AMD held a financial analyst day this past Thursday, December 13, in New York City. One of the main themes of the presentations, by key executives, was that of moving forward from a difficult 2007. Yet, despite the ‘glass half-full’ mentality that was prevalent throughout (indeed, one of the execs used those words exactly) and somewhat apologetic tone, there was too much excuse-making and, surprisingly, a lack of transparency on their future plans for profitability. We’ve already seen what a lack of transparency and uncertainty can do to a company’s stock – take a look at Citigroup. For a long time, the financial community was on edge regarding Citi’s structured investment vehicles. Not knowing what material impact moving those investments onto Citi’s balance sheet would have made investors nervous. And no company wants nervous investors. Especially not AMD.
I don’t plan on summarizing the entire multi-hour webcast. For that, take a read through of Ars Technica’s succinct article. The following are some key points I drew from the presentations and a bit of analysis, both technical and financial. Let’s start it off with the overshadowing stories of recent days, Barcelona/Phenom and the TLB errata.
I didn’t follow the Barcelona launch too closely, except to note that AMD would have a much tougher time in the 2S space, while retaining some of the advantages in the higher socket number servers, due to the inherent advantages of its HyperTransport architecture. However, Phenom’s launch I did follow with great interest.
Perhaps most disappointing was the way AMD handled the enthusiast community. Instead of providing review samples to the various review sites, AMD wanted to ship everyone to Tahoe and have all the benchmarking done there. While this sort of event may have been appropriate several months ago, this was to be the case at the point of retail launch. Whether it was intended to ensure more favorable reviews, or perhaps just ensure that people weren’t going to be comparing the Phenom 9600 to a QX9650. Regardless, when a large tech site like AnandTech outs them, it paints an ugly picture. While the enthusiast community is relatively small, compared to the OEM and even retail businesses, an enthusiast is typically the neighborhood consultant for all things computers. They have more say than over a single CPU or video card.
I do agree mostly with what was said throughout the presentations about the TLB errata – errata aren’t uncommon in processors and the whole situation was perhaps blown a little bit out of proportion by fear-mongers. Yet, the situation will be damaging both from a financial and image point of view. Barcelona was already a bit late to market and now with the TLB issue, AMD has had to selectively ship their new server processors to customers, while rejecting others. This is obviously the best they can make do with the situation at this point; however one has to wonder how smart it is to ship the desktop variant, Phenom, to customers without so much as a warning label about the TLB issue. One of the places the errata can rear its head is in virtualization situations, and while I wouldn’t say it’s prevalent among desktop clients, it is definitely gaining traction among end-users. Until the revision with the hardware fix ships (late Q108), a patch is being used to overcome the errata.The BIOS work-around being send out was recently benchmarked by Tech Report. The results were far from pleasing.
Furthermore, unlike the RV670 launch, which I will talk about a little later, Phenom was launched with the intention of taking Intel’s quad cores head on. In fact, between the new ACP (Average CPU Power) measurement, the ‘true quad-core’ design tag-line and pricing, you’d think Phenom was more than competitive with Intel’s CPUs. Unfortunately, this isn’t the case – and most enthusiasts aren’t buying it, both figuratively and literally. It’s well and good to be able to tell the financial community that you’ve raised ASPs by selling to the few that are willing to buy it, it makes much more sense to price it competitively (It seems like AMD’s realized this, with an somewhat unexpected price drop on Phenoms). Speaking of ASPs…
Okay, the one main point here is that this was a financial analyst day, so it makes sense that the financial analysts will want to know certain things. Be prepared to answer them. On numerous occasions, management attempted to push off further details to the Q4 earnings conference call. That makes it seem like there’s something worth hiding.
Large OEM wins, like Dell and Toshiba, were touted as signs of progress for AMD. While acceptance by these manufacturers have opened up a large window of opportunity for AMD, I wonder at what cost? It’s not uncommon for large contracts to come at the cost of ASPs. Obviously Dell isn’t buying AMD processors at nearly the same price as I can buy them. Price wars aside, I can only wonder what sort of margins cut AMD has taken to fill orders from OEMs. Surely, it comes at a cost compared to the relatively fat margins AMD had selling mostly to the channel.
Capital expenditures (which includes R&D) is one area AMD hasn’t gone easy with cutbacks. They are on track for $1.7 billion this year, but plan on cutting that back to $1.1 billion in 2008 – in effect, cutting their spending by 1/3. To note, this does not include any changes in its current asset-utilization plans, meaning the cut is not as a result of further asset-light or asset-smart shifts. AMD continues to work with UMC and TSMC for its graphics product group, while Chartered Semiconductor is helping with multiprocessor units. So the key cuts that will be happening are to R&D and possibly a slower switchover to 45nm at Fab 36.
I think the cuts in R&D have already started to present themselves in AMD’s roadmap slides. Only 5 months ago at the Technology Analyst Day in July, Rick Bergman showed some slides regarding upcoming platforms in 2008. These included a new high end chipset as well as a R7xx series card. The most recent update, a couple days ago, shows that both the chipset and the R7xx series of graphics products have been pushed off to 2009. Instead, 2008 will have to make do with the R680, which is essentially two RV670 chips Crossfired on a single PCB. I think that’s going to make 2008 a pretty dull year for new computer hardware launches.
Where’s the transparency?
When a company’s in a tough position, the last thing it should do is bottle it all up and present a facade to the world – it makes investors nervous. I already mentioned the Citigroup example earlier. In the days when AMD’s Athlon64 X2’s were beating up on Intel’s Smithfield and Presler dual cores, Intel did a massive about-face and showed the world what was brewing its lab (Conroe). While Intel decided to show technology-transparency, there are other forms. For one, financial transparency would be nice to show at a Financial Analyst Day…
There’s no way around it, AMD’s analyst day was lacking in transparency. I was extremely shocked when Mr. Ruiz attempted to sidestep a question from the Caris representative, regarding profitability – whether it was operational or net profitability they were targeting for Q308 and all of 2008. For godsakes, it’s a financial analyst day – that was a straightforward question. It did nothing but make it seem like he had something to hide by not answering the question and postponing it until the Q4 conference call. Later, when pushed, Bob Rivet finally said it’s operational profitability they’re aiming for, meaning not factoring in interest charges (which is important for a debt-laden company).
Presentations are great for a company to tout its wares and sidestep trouble points, but the part of the whole day that I was most interested in was the Q&A, when management would be asked the tough questions. Unfortunately, the Q&A session was a big disappointment. On more occasions than I bothered to count, management belittled the person asking the question, doubting the information they used to formulate questions (I’m not sure how many times I heard management say ‘I don’t know where you got that information, but it is completely incorrect’ and just sit back down without answering any part of the question or address the point) and generally avoiding any of the actual topic points by pushing blame onto the asker. It was absolutely shameful.
Take a page out of your competitor’s book, AMD. Shape up and sprinkle a little more transparency around.
Rocket science and magic
I chuckled when I heard Mr. Ruiz describe processor design as a combination of rocket science and a bit of magic. Yet despite being quite comical, it was also disappointing to see him hiding behind fancy prose. While processor architecture is a very advanced science, it’s the market AMD has been in for tens of years now. Not anticipating the ‘unpredictability’ of the technology is an awfully large oversight for a company that prides itself on an advanced CPU architecture.
I was very surprised at Mr. Ruiz’s barely-suppressed outburst regarding the recent fall in AMD’s stock price. His reference to past achievements by the company were just that, past achievements. The stock market works on predictions and future potential, neither of which have been particularly flattering for AMD. The past gains presented themselves in AMD’s stock price in the past. In fact, looking back five years, AMD’s stock has done better than Intel’s. What do you have to say about that, Mr. Ruiz? And at its highs of the past 5 years, AMD’s stock more than quadrupled, while Intel’s has languished. I bet Mr. Barrett and Mr. Otellini wondered why that was the case, despite making money, hand over fist during that time. It was all about the future and it still is.
The Good – Graphics Product Group
The one bright spot in the day was the previously-ATi group, aka AMD Graphics Product Group. Although R600 (the HD2900) was hardly a smashing success, AMD followed it up rather quickly with the RV670 series of cards, which hasn’t regained the speed crown for AMD, but has provided a very compelling alternative to NVIDIA’s recent launch of the 8800GT/S cards. The key here is, unlike the botched Phenom launch, AMD has promoted it as an amazing price/performance/power consumption package and has priced it at very attractive levels. It provides almost the same level of performance as the 8800GT, but at a lower price. Feature-wise, DX10.1 and the title for first 55nm GPU aren’t exactly ground-breaking, in terms of what it means for the consumer, but they are marketing points nonetheless.
Continuing with the Graphics Group, hybrid Crossfire is very exciting. It is aimed at the lower end of the market. A recent test paired it up with an unreleased card, based on the RV620 core, and saw some very impressive performance improvement. I see the mobile market as the main area of potential for this technology. The laptop can run on integrated graphics alone when working in the OS environment to conserver power, but if a game is fired up, the discrete core is automatically enabled and Crossfired with the integrated graphics.
Hybrid Crossfire currently only makes sense for lower end cards (and hence, the majority of laptop systems with dedicated graphics) since there is some overhead with enabling Crossfire, so higher end cards may actually see a performance decrease with Hybrid Crossfire. For most desktops, if gaming is required, it will probably make more sense to just buy a slightly more expensive card, although there may be a play in the ‘green computing’ market.
I’m sure it goes without my saying – AMD needs promote the idea to mobile OEMs. ATi has traditionally a very strong position in the mobile space, although it stumbled a bit this past year, and this sort of technology is exactly what it needs to be leveraging. NVIDIA also has some form of the Hybrid SLI coming to market, but AMD is the first to show it off and with a 1H08 introduction date, it’s quite possible they will be first to market with it.
This was actually the first AMD Financial Analyst Day that I watched. It may seem like I’m being hard on AMD, but to be honest, I’m just being a little selfish. The fact is, without AMD (or even a vastly diminished AMD), we (consumers) lose. Lower R&D means slower product introductions, which means slowed update cycles and fewer price cuts that have driven the industry.
The striking cutback to capex (such a crucial component of a microprocessor company) leaves me with the impression that the situation at AMD is a little more frantic than management let on – let’s call the glass one-quarter full?