AMD: Consoles, Kaveris, And Wafers

| About: Advanced Micro (AMD)

It has been a busy quarter for Advanced Micro Devices (NYSE:AMD) so far. Console sales are rolling in, Kaveri is set to ramp at the end of the quarter, the company hosted its annual developer conference, and a $2.50 target price has been reiterated.

This article will include an analysis of the revenue stream being generated by consoles, the mix of upcoming products with a focus on Kaveri, and the potential outcomes of Sanford C. Bernstein's concerns over wafer commitments.

How Much Is AMD Going To Make On Consoles?

If you pull up AMD on the iPhone stock app, the first two articles you'll see (at the time of writing) are an article on Barron's regarding wafer purchase requirements, along with an article on WSJ entitled, "Console Bet Played Out At AMD." I am not a fan of sensationalistic titles, and I would say the WSJ article qualifies since only one of the consoles is currently shipping. I'm not quite sure how something is played out before it starts.

Sony (NYSE:SNE) had a record launch with the PS4, to the tune of 1M units within 24 hours. Microsoft (NASDAQ:MSFT) releases its Xbox One this week. Sony is expecting to sell 3M units this year (Sony met 1/3 of this goal within the first 24 hours), and 5M by March 2014. Colin Sebastian of Baird is forecasting a similar 2.5M-3M units for each next gen console by year's end.

Assuming both console makers are able to meet an estimate of 5M units each within the first ~6 months of launch, AMD can expect to ship 10M+ total APUs during this time to the console makers.

According to iSuppli (via Forbes), Sony pays ~$100 per APU. Chipworks states the chip is 348 mm^2, and is "likely" built on TSMC's 28nm HP Hi-k metal gate process.

Image Courtesy of AnandTech

The Xbox One APU is built on the same process, and is slightly larger at 363 mm^2, meaning it is likely more expensive. But for ease of calculation I will use iSuppli's estimate of $100 for both console APUs.

To tie this all together, consoles have a fairly long life cycle, and during the Q3 conference call AMD CEO Mr. Rory Read stated sales typically peak during the third year. During the same call, the company CFO Mr. Devinder Kumar put console operating margins in the mid teens, and stated in the Q&A session they could rise as volume ramps and cost structures improve.

A little arithmetic based on the assumptions laid out above forecasts console revenues for the first ~6 months of sales at ($100/unit * 10M units)

At 15% operating margin, this equates to $150M going to the bottom line. Here is where I would like to look at a little fundamental analysis of console sales.

Source: Google

From the bottom in November 2012 until April of this year when the Street took note of console wins, the market priced AMD around $2.30 on average (rough estimate based on eyeballing the chart). Currently AMD sits ~$1 off this low.

Using 750M shares, the $150M estimate of profits from console sales during the first 6 months (note, not full year), we arrive at an EPS of $.20. A $1 rise in share price based on $.20 EPS corresponds to a P/E of 5x.

Electronic Arts (NASDAQ:EA) also predicts 10M total next gen console units sold by March of 2014. If Sony and Microsoft combined reach 15M total units sold within the first year, based on the same estimates as above this would correspond to a revenue of $1.5B, or consoles contributing $.30 EPS over a full year (note this year of sales in this example does not run with the calendar year), which would correspond to a P/E of 3x based on a $1 rise in share price.

I am not using this calculation as a basis to suggest a price target using sector multiples, but rather as an indicator to gauge how much the market appreciates the earnings power of consoles.

Wafer Concerns

Image Courtesy of ExtremeTech

Recently, Dr. Stacy Rasgon of Sanford C. Bernstein reiterated a price target of $2.50 for AMD, citing concerns of a remaining ~$400M of the $1.15B required for wafer purchases during 2013. The concern isn't necessarily meeting the $400M requirement, but rather the potential for subsequent inventory write downs going forward. He also noted that console chips are made by TSMC (NYSE:TSM), not GlobalFoundries.

Prior to AMD's APU 2013 event I published a very detailed article regarding the upcoming Kaveri architecture, and in this article I had pointed out one of the importances of Kaveri is that it would be one of the few products produced at GlobalFoundries that could satisfy purchase commitments. GPUs and the Jaguar chips appear to all be produced at TSMC. GlobalFoundries produces 32nm CPUs and APUs.

One thing not mentioned in the concern is Kaveri production, which based on leaks (which may turn out to be incorrect) is set to begin in December.

Source: TechPowerUp

Based on this information, I believe 1 of 4 outcomes is likely, with 4 being the most interesting:

  1. AMD has inventory rise between Q3 and Q4 in accordance with Bernstein's concerns
  2. AMD incurs a penalty and reduces wafer requirements for Q4 in lieu of building inventory
  3. A combination of 1 and 2
  4. AMD manages the situation inline with the comments given during the Q3 conference call, and meets wafer purchase requirements without increasing inventory

If you take one of the first three scenarios laid out above as a worst case, this still looks much better than 2012. Regarding expenses surrounding the restructuring and reducing of wafer commitments, AMD paid roughly $700M in March 2012 and took another $165M (see the footnote in the link for details) charge in December 2012.

During the Q3 conference call, Mr. Kumar stated there should be an announcement regarding wafer pricing for 2014 somewhere between 30 and 60 days after the Q3 conference call, which means we should have a more solid answer sometime between now and the middle of December.

Exploring Scenario 4

Scenario four is the most interesting. There are two potential, simple explanations. The first would be to manufacture more products at GlobalFoundries.

In an article on Anandtech, Anand summarizes AMD's future direction as trying to create easily synthesizable architectures.

Turns out one of the design goals of GCN (AMD's IP blocks that are used to make GPUs) is to create an architecture that can be scaled to fit different needs. These presentations I am quoting are from 2011 and 2012, and in 2013 we see their fruits. AMD has used the company's Jaguar core architectures now in both Kabini and Temash processors, as well as the PS4 and Xbox One APUs. And the GCN cores pictured above are used in the company's GPU line, and are placed alongside the Kabini/Temash/console APUs, and are to be featured in the upcoming Kaveri chip.

The above image is taken from AnandTech, and in the article it is stated that Kaveri should be available for retail starting on January 14th.

Exploring this a little further, the thought now is that Kaveri is taped out and is being produced at GlobalFoundries. Historically, AMD GPUs have been produced at rival foundry TSMC. The Radeon 6000 series was based on an architecture known as VLIW4, and was produced on TSMC 40nm. These VLIW4 cores were ported to GlobalFoundries to launch various APUs such as Llano, Trinity and Richland, but they were laid out on a specialized process (32nm PD-SOI) which was more designed for CPUs. By the time these APUs were up and running, AMD was a few months away from rolling out new graphics cards based on GCN so it made little sense to try and ramp a GPU line that was being replaced, until now.

With the legwork done to manufacture GCN cores at GlobalFoundries on a more standardized process, it is a little more feasible to move some limited GPU production to GloFo in order to help meet wafer requirements.

"As part of our prior amendment, we are also committing to make some GPU products at or below [the] 28nm technology node at GLOBALFOUNDRIES." commented Devinder Kumar on the situation.

The above comment, taken from BSN, is from the December 2012 conference call concerning the WSA amendment.

So, if AMD has moved some GPU production to GlobalFoundries, this will assist in fulfilling purchase requirements. However, I think this is a bit of a stretch to expect this to be the case for Q4, but has the potential to alleviate some concern going into 2014. While cutting OPEX, AMD has limited resources, and has worked on two console APUs, Kaveri, Kabini/Temash, the new Hawaii GPUs, as well as tweaked Kabini/Temash in preparation for a launch next year of new lower power APUs.

The second plausible explanation is simple: Kaveri production begins in December, making it much easier to meet wafer concerns. Kaveri is initially aimed at the desktop, and this is where AMD has been stating the company has had good growth.

Regarding Kaveri production, I specifically mention the date of availability as January 14th for a reason. Once the old dude above is finished looking at the wafer, there are a few additional steps needed to actually make it usable. The wafer has to get diced, and the chips need to be tested/binned and packaged. This takes some time, and therefore in order to make chips available by January 14th, production would have to begin in December.


We are finally starting to see indications of how strong the demand is for next generation consoles, and based on various forecasts and Sony's launch strength, demand seems to be quite good. Using iSuppli data, AMD is likely making around $100 in revenues per console at an operating margin of 15%. And according to an article on Bloomberg, GameStop alone has 2M+ people on a "first to know" list waiting on consoles, indicating a healthy demand.

Comparing my estimates above to some analyst's expectations from earlier this year, Barron's notes that Mr. Vivek Arya stated the following:

Our published AMD model assumes 5.7mn console APUs at $81 a piece in 2013 with units growing to 15.4mn in 2014 but ASPs declining to $75. We also assume a 15% operating margin contribution from console processor sales given our expectations of ~30% gross margins for such chips, ~15% spending on R&D to make upgrades process node shrinks, graphics, processor architectures upgraded and virtually zero spending on SG&A. This results in a baseline EPS contribution from consoles of $0.09 cents in 2013, and $0.22 in 2014. AMD's core businesses are still expected to lose $(0.26) in 2013 and $(0.09) in 2014.

So an EPS contribution of between $.20 to $.30 is inline with projections made in July. In the same article, Mr. Arya also states revenues for full year 2013 are estimated at $4.98B.

AMD's combined revenues for the first three quarters is $3.71B, and on the low end of guidance (2% sequential revenue growth), AMD's full year revenue would total $5.2B, well above the $4.98B estimate.

But unfortunately, I think share price reaction is a telltale of investor sentiment. Even when AMD has good news, there is typically some sort of negative press regarding the information later. Take for example the Verizon (NYSE:VZ) server deal. There was a discussion as to whether Intel (NASDAQ:INTC) or AMD chips were used in deployments, and it is taking comments from AMD's VP of Servers on VR-Zone to "set the record straight":

Feldman explains that before SeaMicro was acquired by AMD, and before Verizon's cloud deal had been announced, it had sold Verizon Xeons. But things changed after the acquisition and once Opteron-based systems came online.

"After the acquisition, and once we built an Opteron-based system, all forward-going orders have been on Opterons," Feldman said.

For the Verizon cloud deal, Feldman said that Intel simply couldn't meet Verizon's requirements.

And my guess is that Mr. Feldman is chomping at the bit to get his hands on the server version of Kaveri.

The slide above, taken from the website BitsnChips, states that the server version of Kaveri is "ideal for fabric connected systems." I highly recommend reading this article by fellow SA contributor Mr. Charles Cooper, which explains the importance of AMD's fabric technology in a manner that is easily understandable, as well as being a joy to read.

If AMD is able to get Kaveri production running in December, it becomes easier to meet the upcoming wafer requirements. Based on comments during the Q3 conference call, I am awaiting a press release from AMD detailing 2014 pricing. Note the last bullet on the slide, if translated, reads something to the effect of "Available 1H 2014", I believe.

In summary, consoles are a great source of steady revenue for AMD, and there will be no competition from Intel or Nvidia in these units until Sony and Microsoft roll out a future generation. Consoles actually allow AMD to enjoy having a moat in their GVS segment. But for the market to truly appreciate the added value, I feel AMD's numbers will have to speak for themselves, and speak loudly. Commentary as to why the stock price dropped post Q3 earnings call included PC industry worries and sentiment that the beat wasn't big enough.

As long as there is something that can be construed as negative during the conference call, bears will continue to have reasons to remain bearish. During Q4, this could come from Computing Solutions revenue, as consumers hold off on purchasing newer AMD products in anticipation of the upcoming Kaveri unit, which could also give Intel the chance to capture some additional market share in the PC space.

If there is a negative reaction to any news regarding AMD's commitments to GlobalFoundries, it will likely have the biggest effect this quarter. But regardless of the outcome, barring any detrimental news in the near future, 2013 has been a much better year for the company than 2012.

AMD's focus on easily synthesizable, scalable architectures could allow the company to more easily meet commitments going forward, as well as make the company more competitive in traditional businesses. While I think we may not see GPUs roll out during Q4 at GloFo based on the time it would take to get production up and running, moving limited GPU production to GlobalFoundries, along with the Kaveri launch, should serve to alleviate some of Dr. Rasgon's concerns regarding the WSA 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.

Additional disclosure: I am long both shares and options in AMD and actively trade my position. I may add/liquidate shares/options at anytime. I may add to my INTC position in the next 72 hours.