AMD Pulls Out of its Nose Dive

by Mark W. Hibben

The long steep decline in revenue of AMD has finally come to an end, thanks to the start of processor deliveries for the new Microsoft and Sony game consoles. Ramp up of production of the custom game console processors can be expected to rerturn AMD to profitability in Q3. The question for investors now becomes, is this profitability sustainable?

Pulling Out of the Dive

As the chart below shows, AMD has atleast for now, pulled out of the revenue nose dive that its Computing Solutions division has suffered since the beginning of 2012.

The question of sustainability was evidently on the minds of AMD investors as they mulled over AMD's Q2 earnings report during after hours trading. AMD shares had closed up nearly 6% only to give most of it back after the markets closed.

If there was a single specific catalyst for the change in sentiment it must have been a question by an analyst asking whether the new revenue from console APUs (Accelerated Processing Units in AMD-speak) would only just offset the loss of royalties from the previous console generation. It was as if all the air had been suddenly sucked out of the room, leaving the AMD execs gasping for breath. Needless to say, they wouldn't go into details.

In fact the revenue streams do not work out to be comparable, even if you assume the low end of the unit shipments for console APU's of 5 million per quarter, which agrees with the historical average for combined console shipments by Microsoft and Sony since the introductions of their current generation consoles. The analyst had estimated that AMD made about $100-150 million in royalties per year. At 5 million console APU shipments per quarter, a $40 ASP, and assuming 35% gross margin, the gross profit per quarter works out to be about $70 million, for an annual gross profit almost double the assumed royalties.

But the question took away the celebratory air of the conference call, an left unanswered whether AMD could sustain profitability after Q3.

A Closer look at AMD's Q3 guidance

On the surface, AMD's guidance for Q3 was refreshingly positive: a sequential increase of 22% in total revenue to $1.4 billion, a reduction of operating expenses to $450 million, and a gross margin of 36%. Turning the crank on the numbers yields an operating profit of $59 million. Not great, since it won't even represent an improvement over AMD's 2012 Q3 results ($1.269 billion in revenue, operating income of $131 million), but much better than the recent string of losses.

Most of the $260 million increase in revenue for Q3 will necessarily come from console APU shipments. Revenue from the graphics division will show probably show negligible growth, hovering around the $320 million mark. There are signs that AMD's share of the PC and server processor market has stabilized at about 15%, but I don't expect it to grow substantially next quarter either.

This leaves console APUs with the burden of carrying the full $260 million increase. By virtue of the revenue generated by the Computing Solutions division in Q2, it's clear that AMD has already shipped a substantial number, about 5.5 million console APUs. They would need to in order to support planned build ups by Sony and Microsoft for the holiday season, which have to be started months in advance of shipments to retailers. In order to reach the target revenue of $1.1 billion for Computing Solutions, AMD will need to ship about 12 million console APUs in Q3.

That's a lot of APUs, but combined sales of 20 million consoles in the first quarter of availability is probably reasonable, and I'm sure that Microsoft and Sony would rather have too many units available than not enough. What happens after that is anyone's guess, which is why AMD execs wouldn't speculate on Q4 revenue or console shipments. They simply don't know.

Q4 and Beyond

By Q4, the console makers have all they APUs they need for the near term, and APU shipments probably fall back to the 5 million/quarter range. Unless something dramatic happens in the PC side of the business, that isn't enough to sustain profitability, since AMD's revenue basically reverts to the roughly $1.2 billion of Q2.

When I first wrote about AMD back in April, I speculated that console APU's might be able to save AMD, but only with further cost cutting. Going forward, AMD will need to make further operating cost reductions to the tune of $30 million in order to break even in Q4 and beyond. Naturally, this is something that received special emphasis in AMD's presentation, and investors should watch AMD's cost cutting performance in Q3 to ensure that AMD is on track for further cuts in Q4.

A little expansion in the PC and Windows Tablet areas would be nice, but here the future looks very cloudy. The very mention of Bay Trail seemed to produce yet more managerial asphyxiation among the AMD execs. It's tough competing with Intel's latest and greatest mobile chip when your foundries (either Global Foundries or TSMC) are a generation behind in process technology. AMD is readying some processors for the Windows 8 ultralight and tablet markets, but going up against Bay Trail is going to be tough.

AMD's main advantage is what got them the console APU wins, very capable internal graphics processors married to very capable CPUs. In neither category would the internal processors be the best on their own, it just isn't possible (for AMD) to build best in class graphics and general purpose processing into a single piece of silicon. But in side by side tests, AMD's best (the A10-6800K) trounces Intel's best (the 4th gen Core i7 4770K) in graphics intensive gaming. AMD was simply willing to build a more capable graphics processor into their machines than Intel was. But this could change.

In the mean time, AMD APUs for mobile may find an audience with mobile gamers, and eventually, a variant of AMD's console APUs will become available for PCs, assuming that Microsoft and Sony haven't locked AMD into too restrictive a licensing agreement. There's every indication that the Xbox One and Playstation 3 game consoles have excited gamers worldwide with their combination of price and performance. These could provide a halo effect for AMD APU based PCs, but only if consumers can buy performance equivalent or better than the new consoles.



Is the Cost of Intel's Entry into Mobile too High?

by Mark W. Hibben

Continuing PC Declines

Intel reported revenue of $12.8 billion and operating income of $2.7 billion for 2013 Q2. Revenue grew 2% from Q1 but was down 5% year over year, but more surprising was the nearly 30% decline in operating income from the year ago quarter. Intel's entry into the mobile business is proving costlier than expected, and Intel management guided slightly lower for the second half.

Investors were prepared to see some year over year decline in revenue for the PC Client Group based on the Gartner and IDC reports recently released on worldwide PC sales for Q2. Both showed year over year declines of about 11%. Intel reported a PCCG revenue decline of 7.5% year over year, but left for their SEC filing the number for the group's operating income. Operating income has also been falling.

The research numbers don't include tablet devices, and so probably reflect a continuing consumer shift away from traditional PC form factors. Since Intel's initial Haswell offerings all required fans of some type, whether desktop or mobile, Haswell has been restricted for the time being to the traditional PC market, where for whatever reason, it couldn't make much of an impact.

Mobile Un-profitability

But the continuing PC decline is only one part, and perhaps not the most important part of the explanation for Intel's earnings decline. As Intel ramps up production of its next generation Bay Trail Atom processors, as well as "fanless" Haswell mobile processors, and prepares to enter production for its most advanced 14 nm process, operating expenses have increased to $2.7 billion. Operating margin has declined from 28% in 2012 Q2 to 21% for the current quarter.

While costs associated with the introduction of new mobile oriented chips have increased, revenue for the Other Intel Architecture (OIA) group has steadily fallen. OIA includes all processors (Atoms) for tablets and smart phones, as well as mobile device radio chips. As the chart below shows, OIA has been a money loser in every quarter since 2010 Q1, except for one quarter, 2010 Q3.

Although Intel didn't release operating income (or loss) numbers for its segments, when the SEC 10Q comes out, it will probably continue to show a loss of about $600 million for OIA.

So far Intel investors and analysts like myself have tended to focus on the technical advantages of the latest Haswell and Bay Trail Atom processors, which both use Intel's most advanced 22 nm process, a process that continues to be unavailable in the ARM world. When these start appearing in consumer devices in the 3rd and 4th quarters, they will offer better performance and be more energy efficient than anything currently available in the ARM SOC world. So Intel will just crush ARM, right? Maybe not.

All the discussions about processor performance ignore some realities that have existed since the introduction of the Apple iPad in 2010. From the very beginning, the SOC in the iPad was demonstrably slower and less computationally powerful than competing (and less expensive) Netbooks, and we all know what happened to the Netbook. Consumers weren't evaluating the iPad or Android tablets on the basis of processor benchmarks, but looking in a more wholistic way at the entire device experience, including apps and services.

Processor performance certainly contributes to the overall experience, but indirectly, so processor performance doesn't really have as much leverage on sales appeal as many have assumed.

Intel has already proven that it can build mobile device chips as energy efficient as the latest ARM processors in its Clover Trail Atom series based on Intel's older 32 nm process. Anandtech's testing of a Clover Trail processor (Z2760) showed that it was just as energy efficient while being more computationally powerful than competing ARM processors. Tablets with iPad like thinness such as the HP Elitepad 900, which uses the Z2760, have been available since the beginning of the year.

Yet there hasn't been a mass exodus of tablet makers from ARM to Intel, nor have consumers switched on masse to Intel based tablets. The most notworthy adoption of Clover Trail has been Samsung's announcement that it would be in the next Galaxy Tab 10.1.

The reason for this is simple, and the hint for the reason is present in OIA operating losses. Manufacturers haven't switch to Intel Clover Trail because it didn't offer a significant cost advantage over ARM. Cost is the last major battle to be fought between ARM and Intel, and based on Intel's OIA track record so far, the outcome is far from certain.

Manufacturing costs and ASPs are secrets that chip makers keep closely guarded. The best information I've been able to glean about ARM chip costs comes from iHS-iSupply teardowns of tablets such as the under $200 Google Nexus 7 with a processor cost of $21 for the nVidia quad core Tegra 3 processor. Likewise, the Amazon Kindle Fire processor costs $14.65 for the Texas Instruments OMAP4430.

Since these were purchased parts from external manufacturers, the costs include design and capital equipment costs, and therefore represent a reasonable basis for comparison with Intel. Can Intel beat ARM processor makers on price, while making a profit? The answer is very much a function of sales volume for Intel chips. Outside of Intel, I doubt anyone knows what the sales volume threshold of profitability will be for Bay Trail, let alone whether that threshold will be met.

Q4 and Beyond

Intel's management team, headed by newly appointed CEO Brian Krzanich were very bullish on Bay Trail and fanless dual core Haswell processors, and these will have an impact by elevating the performance of thin form factor tablets. I expect them to carve out a significant market share for Intel in the tablet space in Q4 and into 2014.

How significant depends a lot on how the ARM processor manufacturers and foundries respond to the Intel challenge. Taiwan Semiconductor Manufacturing Company (TSMC) and IBM's Common Platform partners Samsung and Global Foundries are working feverishly on their own 14 nm processes, which should rival Intel's in power efficiency. If the ARM foundries can get the process on line by the middle of 2014, they should be able to hold the line against Bay Trail and its successors. If not, tablet device makers will begin to abandon ARM for Intel in 2014, thereby finally bringing profitability to Intel's mobile business.