the majority of its business the x86 cpu market, in which the competition is a behemoth, and *2* being a hype-master during impending bad times).
Okay ... author dude (I forgot your name already, too much ADD in my brain!), you assert that AMD will make an earnings warning in September. Allow me to address how I think you may be right, and how I think you are certainly wrong.
First off, I will be beyond surprised if AMD posts a loss in any quarter of this year. This assertion of mine is not based on some blind faith that "hey, gee, Jerry the Great will pull it off!". It is based on two factors, one which you slightly addressed in your article, the other which you totally blew off (well, ignored).
Okay, I'm already made my point (I hope) about how high bin splits help dramatically towards achieving profits, even in times of high spending. What I haven't noted is that AMD is a *huge* player (perhaps I embellish slightly) in the Flash market. AMD's FASL joint effort with Fujitsu (I think that's who it is?) is quite possibly the largest producer of Flash in the world. And currently, AMD makes from Flash the same revenue as (a higher amount than, in the past) they make from microprocessors. Flash is currently an immensely supply constrained, booming market (partly due to the explosion of cell phones, set top boxes, PDAs, etc..) for which both AMD and Intel are viciously working to increase production capacity. AMD just announced tentative plans to create a Flash fab in Japan, and they're going to dedicate much of their Austin fab towards flash as their new Dresden plant ramps up. Basically, this will equate with hundreds of millions of dollars in "printed cash" -- free revenue (not *free* free; I simply mean it'll be a huge boost) that will allow AMD to leverage cash for when times are bad for their microprocessor division.
Okay ... that aside, I cannot at all see AMD posting a loss in any quarter this year. It is virtually an impossibility. The only way it will happen is if something drastic happens, like an alien invasion. Even if Dresden was firebombed -- again (apologies, this is somewhat in bad taste, I do not mean to offend) and AMD totally lost FAB30, AMD would still pull profits every quarter this year.
But they might preannounce an earnings shortfall for Q4. Much less of a chance in Q3, since AMD has finally learned the cardinal rule of guidance ("underpromise, guide up slightly, then overdeliver"). If they keep their current tactics, they will start off by forecasting that Q3 will be "down to flat" from Q2, then they will guide from there, so shortfalls are unlikely, even if their earnings do go down quarter to quarter.
Anyway, when will AMD start feeling profit-hurt? This is almost too easy to answer, if we forget what I said about Flash above (Flash could offset losses possibly, but I'm not going to stake my name on this assertion). Okay, the only thing Intel can realistically do to cripple or hurt AMD is to release their next generation "Willamette" design. From the technical observations of my peers, I have come to the conclusion that Willamette will be relatively slow in performance at any speed grade (especially in floating-point applications, where it appears to be only equivalent to the Pentium III, SIMD-aside -- with maximum SIMD optimizations, which are an unrealistic option, it appears to only achieve parity with the Athlon). But Willamette has one feature which is critically important: It is "hyper-pipelined". Just like how being "superpipelined" gave Intel's P6 an advantage over the K6, "hyper-pipelining" will give a clock speed advantage of the Willamette over the K7 core. The degree of this is unknown, but due to the concept of diminishing returns, it is not expected to be as severe a difference as it was with the P6 over the K6. Nonetheless, the situation currently with K7 and P6 will likely be reversed, and possibly more -- AMD will be straining every resource to reach x MHz while Intel will effortlessly pump out 1.05x MHz parts with relative ease.
This will, of course, be modified by process technology. Intel is currently ahead in process technology, having smaller effective gate widths (Leffs) and faster switching transistors, according to the more technically-oriented friends of mine. AMD's HiP6L Dresden process will probably bring them to approximate parity with Intel's current 180nm process. AMD looks to be putting together (with Motorola) a 150nm HiP7L process. If Intel doesn't have any significant process enhancements at that point (they are planning for 130nm at end of 2001 I think, but I don't know what they're doing before then), then AMD will be at a process advantage, which could allow them to achieve parity with Intel, frequency-wise.
But assuming that AMD fails at this, and is using their HiP6L and CS50 technologies for longer than expected, then what happens? This is the scenario I am anticipating by default, because anticipating the best case scenario is generally a bad idea, right?
Okay. Willamette will be a stronger ramper than Athlon. When Willamette reaches volume production, then (and only then, and this is assuming AMD is weak in the process department, and this is also assuming that Flash suddenly decides to level off in demand) AMD's profits will start to wane. It will likely not be a sudden, one quarter jump to constant losses, but the dropoff may be significant and the stock could suffer in this scenario; it could suffer strongly.
When will Willamette reach this stage? Well, Intel is expected to "introduce" Willamette in the Fall of this year. But they have publically admitted that only hundreds of thousands of parts will ship in y2k. With six times the fab capacity, it is actually surprising that Willamette will scale more slowly than AMD's Athlon did last year, but that's immaterial to the discussion.
Willamette will reach the point of volume shipments at the high end probably some time in Q1. Not before. And even if their ramp is above schedule, Willamette will be overly costly (large die, etc..) and will not strongly impact AMD's bottom line until Intel gets it shrunk a little bit (or finds some other way of managing costs).
Essentially, AMD has a totally clear y2k. Intel will not be able to push down their profits until they have a seriously competitive core, and though Intel will have "bragging rights" for the Willamette as early as, say, August or September, they will not have "revenue rights" (interesting term that Jerry Sanders recently coined) until at least January 2001.
Anybody going back to the old "but the management sucks!" routine is being silly. The management has always had worthless crap to work with. AMD's always been the "second source", the "clone" guy, bereaved of resources necessary to make it a major player. Whether or not the management has given up opportunities in the past -- whether or not they've purposefully attempted to destory themselves over and over again -- is irrelevant. Their actions over the past two years have been almost perfectly in tune with the necessary measures required to compete against a company much larger than themselves. And those actions have bought them a free year of profit. After 2000 all is uncertain, and if this uncertainty remains towards the end of the year, I do plan to relieve myself of the shares I purchased to take advantage of this completely obvious and foreseable strong year for AMD. But for now ... no, AMD is in no risk at all.
Oh, crud. I just wrote a novel, didn't I? Okay, apologies for that. Please feel free to remove this message if you deem it too long. I just hope that the stuff I wrote here can clear up some issues regarding AMD, and I hope I didn't come off as a total fool.
-JC
PC News'n'Links
jc-news.com
IP: 209.9.36.36
TaeKim
Web Admin posted 04-11-2000 13:43
--------------------------------------------------------------------------------
JC, with your permission can we publish it as an article? Seriously, just give it a title.
------------------
Tae Kim
Founder, Editor-in-Chief
TigerInvestor.com
tigerinvestor.com
IP: 208.218.153.234
tigerinvestor.com
Milo |
