Joel Hruska of ExtremeTech constructed the above graph to depict how Advanced Micro Devices' (NASDAQ:AMD) revenues have changed between 2012 and 2013. Gartner's data shows that the overall PC market declined ~10% during 2013, whereas AMD's CS (Computing Solutions) revenue declined by just over 20%, considerably faster than the overall PC market.
This contraction in CS revenues forms a central part of the bear thesis against AMD and will likely continue to do so unless AMD is able to garner some more semi-custom design wins.
In this article I would like to explain my recent focus on the PC space and why I believe it will be one of the key points for Thursday's earnings call. I will begin by defining AMD's various markets, followed by quickly sharing the various resources for finding market share data. I will also briefly touch on ways in which AMD has mitigated near term risks, leaving PC market share as one of the main concerns.
Sizing the Markets AMD Operates In
Each financial sum can be found in the companies' respective quarterly findings. These figures are only meant to serve as a very rough approximation of the size of various markets. I've provided the following two sections for the convenience of the reader, but the sections can be skipped as they are not central to the article.
AMD essentially operates in two separate duopolistic markets: Intel (NASDAQ:INTC) and AMD in processors; Nvidia (NASDAQ:NVDA) and AMD in graphics. This makes it fairly straight forward to calculate the total available dollars in each market. Note that these numbers are very rough and should not be used for precise calculations, but are rather meant to serve to give an idea to the relative size of each market.
Nvidia generated $947M from GPU sales in the most recent quarter. Prior to console sales AMD generated roughly $300M per quarter in GPU sales. Including both professional and consumer chips, the entire graphics card market is in the neighborhood of $1B to $1.2B quarterly.
In 2012, Nvidia underwent an accounting change to combine both consumer and professional GPUs operationally into simply GPU revenues. The quarter this change was made (quarter ended October 2012), Nvidia booked $740M in consumer GPUs and $220M in professional GPUs. Based on CFO commentary from both AMD and Nvidia, we know the professional GPU market has been growing, which would mean the professional GPU market is around $300M per quarter.
These numbers were backed up by AMD's CSO Mr. John Byrne during a recent conference. In the conference, Mr. Byrne pegged the professional GPU market between $250M and $300M quarterly.
In the PC space, AMD generates roughly $700M per quarter from PC sales and servers. Intel generates roughly $9B per quarter from PCs and $3B per quarter in data center revenues.
In short, the GPU space is the smallest market at around $5B annually, followed by a $12B or so data center market and a $35B or so PC market.
Note that consoles are a big win for AMD, but I am not touching on them in this article as the topic is extensively covered.
Market Share and Volume Data
The research reports, which can cost several thousand dollars, published by these entities may be cost prohibitive for retail investors such as myself. But various companies, AMD for example, will commission entities such as Mercury Research to provide data for periodic investor reports and internal uses. The conference with Mr. Byrne linked to above, for example, likely uses data compiled using Mercury Research.
IDC is owned by IDG, and IDG operates several popular websites; PC World, for example (read the disclaimer in the PC World article linked to). This means authors of those articles will frequently have access to these research reports, allowing us free access to some of the information in these reports.
The article from Gartner linked to at the beginning of this article puts 2013 PC shipments at around 315M units, or ~80M per quarter. The most recent add-in-board data from JPR puts discrete graphics cards shipments at 15M units per quarter. JPR also releases data on the overall graphics market which includes desktops, notebooks, and integrated GPUs.
Risk Mitigation and Remaining Concerns
Four of the biggest risks with AMD that come to mind are:
- Near Term debt
- Wafer purchase commitments
- Reducing R&D budget
- Competition from Intel and Nvidia
Luckily for longs, AMD has taken steps to assuage fears regarding points 1, 2 and 3.
This shuffling of debt is expected to be neutral from a quarterly interest perspective, essentially serving only to push out near term debt without adding more debt load.
Point 2 requires a little more back ground to explain. 2012 was a rough year for AMD as the company was hit with numerous penalty payments from Global Foundries. As recently as November 2013 bears were citing concerns over AMD's ability to fulfill purchase requirements. I made the point that public statements from AMD pointed to the company moving more products to Global Foundries in order to help meet these requirements.
From the technology conference featuring Mr. Byrne linked to above:
The IP was unique to every product, the methodology was custom to every product and what Mark Papermaster and his engineering team has done is, he has built the IP, so that the IP is consistently re-usable and he is building the methodology, the way they design products and methodology is consistent. So when you look at the PC silicon, it means that, 60% is consistent between IP and methodology, which is a wonderful feat of achievement by Mark and his team.
To translate, AMD can't simply walk from one foundry to another and ask them to make the same product because of differences in manufacturing, unless AMD designs the IP with this ability in mind.
The portability between foundries exceeded my expectations as the words "Diffused in Germany" on the heat spreader below demonstrates. Thanks to ExtremeTech poster "pelov lov" for pointing this out.
One of GlobalFoundries main facilities is in Dresden, Germany. AMD has recently inked a purchase agreement for 2014 with GlobalFoundries, indicating that various products will be moved. This will help fulfill the purchase requirements, and the release of socketed Kabini chips being built at GlobalFoundries shows that some production has already been shifted.
Point 3 is more open ended. For example, designing IP from the beginning to be portable between foundries may require more work up front, but then saves money on the back end by preventing engineers from having to go through the re-work of porting the design.
As another example, one thing that I particularly like about the AM1 socket is AMD is taking a product already in the company's portfolio, a notebook chip, and applying it to the desktop market. This means a larger market for Kabini with minimal overhead.
Finally, point 3 and point 4 go hand in hand. Monopolies are frowned upon, so most companies operate with competition. The fear from the bear side is that reduced R&D budget will prevent AMD from keeping pace with the competition. Bulls see a reduced OPEX budget improving operating margins and profitability. In essence, points 3 and 4 can be argued from either perspective depending on the case one is trying to make.
Tying the Argument Back To PC Concerns
AMD's desktop market share has been relatively stable, with the company releasing the desktop version of the Kaveri processor in January, as well as the low powered AM1 processors this month. Note that the AM1 desktop platform will likely have minimal impact to Q1 operations based on timing of the release. However the release of these two new platforms should help to provide some stability in the PC space in the near term.
Contrast desktops to the notebook space where AMD lost almost 5% market share between Q3 and Q4 and you can see the concern.
Prior to the previous earnings report, I noted that some caution was in order, particularly as there were no new compelling products that would contribute to CS revenues. With only 2 SKUs of the desktop version of Kaveri launching this quarter there is still a lot of uncertainty in the PC space.
Furthermore, you can see by looking at market share data that AMD's market share actually increased from 4Q12 to 3Q13 prior to the hit in Q4.
The graph above doesn't show 2Q13 to 3Q13 data, but I had pointed out in a previous article that during this time AMD lost slight market share to Intel, which subsequently corresponded to a large hit to the share price.
From an investment standpoint, this is important to understand because AMD was given no credit during Q2 when the company gained significant market share against Intel. However, in both Q3 and Q4 AMD's share price took a hit due to losing market share.
Finally, despite diversifying away from PCs, the PC market still remains AMD's largest in terms of total available dollars. The PC market is also important from an R&D standpoint, as IP developed for PCs and servers is then used in semi-custom projects, meaning operations in Computing Solutions needs to fund the R&D to make the semi-custom business model a viable strategy.
Soft Notebook Sales Combined With Inventory Reduction Set the Stage for a Weak Q4
To highlight part of this uncertainty I would like to quote CEO Mr. Rory Read from the Q4 conference call (my emphasis added):
The consumer notebook market remained soft in the fourth quarter. We focused on improving mix and reducing downstream inventory with our customers. We have secured a number of premium notebook design wins for Kaveri and also have solid adoption of our next generation low power Mullins and Beema SOCs, which deliver twice the performance per watt of our previous offerings.
During Q4, Mr. Read stated the notebook market was soft and that AMD was trying to reduce downstream inventory. Reducing inventory is likely done in preparation for new product launches. AMD's Richland platform was essentially a tweak of Trinity.
Source: AMD via Engadget
Original roadmaps show Kaveri as a 2013 product, but it seems Richland was launched as a stop gap while AMD worked on getting Kaveri out the door. The importance of this is that AMD has been essentially using a 2012 product for nearly 2 years. OEMs need to burn down Trinity and Richland inventory in order to pave the way for mobile Kaveri.
I believe we witnessed something similar in regards to Intel with the roll out of Haswell based ultrabooks.
In short, Richland and Trinity inventory clogging the retail channel, combined with the weaker PC market, led to weakness in Computing Solutions.
The Timing Of Product Releases and Performance Is Uncertain
Extending the uncertainty from Q4 is the fact that AMD hasn't given any definitive timelines for the upcoming product releases.
I emailed AMD several weeks back trying to get confirmation mobile Kaveri, Beema, and Mullins were still tracking for a first half launch. The response I received indicates that we should see these products launch during the first half of the year, but no specifics were mentioned as to exactly when, other than to expect to see these new products show up with "typical OEM refresh cycles." Also expect to see the OEMs, not AMD, announcing the design wins.
The issue here is the quarter doesn't end until June, and there is a major electronics trade show (Computex) scheduled for the beginning of June. Typically we will see numerous product launches at Computex, so if these product launches align with Computex, it's unclear as to how much these new products could impact Q1 revenues specifically.
Source: AMD via TechReport
There is very little technical information regarding these upcoming mobile products, making it harder to predict where they will fall out from a performance perspective. I managed to dig up what I believe to be a little information on the upcoming Mullins chip, and I believe the information I found, combined with data from AMD's testing labs, points to a solid chip.
Regarding Kaveri, the desktop variants of Kaveri launched to mixed reviews, but reviewers almost unanimously claim that the 45W desktop version of Kaveri is the most interesting. To quote Joel Hruska of ExtremeTech, "the 45W A8-7600 offered 80-90% of the A10-7850K's performance..." Thomas Ryan of SemiAccurate has an excellent write-up of the 45W Kaveri with an emphasis on the performance per watt of the chip.
Because of these reasons, I believe the new low powered APUs, along with mobile Kaveri, have the potential to give AMD a nice performance boost in mobile. Based on recent market share woes in the notebook space, this is where the performance increases are most needed, in my opinion.
While I don't disagree with contributor Sean Chandler's article (a very good read, by the way) that both console sales and GPU demand could be a boon on Thursday, I defined the size of the GPU market to give some perspective to the discussion.
Although GPUs could be a positive, the relative size of the GPU market, combined with AMD's limited GPU exposure in notebooks, cap the upside potential surprise from GPUs.
Other positives for AMD going into this earnings report are the debt restructuring to convert short term debt to long term debt, along with shifting production to GlobalFoundries to ease wafer purchase concerns. In my opinion, these are also very large positives, so don't let the relative length I have devoted to them detract from their importance.
Shifting to PC sales, bear in mind that IDC tracks vendor data. Just because Dell, for example, may have seen improving PC sales doesn't necessarily mean the benefit will be reflected in AMD's financials. Although there is a chance this is the case, it's far from a guarantee.
Jon Peddie Research indicates that its reports serve as leading indicators, which makes sense logically. JPR tracks chip sales, not PC sales. An OEM may sell a PC using an AMD APU, but that APU may have been booked in a prior quarter if the OEM is depleting inventory.
Looking at recent quarters of market data on JPR reveals AMD has exhibited divergent trend lines, with notebook APUs declining and desktop APU shipments increasing, but with notebook decreasing faster than desktop is increasing. Being able to reverse the notebook trendline will be important in placating concerns.
To draw parallels between the launch of mobile Kaveri and Beema/Mullins to the Jaguar platform last year, there was little public information released regarding Jaguar during Q1 2013. However, during the Q1 2013 conference call Mr. Read disclosed that volume shipments of Kabini had begun. Design wins with Jaguar APUs started coming to market during Q2. Further, these new products drove increases in market share.
It's not sage wisdom that AMD, Nvidia, and Intel release chips that find their way into computers. AMD released the mobile Jaguar chips in the beginning of the year and the company gained market share. Intel released Haswell chips in June, with more widespread notebook availability in Q3 and Q4. AMD, having no proper answer to Haswell as Richland was more of a refresh, combined with Bay Trail adding competition directly to Jaguar, served as a recipe for a loss of market share in the back half of 2013.
I often focus on performance of AMD's products because it is the technology that drives the revenue, and it's often this piece of the puzzle that's slightly harder to understand. Furthermore, I don't feel AMD gets a fair shake from many OEMs. Couple this with lazy reviewers or non-tech savvy consumers, and AMD products get a negative connotation for being "slow" or "cheap."
For example, a consumer may compare an AMD notebook to an Intel ultrabook. Ultrabooks typically have higher resolution screens and either a hybrid SSD/HDD or an SSD. AMD designs are also often hamstrung with smaller batteries. As I'm trying to keep this article less technical, just know hybrid drives or solid-state drives are typically much faster than standard hard drives; more so than most consumers realize.
Likely if you simply swapped out the processor from two otherwise equal designs, most consumers would be hard pressed to tell the difference. Yet because of the platform differences, not processor differences, the AMD powered device may receive lower reviews.
AMD's VP of Channel Sales Mr. Roy Taylor published a write-up on AMD's website entitled "The Real Price of Productivity." In it he essentially points out that many corporations overspend on processing power they don't need.
To tie this back into why I specifically focus on AMD's technology, it's because I do not believe most consumers understand the technology they are buying. Intel's brand recognition goes a long way, so if a consumer walks into a retailer to purchase a PC, they may not necessarily know the AMD PC would be better suited for their particular need or budget. Instead, they may choose Intel simply because it's a brand they recognize. This leaves AMD more likely to be purchased by the tech savvy group that understands the technology.
Because there's little performance information available on the upcoming mobile platforms, it's hard to judge their likely competitiveness.
In summary, GPU sales and console sales seem like likely positives for AMD on Thursday. However, similar to last earnings report, there is little information regarding product announcements and design wins that will add specifically to CS revenues.
If taking a short term view on AMD, there isn't really enough information to get a good read on what to expect from CS revenues so some caution is warranted. Intel's earnings call may give some clues.
If AMD has begun volume shipments of the next generation mobile products in preparation for an impending launch, it is likely CS revenues could at least be inline, if not surprise. However, if OEMs are continuing to deplete inventory and the next generation products are launching late in the quarter, CS revenues could come under further pressure.
From a longer term perspective, we should get some information on where AMD stands with its new mobile platforms and hopefully an idea of when to expect these platforms to come to market. Because of the recent decline of notebook market share, and the relative size of this market compared to AMD's other ventures, these products will be important for the balance of 2014.
Disclosure: I am long AMD, INTC. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article. I own both shares and options in AMD, and actively trade my AMD position. I may add to or liquidate a portion or all of my position at anytime. I may also liquidate my INTC shares within the next 72 hours.