AMD Announces Semi-Custom Wins, Projects Sales Decline, Cuts 7 Percent of Staff

It was a mixed set of announcements from AMD last night, with some positive news for the chip manufacturer, but some long-term projections that sent investors skittering. This was Lisa Su's first quarterly announcement after taking over the CEO position barely a week ago, and she opened with good news -- not only did AMD turn a small profit on the quarter -- $17M net income ($63M operating income) on revenue of $1.43B -- it's secured two new semi-custom deals expected to provide roughly $1B in revenue over the next three years. That works out to around $84M per quarter -- not an enormous amount, no, but a nice feather in AMD's overall business cap. AMD expects to ship this hardware for revenue in 2016 and revealed that at least one design is for an AMD ARM core, rather than an x86 chip.

Su also announced that Arista has begun shipping new switches based on AMD's G-series embedded hardware, the already-discussed wins to power the new Retina iMacs, and strong sales of both the Xbox One and PS4. Both consoles are expected to have peaked in Q3, Q4 orders will reflect the drawdown in orders after the post-holiday rush. Mobile GPU sales rose, as did commercial APU client sales. Su also noted that AMD is seeing growth in its professional GPU sales and expects that to continue and that the company recognized $28M in technology licensing revenue.

5K Retina Apple iMac With AMD Radeon M290X Inside

5K Retina iMac With Radeon M290X Inside 

Despite Good News, PC Market Share Continues to Decline

Unfortunately for AMD, all the good news in other segments, while critical to the company's long-term success, is only barely compensating for the decline of its chipset and CPU/APU business. Sales in that segment are down 6% sequentially and 16% year on year. To put that in further perspective -- in Q3 2014, AMD sold $781M worth of CPUs and APUs. In Q3 2011, three years ago, AMD sold $1.28B worth of CPUs and APUs -- a three-year decline of nearly 40%.

While this is wretched performance, it's only fair to note that this is the scenario Rory Read inherited when he took over AMD. There's an old saying often attributed to Digital's ex-CEO Robert Palmer: "Building semiconductor fabs is like playing Russian roulette. You put the gun to your head, pull the trigger, and find out four years later if you blew your brains out." This same rule applies to CPU architectures -- and once it was clear that Bulldozer was as fundamentally compromised as it was, AMD was stuck with a long-term problem and no short-term solutions.

Thus, as bad as AMD's performance in what was once its core market has been, one could argue that this decline was inevitable until the company could bring a new architecture to market -- and therefore, AMD should be judged more on the strength of its pivot and its bets in emerging spaces. I think that's a fair argument given the size of the problems AMD was facing down -- but it's not one that's going to be particularly comforting to investors given that AMD has slashed its R&D budgets even as Intel and Nvidia have dumped more cash into them.

The problem was compounded when Su announced that AMD would be reducing its headcount by approximately 7% with most of this completed by the end of the year. The goal is a savings of approximately $9M in 2014 and $85M in 2015. Lisa Su claims that the company is protecting "the key engineering skills and key engineering investments," but says in the next breath that "we found opportunities to streamline, particularly on the Computing and Graphics side." That's not encouraging if you're already concerned about AMD's core business.

AMD expects a 13% sales decline in Q4, weighted towards the semicustom end of the business, but it declined to break out the exact split. Given that Q4 has typically been an industry high point, this is bleak news. The company is struggling to ramp up its new businesses while managing the decline of its own and dealing with the fact that while it will have new parts in 2015 (including a new low-power x86-ARM combined platform), the really game-changing initiatives are still a year away.

The one big announcement that might've kicked things into higher gear would've been a 20nm GCN 2.0 GPU, but Su was quiet on that front and no analysts asked.

A Company In Transition:

The big takeaway from this is that it's going to be another 12-18 months before we know if AMD's big long-term bets will pay off. If AMD has a new GCN 2.0 on 20nm coming soon, that changes the GPU competitive equation. If the 20nm die shrink of Jaguar as part of Project Skybridge comes along with a nice clock bump, we could see a new wave of well-positioned low-end notebooks. If AMD's ARM server plays start to grab hold, we'll see increased revenue from those quarters.

Ultimately, how this plays out depends on how well AMD is playing its hand behind the scenes.