AMD releases Q2 financial results, and Ontario details galore

40nm supply problems persist

Another day, another quarterly earnings conference call.  AMD came in with their latest and greatest batch of numbers, tidbits of product information, and future guidance for the company.  While your man was unable to squeeze any questions in this time around, (and apparently vuvuzelas are frowned upon during these conference calls for some reason) there were still plenty of interesting nuggets of information dropped along the way, so let’s begin.

AMD is reporting revenues of $1.65B which is 5% higher than Q1 this year, 40% higher than Q2 of last year, and is apparently a record for the company. They are also reporting a net loss of $43 million, or $0.06 per share, and operating income of $125 million. The company reported non-GAAP net income of $83 million, or $0.11 per share, and non-GAAP operating income of $138 million. The discrepancy between GAAP, and non-GAAP numbers is attributed to AMD’s stake in Global Foundries, and how it is reported internally vs. how GAAP dictates that it be recorded.  The bottom line is that AMD had a respectable quarter from a financial standpoint.



 Much was discussed about AMD’s first upcoming Fusion product.  While most of it was information we already knew, here is a recap of what was discussed about the new APU.

  • It will be initially built at TSMC on the 40nm bulk process  (Same as the HD 5000 series GPUs)
  • It will be the “low-end bookshelf product” of the Fusion line , with Llano capping off the high end for mainstream computing
  • Targeted towards netbooks and SFF PC’s  (Tablets need not apply)
  • Will launch ahead of schedule in Q4 2010
  • Dirk Meyer specifically declined to offer up GPU die area numbers for Ontario, however enough other information was dropped that we can make an educated guess in this matter:
    • “Performance will be essentially what you see in the low-end mainstream notebook chips today”
    • “Ontario will take up half the die area of today’s chips”
    • A “low-end mainstream” chip today is the Athlon II Dual-Core Mobile at 117mm^2
    • Half of that would put the die size of Ontario at around 59mm^2
    • Ontario is AMD taking aim at Atom, and the market segment it currently monopolizes
    • Graphics performance of Ontario will be a “Game changer” in the industry (according to AMD)



Not as much was said about Llano during the call as its smaller brother Ontario.  Llano will be the mainstream component to AMD’s Fusion platform plans, finding its place in both notebooks and desktop systems.  Llano will bring improvements to both CPU power, and GPU power  (remember that both CPU and GPU are built into the same piece of silicon, henceforth known as an “APU”.)  The big news with Llano is that it is being delayed by a “couple of months” and will be released in the rather ambiguous time frame of 1H 2011.  The reason for the delay was the company acknowledging problems with the learning curve for GloFo’s new 32nm SOI HKMG process technology.  They are producing chips on this process, but yields are not good enough to consider launching the chips for revenue any time soon.



In Intel’s recent financial conference call, much of their record success in Q2 was attributed to a massive upswing in server chip sales.  The same is not true for AMD according to this recent call, and Dirk Meyer seemed positively livid about it.  Their recent Magny-Cours Opteron 6000 series launch was apparently marred by OEM’s dragging their feet in bringing AMD server designs to market which caused AMD to miss out on much of the recent server buying spree.  Meyer acknowledged that in the coming quarters AMD Opteron chips will be crawling out of a market share hole that they have not experienced for years.  They expect server sales to be on the upswing now that products are available for purchase from server OEMs, but with the strong performance shown Intel’s Nehalem EX chips in the 1P and 2P space, AMD has an uphill battle on their hands in the coming quarters.



While AMD’s graphics division was applauded during the conference call, raking in $440M in revenue last quarter, the company admitted that they were still supply constrained in their GPU division.  To date they have shipped 16 million DX11 GPU’s, but demand continues to outstrip TSMC’s ability to crank out wafers.  An interesting question was thrown out during the call, asking how come ASP’s (average selling prices) weren’t as high as they could have been on GPU’s given that typically when demand for a product is greater than supply prices tend to go up.  According to AMD, they signed on a number of new notebook OEMs  (Sony being the most notable company to use AMD chips again), and since notebook manufacturers solder GPU’s onto the motherboards of their units typically, they can’t ship systems if they are missing the GPU component.  That being the case, AMD allocated more production capacity to mobile GPU products (and away from the higher end GPU’s which garner higher ASPs) this past quarter to enable notebook makers to ship more AMD based systems.

The continued tight supply of chips coming from TSMC is worrisome, given the fact that Ontario is launching later this year  (along with Southern Islands) on the same 40nm TSMC process node.  It’s a safe bet that AMD is running quite a few test wafers of both new chips through the lines at TSMC right now which could be contributing to the shortage of HD5K products, but while the AMD brass did their best to assure the investment firms on the line that these supply problems would ease up towards the end of the year, that has been the story ever since the launch of the Evergreen family of GPU’s and it is doubtful that without another fab coming online for TSMC that the issues will magically disappear.  Dirk Meyer put on a good show to convince people that AMD still hearts TSMC, but we get the feeling that GloFo’s 28nm bulk process can’t come online soon enough.


Global Foundries

AMD still owns a 28% overall stake in Global Foundries in case you were wondering.  Not a lot was mentioned about the chip fabrication facility during the call, but there are a couple points worth mentioning.  The delay in 32nm has got to have several execs at AMD banging their heads on the conference table.  While delays are inevitable in this industry, GloFo is in an especially difficult position right now.  They are responsible for churning out AMD’s current 45nm wafers, while upgrading lines to begin revenue production of 32nm, while trying to attract and retain new clients and build products on the proper nodes needed by those customers.  They are in a transitionary period right now and could perhaps be given a little slack for the time-table slip.  More details about GloFo’s future plans were given at SemiCon West, and can be found here.


Wrap Up

AMD had a strong quarter, and projects a strong finish to the year.  The problem they have however, is that outside of their graphics business, their CPUs have more or less stagnated while Intel keeps tick-tocking away.  You can make a valid argument for the Phenom II x6, or Magny-Cours being a nice refresh, but aside from tacking on more cores, there is not a whole lot more to them from the consumer perspective.  We’ve been anxiously waiting to see this “Fusion” thing in action for a long time now and frankly, instead of “The Future is Fusion”, we need to make that “The Present is Fusion” if AMD wants to remain competitive going forward.  The good news is that AMD seems to recognize this with the “ahead-of-schedule” launch for Ontario.  The bad news is that their mainstream Llano chip will be launched months behind Sandy Bridge giving away a large chunk of potential market share.  Another ominous sign from this call was the utter lack of any Bulldozer talk.  It was mentioned in passing, but never focused on.  Part of the blame can be placed on those asking the questions, as they never brought up Bulldozer, but the fact that they voluntarily skipped over it in their opening remarks is cause for concern.  32nm is delayed, hence Bulldozer is delayed as well, but it would have been nice if they had at least discussed it in the same fashion as they addressed Llano, especially considering AMD’s self-admitted decline in the server world.S|A

The following two tabs change content below.