Before you read this, please have a look at the earnings and read the webcast transcript:
http://www.amd.com/us-en/assets/content_type/DownloadableAssets/Q107Results.pdf
http://seekingalpha.com/article/32901
1. CPU biz revenue is down 31% YoY and 38% QoQ. Not only has there been ASP decline but significant share loss. iSuppli indicates a massive 4.5% share gain for Intel though I'm sceptical it's such a large swing in 1 quarter. Intel would not have bet on such a large gain and considering ASPs are stable they would have come in above their guidance if this had happened:
http://www.fabtech.org/content/view/2739/2/
(Thanks Ho Ho for the link)
2. Inventories are up 15% from 814 mln to 937 mln. The explanation is this is desired build up in the channel to ensure stable supply and that these are products the market still wants. There are a couple of gaps in this theory. First - if the market wanted these products they would not have lost share and missed their unit targets by such a large margin. Second, 40% of this build up is 65nm parts. This doesn't make sense. The first thing you'd want to sell off are lower cost 65nm parts which means takers for the parts are hard to find. Essentially if you add 30% gross margin back to the inventory number AMD already has all the parts they need to fill their Q2 demand. Last year when they were making money inventory was running at around 30% of revenue. Now it's running at 76%.
AMD has two choices here - drop prices further to clear inventory. In fact, they might start giving away some parts at cost simply to re-gain market share which prevents Intel from making a corresponding sale. Alternatively, they will have to take a big write down in the next couple of quarters.
3. Accounts receivable are down QoQ from 1.14 bln to 667 mln which is more in line with what the number was through most of last year. Which means even units in the channel have been worked through and yet inventory is building = lowering demand for parts.
I am extremely sceptical at AMD's confidence on the quarter being flat to up. They are displaying their naivete or mis-leading the market yet again because in a seasonally down quarter what this means is they will have to re-take significant share. However, Intel will not be standing still and the impending Core 2 price cuts along with the Conroe single cores being introduced as Pentium and Celeron parts will squeeze AMD. Specially since the die size on that single core part is going to be super small making it super easy for Intel to price AMD at the lower ends if necessary.
Overall, Q2 will continue to be tough for AMD. There is no reason for them to re-gain share unless they start giving the product away at cost. They may see some desktop share recovery in the channel but the revenue upside could be offset by having to lower prices even further as Intel cuts prices on their Conroe parts for the first time. They will feel either share or revenue pressure on mobile with Intel's new Santa Rosa launch coming up as the OEMs replace older SKUs which means with nothing new from AMD, some of their Turion based SKUs may get bumped off. Server is also all pricing for them now...if they can go low enough then they might get some share back there. But Intel will not give in so easily and what I would be expecting AMD to do is try and stall the market to wait for Barcelona. Unfortunately, the absence of a public demo till date makes their claimed performance numbers unverifiable.
I think AMD will have to make a choice on whether they try and re-gain share or hold ASPs and hopefully revenues. Either way, up is unlikely. Flat is possible if they get relief from the new ATI products. While it was unsaid in the Intel call, I can bet you Intel has tasted blood and will continue to try and re-gain market share using their upcoming price drop.
At the end of the day, Intel's real advantage is not just they have a kick-ass new product, it is they have a product that is significantly cheaper (smaller die size, more flexibility due to common uarch, etc) to manufacture running on a process that is close to a year ahead of their competitor. This gives them tremendous room to navigate on pricing/performance/watt and that's what the market has become about.
8 comments:
Good financial analysis 180!
One significant turn of events that a lot of financial analysts have been overlooking is the fact that Intel has become so competetive product and process wise that AMD's business model no longer works. For the first time in history, AMD's loses are so massive that Fab improvements and upgrades are affected.
The shift is so great that now we're hearing AMD being serious about subcontracting (asset lite).
While this move might allow AMD to survive long term, this will definitely limit AMD's capability in gaining any significant market share. Subcontracting will make AMD become less nimble and will be at the mercy of another foundry. This tell's me that AMD has completely given up on the 30% market share idea.
May I request you to rename your blog to something other than "Sharikou 180"? Sharikou is obviously a fansite, Scientia less so - this is possibly the only balanced (possibly Intel slanted) viewpoint.
180 - good analysis overall but I'm not sure some of your inventory comments are accurate.
Inventory is generally quoted as the sum of finished goods, work in progress and raw materials. I haven't looked at AMD's inventory specifically but in Intel's case only about 1/4 - 1/3 of the reported inventory # is actual finished product. Given the cycle time of wafers through the fab I would expect AMD to have a similar ratio of finished goods/overll inventory #.
I would suspect part of the inventory buildup is additional WIP as F36 continues to ramp, you also now have a full quarter of whatever graphics inventory (though this should be minimal impact - as this is outsourced thus no WIP, raw material, just finished products). I would be far less concerned about the inventory buildup - espwcially given the other #'s. I haven't run the #'s but I doubt if the inventory is a full quarter of finished product needs.
Also the isupply #'s are I believe overall market share (not just x86) and I believe they are revenue share, so these #'s are a bit different from the Q4 75%/25% market share split. Though a 4-5% drop in 1 quarter regardless of how the # is quoted is rather shocking.
"They are displaying their naivete or mis-leading the market yet again because in a seasonally down quarter what this means is they will have to re-take significant share."
They could be correct - you are assuming ASP's are flat to down, while everyone focuses on desktop pricing mobile continues to grow (and was the one area where AMD actually did well) - AMD may be thinking this (or server) grows in Q2. This will also be pretty much the first full quarter where AMD would see 65nm production savings. I don't think flat is completely out of the question (though I wouldn't bet on it).
"it [Intel] is they have a product that is significantly cheaper (smaller die size, more flexibility due to common uarch, etc) to manufacture running on a process that is close to a year ahead of their competitor."
This is what almost AMD fan likes to believe is not the case, but it is unfortunately a fact of life. AMD try to break this advantage thinking they could get 3 fabs up in the next 5 years (F36, conversion of F30-F38 and NY fab). However until AMD is stable financially AND they have 2-3 fabs up and running Intel will always have the production, and barring any product breakthroughs, pricing advantage. Unfortunately AMD seriously misjudged the situation (arrogance?) thinking they could acquire a company that was 60% of their size ramp 2 fabs, start building a 3rd and at same time remain profitable.
I really don't think the average fan understands the process technology differences. Folks like Sharikou, Scientia like to quote the gap as the time when Intel release product on a given node and the time AMD does. And Scientia will tweak it even further by using the time AMD releases their first product, and use the time he thinks Intel has significant volume (thus he likes to claim Q4 for Intel and Q3 for AMD - when it really will end up being Q3'07 vs Q3'08). However even this is only 1/2 the story....
Intel's process when they release product is fully developed and achieving all of the process technology specs (they work on yield and some minor process improvement tweaks after transfer to HVM, but overall the process is hitting all of the key tech metrics)
AMD on the other hand goes with the CTI (continuous transistor improvement) approach. At the outset of a product release on a new node, their process is NOT (I repeat not) at the mature technology targets and it is another 6-12 months before it gets there. Thus while on paper it seems AMD is only a year behind on 65nm, they are actually a bit more than that.
The next obvious example of that is 45nm - the initial AMD procss will essentially be a "shrink" of 65nm with minimal transistor improvements save for the power savings due to the lower Vt's on the 45nm node. You saw this on 65mn - power on AMD's initial 65nm products goes down but speed does not go up - over time it will as they start implementing these improvements. Fans like to say they were "focusing" on low end parts, but if 65nm were hitting process specs AMD would be much better doing the high end parts on 65nm to keep TDP in check (or lower it). The process just wasn't there (in terms of transistor performance) at the release.
Now look at 45nm - AMD has said they will at best implement high K/metal gate at the end of the tech node (late 2009/early 2010). This will likely mean minimal transistor speed improvements until then. Thus when 45nm is released, just like 65nm, you will see power improvements, but no real clockspeed improvements (for a given architecture). People may also start saying what about the low K claims on the backend - basic problem here is that switching speed is still mainly gated by the front end process not the backend (metallization). AMD will need the lower K dielectrics in the backend when they get around to the transistor improvements in the front end, but without both they will not see significant transistor speed improvements.
To wrap this long post up - Look at the Barcy roadmap, there is very little speed bump on the upper end part is expected between release in Q3 this year and Q2'08 next year.
Now look at Intel's 45nm shrink which looks like will start at at least 10% higher clock (3.33), this is because their technology improvements are put in up front as opposed to mainly just shrinking the linewidths and doing the transistor improvements later. This is obviously harder to do (you are cramming more development into a shorter time period), but it shows how strong Intel's process development group is.
"Subcontracting will make AMD become less nimble and will be at the mercy of another foundry."
Good points roborat - it also makes AMD less profitable as Chartered has to make some money as well. This would give Intel a bit more pricing power as AMD also will be paying a "foundry cost" on the chip. It's not like Chartered are selling these chips at cost to AMD...
It will also add other significant costs unless they stick to the Chartered/IBM /et.al consortia as there would be a significant development cost to outsource the design on a different 65nm (or 45m) process like say TSMC.
My guess is AMD will stick with F36 as main production and use it for 45nm as well and then try to negotiate some upside capacity at Chartered should they need it. I think the current financial situation kills the NY fab (which is a good thing for AMD as I don't think they needed it anyway), but it also probably kills any ideas of trying to bring the ATI product manufacturing in housw. I would also guess F30 will start the 300mm conversion and be used for any unique 45nm tooling that they don't have in F36 (like more immersion litho tools and ALD tools for eventual highK/Metal gate work) or to provide more upside capcity if things turn around. I would think they would just do the conversion, put the wafer handling system in place and just hold off on tooling (which is the major expense of the conversion anyway)
Nice posts (Anonymous) guys - illuminating and informative!
cheers...
It is worth noting that AMD's current inventory levels are about 75% of their quarterly sales.
I think Intel at its worst was only up to about 50% levels.
I can imagine the catch 22 situation the AMD management is in. If they ramp Barcelona too quickly, they will have to practically sell old inventory at scrap prices. If they ramp up too slow, they will continue to lose market share.
Going to a foundry is not all that bad, as TI. Asian Foundries, being in located in low cost economies, have a way cheaper cost model than a US/Europe based Fabs.
On the contarary, the Graphics biggies have had most of their product released delayed to some extent. Is it because of their foundry based Fabbing model ? Don't know !
Post a Comment