Advanced Micro Devices (NYSE:AMD) seems to be back on track, reporting a second quarter profit when a loss was expected, and promising new product releases into 2017.
But most investors think of AMD as the perennial loser to rivals Intel (NASDAQ:INTC) on the CPU side and Nvidia (NASDAQ:NVDA) on the graphics side. Because AMD has been. Every time in the past that AMD seemed to be getting a decent share of the market, one way or another its rivals knocked it back down again. Shorting AMD has been a profitable hobby for some tech investors during this last decade.
I have owned AMD stock since 2004, so I have followed AMD's up and down cycles. However, after Lisa Su was promoted to CEO in 2014, I began accumulating stock again, carefully. She impressed me as the right person to run the company. She took the product road map she inherited, and the AMD team, and came up with a creditable transition plan.
After the acquisition of ATI, back in 2006, the plan had been to beat Intel and Nvidia by combining serious graphics and CPU power on a single chip, dubbed an APU (Advanced Processing Unit). But it took longer to merge the two than expected, and the resulting chips were panned as having mediocre CPUs and GPUs. They were consigned to the middle range and low end of the market, where margins were poor. Also Intel was ahead in process technology, and so could put more gates on a chip, and integrated good enough graphics on their chips.
Lisa Su and crew had to prioritize projects while getting the most out of a relatively small R&D budget. By winning the gaming console market, keeping APUs alive, and keeping competitive in stand-alone GPUs, the AMD team positioned itself for a comeback in products with better margins, like server chips. In 2016 and 2017, we are seeing the fruits of those efforts.
Attacking the middle first: a brilliant GPU play
Sapphire Radeon RX 480
The recent introduction of the Radeon RX 480 graphic chip and cards appears to have gone well. It probably boosted Q2 revenue, although Q3 will be the first full quarter. Lisa Su said there was a good supply on launch day and that the demand had continued to be strong. She added some dealers have run out of supply, but the 14 nm yields are good so AMD hopes to catch up with demand and also launch the other products in the line. Before the conference, I had confirmed that many online retailers were sold out and were taking advance orders only.
The RX 480 products are mid-tier. Their typical buyers would be serious gamers and graphics artists with limited budgets. But that is where the unit sales bulge in the market is. Sure, everyone wants a $1000 graphics card, but given that it will be just run-of-the-mill in 2 years and nearly obsolete in 3, most people budget something under $300 for video cards.
In the past, in every cycle I can remember, AMD (and its predecessor ATI) and Nvidia would launch a new generation of GPUs by bringing out the top of the line first. This makes sense. The top of the line will beat (in ability to run games or whatever) the prior generation's top of the line. It will have high margins despite low yields of good chips from the dies. When a new chip starts coming out, especially after a die shrink, a substantial percentage of chips have defects. That can make them useless, or sometimes they will qualify for the middle or bottom of the line SKUs. It works psychologically too: the opinion leaders get the fastest, most expensive chips, creating envy all the way down the price chain.
That is what Nvidia did this time around with their latest generation, as represented by the GTX1080, launched May 27 and built on a 16 nm TSMC process. The 1080 costs $599 and draws 180 watts.
The RX 480 draws 150 watts and costs $199, or $239 with extra memory. Is the 480 as capable as the 1080? No. But relatively few people can justify paying an extra $300 for the increased performance. We won't know how AMD's direct competition with the 1080 will compare until it is introduced, but it is likely to be in the same ballpark, and also sell a lot less units than the 480.
By showing the power of the new architecture (Polaris GCN 4) and process node, AMD is giving mid-level users a reason to upgrade graphics cards, or to look for AMD if they decide to get a higher performance card.
Did I mention the RX 480 can run the new VR (virtual reality) goggles and games? So can the Nvidia cards, of course, but AMD will let you do it for a more reasonable price.
On the CPU side, Intel still has a process technology lead over AMD, but it is about to become narrower than at any time in recent history. AMD CPUs at 14 nm should be available in early 2017. Intel will start moving its high-end CPU products to 10 nm, but that is probably a late 2017 story. So AMD has moved from being about 2 process nodes behind to lagging the cutting edge by about half a node. This should allow AMD to be more competitive for mainstream CPUs, and with better profit margins. Margins have been dismal at AMD for most of this decade.
Gaming console concentration not looking so stupid
Microsoft Xbox One S
Nvidia was pretty dismissive of AMD's sweep of major gaming consoles a couple of years back.
Nvidia proclaimed that gaming consoles were dying out. Two trends were to minimize their sales: the move to gaming on smartphones and tablets, and the superiority of the PC experience for high-end gaming. In addition, the CPU/GPU chips for the consoles have relatively low margins, which are not attractive to either Intel or Nvidia.
And yet gaming consoles are still selling, and new models are being brought out including the Xbox One S in August. Game designers who want to fill big screens with action want their games to run well on the consoles as well as PCs, so increasingly game optimization tends to favor AMD. In the past, it tended to strongly favor Nvidia, and of course, the battle is far from over.
Game consoles have been the solid base of support for AMD, allowing it to reinvent itself. Sales are seasonal, peaking in Q4, so chip shipments peak in Q3. That is more seasonal than the PC and graphic cards businesses, so AMD tends to have dismal Q1s.
If AMD can gain market share (it is a distant second now) in GPUs, and do it with better profit margins, the pressure will be on Nvidia. Nvidia has a lot of cash and (so far) superior branding. It lacks high-end CPU technology, other than ARM (NASDAQ:ARMH) (which both AMD and Nvidia license), so Nvidia will likely continue to see erosion from the bottom, where integrated graphics are already quite good, at the same time that it hits stiffer competition for mid and high-end gamer dollars. However, Nvidia is very strong with datacenter graphics computing and in the smart automobile industry. It will not cede the gamer space without a major fight.
Servers for 1.3 billion people in China
The single greatest overall problem for AMD during the last decade has been low margins. Those margins reflect the lower-end categories of products that AMD chips have gone into. Generally, server chips have the highest margins (and high-end servers have higher margins than mid and low-end servers). Within notebooks and desktop PCs, the high-end chips have the highest margins. Same in GPU chips. When operating margins go below zero, no amount of volume sold into the low-end market will make a company profitable. Without profits R&D budgets suffer, continuing the downward spiral.
Several times, AMD has broken out of this spiral, only eventually to be ground down by Intel and Nvidia. That could be what will happen this time. But I think AMD may have finally turned the tables, at least in some specific areas.
The population of China is about 1.3 billion, larger than the combined populations of the United States and Europe. China has the second highest GDP in the world, and the highest if adjusted for purchasing power. Despite all the talk about free trade, China cannot import many top-end electronics products, including Intel's CPUs, due to U. S. Defense Department concerns. See Intel China Ban.
But China is ready to produce its own server chips, and will work with AMD to do that.
The announcement of the agreement is short enough that I'll begin by quoting it in its entirety (it is within the Q1 2016 results press release):
AMD licensed high-performance processor and SoC technology to a newly-created JV it has formed with THATIC (Tianjin Haiguang Advanced Technology Investment Co., Ltd.) to develop SoCs tailored to the Chinese server market that will complement AMD's own offerings. The $293 million licensing agreement is a meaningful step in AMD's IP monetization strategy intended to accelerate the Company's growth and better monetize its valuable assets. Payments are contingent upon the JV achieving certain milestones. AMD also expects to receive royalty payments from the JV's future product sales.
"Our new licensing agreement is a great example of leveraging our strong IP portfolio to accelerate the adoption of our technologies more broadly," said Dr. Su. "The joint venture with THATIC provides AMD with a differentiated approach to help gain share in the fastest growing region of the server market."
Once the server chips are available, probably in 2018, I would not be surprised if the Chinese government "encourages" businesses to use the Chinese chips. Since AMD is providing intellectual property, and does not have to make the chips, the royalty payments should greatly add to AMDs margins.
Right now, AMD has only a tiny share of the server market. Between the China chip and the upcoming Zen server chip, it can only gain market share. How much market share it can gain will be a key question in 2017-2018.
AMD is a large, global company with a variety of products. I am not attempting to cover every aspect in this article, just the ones that I think will have a major impact over the next 3 years or so. Well, I will add just one: the new professional graphics card sampling now that has a full terabyte of memory. "8K raw video timeline scrubbing was accelerated from 17 frames per second to a stunning 90+ frames per second. As content creators rapidly adopt 8K resolution to future-proof their content, a 5X performance boost will improve quality of life, productivity and efficiency in the editing process." That will be useful in high-definition video editing and all sorts of high-intensity computing.
AMD's Q2 upside revenue and earnings surprise is well covered, for instance by AMD's Turnaround Continues, so I won't go over that.
AMD is not out of the woods, so I would still classify it as a risky investment. The upside, however, is large. Neither desktop PCs nor notebooks died as predicted. Many people need their power both for entertainment and for work. While some screens have shrunk, some have grown in size and resolution, requiring better graphics processing.
Gaming has become a big, big industry, and the end of growth is not in sight. VR games will add to it, and they require much more graphics processing power than even large screen games. AMD has set itself up to grow with VR at price points masses of users can afford.
AMD's debt, which seemed to loom so large just a couple of years ago, now is beginning to look like a well-spent startup investment. At the end of Q2, AMD had about $0.96 billion in cash and $2.24 billion in debt. That is very manageable provided AMD can stay cash flow positive on an annual basis.
AMD's stock shot up after Q1 results were released and again after Q2 results were released last week. It closed on Tuesday, July 26 at $$6.98, up 166% since the close at $2.62 the day before the Q1 results release. The last time AMD was over $6 per share was in 2012.
AMD's stock still looks undervalued to me, accounting both for the forward-looking positives and for the knowledge that Intel and Nvidia are still in command of the market. I expect it will bounce around between $5 and $10 per share until it is either clear that (1.) AMD is really taking significant market share, and at a healthy operating margin, or that (2.) it has been foiled, again. If it's the former, then we will get back to multiplying EPS by whatever ratio we think is appropriate to value the stock.
I'm thinking of AMD as a successful looking VR startup. Its GPUs can run VR games. Its next generation professional GPUs and server chips can generate VR content. And likely in a couple of years, its APUs will be able to run VR as well.
Disclosure: I am/we are long AMD.
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.