Advanced Micro Devices (AMD) is in serious trouble. In the most recent quarter, the firm swung to an operating loss in its core CPU/APU division to the tune of $114M. This was due to a combination of inventory write-downs on stale inventory, a drop in market share from 18.8% to 16.1%, and lower average selling prices across the board. Intel (INTC) has proven that it is the clear leader here, and given the broad industry shift to more efficient designs in everything from notebooks to big-iron servers, it would take a significant misstep on the chip giant's part for AMD to even come close to building a comparable solution, let alone a better one.
However, AMD -- thanks to its $5.4B acquisition of ATI Technologies -- has traditionally had a strong hand in the discrete graphics business. Discrete graphics chips are used in gaming computers, professional workstations, and even in supercomputers (albeit slightly modified to be floating point co-processors rather than graphics cards). In this space, there's only one competitor: Nvidia (NVDA). Generally speaking, Nvidia has traditionally been the market share leader but the gap between them was generally not too wide (usually a 40-45% share for AMD and 55-60% for Nvidia).
However, the recent market share reports confirm that the balance of power is worsening for AMD.
Q3 2012 - Things Are Looking Lopsided
In Q3 2012, according to the latest data, AMD's share in desktop computers dropped to 35.7% from 40.7%, and in notebooks, it saw share plummet from 44.8% to 34.2%. Of course, as there are only two players in the industry -- it shouldn't be too hard to figure out where those missing percentage points went.
In the quarter, net revenue for graphics was $342M, with net income a startlingly low $18M. In contrast, Nvidia's gaming graphics division saw sales of $739M and an operating income of $193M. Nvidia's professional solutions division -- which is broken out separately for Nvidia but lumped in with graphics for AMD -- turned in a solid $220M in sales and operating income of $100M.
The balance of market share and profits in the graphics business is starting to look eerily like that between Intel and AMD in the CPU space. This has very serious implications.
So, What Does This Mean?
In a nutshell, this means that Nvidia simply has a lot more money to reinvest in its core business. That means stronger R&D, which leads to dramatically better products, better marketing of said products, and a much healthier competitive landscape from the perspective of Nvidia.
An example of this was actually seen with the latest generation of graphics cards. AMD put out its biggest, baddest Radeon 7970 with a 365mm^2 die size and for the first quarter of 2012, managed to take the performance and performance/watt lead with its products. However, Nvidia was able to successfully compete on performance and performance per watt with a chip that only weighed in at 294mm^2. With a 19% smaller die size (this means cheaper to produce), Nvidia was able to sell cards at the same price point as its rival. This meant better margins and higher operating profits than its peer.
Explaining AMD's Market Share Decline
Many will say, "well, AMD's products perform similarly to the Nvidia products in the same price point." Surprisingly enough, they would be correct. On paper and in most gaming benchmarks, AMD's products are extremely competitive across the board. The gap in terms of performance that we see on the PC processor side is simply not there. So what's going on?
Well, it really comes down to a number of things on the consumer side:
On the brand equity side of things, Nvidia simply takes the cake. Why? Nvidia, first off, has a number of programs and promotions that help customers feel like they are part of a legitimate ecosystem. For example:
- GeForce.com - This is a site that Nvidia keeps up that has blogs, game demos, driver updates, promotions for the latest games, contests to win prizes, and community support forums. AMD's Radeon support site offers nowhere near the level of user experience and support. It's also not called "Radeon.com" or anything like that.
- The Way It's Meant To Be Played - Nvidia's marketing campaign -- The Way It's Meant To Be Played -- which essentially acts as a "seal of approval" from the game publisher, acts as a powerful "reminder" to gamers that in order to get the full experience of the game, one needs Nvidia hardware.
On the latter point, Nvidia's strong developer relations and software teams do actually enable a few fundamental features critical to the user experience that AMD does not offer. For example:
- PhysX - This enables GPU-accelerated physics effects in games that are programmed to support them. Nvidia's software team works with game developers to smoothly and seamlessly get these things integrated. Gamers who want the best generally don't want to have to keep checkbox features unchecked, which is why the scales often tip in favor of Nvidia.
- 3D Vision - This allows for stereoscopic 3D in PC games, provided that one owns a 3D Vision Kit (a pair of 3D glasses) and the right monitor. A niche market, but something gamers will consider when plunking down hundreds of dollars for their next generation graphics cards. Nvidia works with the game developers to make sure this is implemented in many of the major titles.
Okay, What About In HPC/Professional?
In the professional space, Nvidia is eating AMD's lunch because it simply builds the better mousetrap with much better software/driver support. In a recent review of AMD's latest FirePro professional workstation cards, the verdict -- especially compared to Nvidia's products -- was not favorable:
To be perfectly candid, we found the W8000 and W9000's performance disappointing, and their proposed price/performance ratio isn't so hot either. If AMD had kept to its previous price structure, the W9000 would've slipped in nicely at $2500 and the W8000 at $1149. At $4000 and $1600 against the Quadro 6000 (also $4000) and the Quadro 5000 ($1849) though, AMD's prospects at this time aren't very good. The situation may change with future driver revisions if AMD is able to wring more performance from the GCN architecture in professional applications, but the current state of the software is what it is and the W8000 and W9000's performance didn't scale as expected.
Professionals using these cards to make serious money will not mess around with second best. And, unfortunately for AMD, that's precisely the position that it is in.
But The Next Generation Of Products Can Fix This, Right?
This brings us to the next question: can the next generation of products tip the scales? Well, it's absolutely not too late to try to change course and really buckle down and focus on the graphics space. All of the puzzle pieces are there. The problem, though, is that AMD is spreading itself too thin to be truly effective at anything that it attempts. It doesn't have enough time or money to try to sow a million seeds and hope something starts sprouting.
AMD is in the process of axing its staff as we speak in an attempt to cut costs. While it lays off engineers (and I'm sure the very best and brightest that remain are giving Intel, Qualcomm (QCOM), Nvidia, Samsung (SSNLF.PK), and others a call looking for new jobs), it is simultaneously trying to develop two distinct lines of X86 processor cores, an ARM (ARMH) based system-on-chip for servers, X86 SoCs based on the two aforementioned lines of cores for everything from entry level PCs to servers, and next generation graphics. All while running either money-losing or barely profitable businesses!
Simply put, AMD is trying to do too much with too little, and it needs to buckle down and focus on something. Unfortunately, discrete graphics and HPC accelerators don't seem to be what management -- or Wall Street -- wants. Even if it's the one place where AMD could truly bring the fight to Nvidia.
Nvidia is quickly about to become the Intel of discrete graphics in every sense of the word, and I suspect that the firm will continue to bleed AMD's market share dry until that $18M operating profit swings to an operating loss. And when the discrete graphics part of AMD stops innovating because it has simply run out of money, then this will not only have serious repercussions for its competitive position in the GPU space, but in the integrated graphics space, where Intel is doubling graphics performance with each new processor. There will be no low end for AMD to retreat to since it faces competition from Intel and its own APU products down there.
Things do not look good for AMD in the discrete graphics space. It lost the X86 CPU space by falling too far behind, and it is now about to lose the graphics war for good if it does not continue to innovate where it counts by developing products that carry high margins and sell well. Shifting focus to ARM-servers with SeaMicro fabric all sound great... until you realize that this will likely be a super cutthroat, low-margin business that will be swallowed whole with Intel's upcoming "Avoton" server SoC. Built with Intels' very own, license-fee-free, 22nm cores.
At least Nvidia doesn't have a manufacturing process lead.