What’s Ahead For AMD in 2019
2019 is going to be a major year for both of the top players in the PC market — but it’s going to impact them in different ways, and the two companies are playing for remarkably different stakes. The Intel-versus-AMD story remains a major component of the tech community, but it’s evolved in some surprising directions compared to what one might have expected five years ago, and the two companies are playing for somewhat different stakes these days. We’re going to talk about what’s coming next for both AMD and Intel, starting with AMD.
AMD: Betting it All on Lucky 7
For AMD, 2019 is all about 7nm and the manufacturing and technological innovation tied to its next-generation process node. These transitions have historically been pain points for the company, and its long-term record over the past 17 years is scattered with a number of difficult shifts, either between microarchitectures, nodes, or both.
This is not to say that AMD has never had a successful transition, but these inflection points have tended to reveal strategic weaknesses in the company’s ability to maintain performance or product leadership. AMD’s most central challenge for the year is to continue building on the success of Ryzen and continue taking market share in desktop, server, and mobile. We don’t have a timeline for 7nm Ryzen or Epyc launches yet, but Rome (the 7nm Epyc die shrink) is expected in volume in the back half of the year, with Q3 a likely target. If AMD launches Ryzen first, that suggests a 7nm debut in Q2. If it leads with Epyc, we might also look to Q3 for Ryzen.
AMD will launch Navi at some point in 2019, though that GPU is rumored as a midrange part rather than a high-end challenger, and there are rumors of 7nm Vega chips for the consumer market as well. The focus, for AMD, will be on demonstrating a solid set of overall architectural improvements across both CPUs and GPUs to win new market space in server, desktop, and mobile. The focus for Ryzen thus far has been on server and desktop markets and presumably this will continue.
At the same time, we expect Lisa Su to focus on continuing to build AMD’s margins. This has been a major focus on AMD conference calls and in Q&A sessions with investors. AMD’s stockholders are well aware that, the stock’s recent performance notwithstanding, AMD has historically suffered from erratic performance as the company’s fortunes have risen and fallen. Strong margins — AMD’s are currently up to ~40-41%, compared to ~34% in the recent past — are an important mark of overall health.
The margin question is complex, with a variety of pushes and pulls. Su’s tenure at AMD has been marked by a few distinct trends. Under her leadership, AMD has pursued multiple long-term licensing and product manufacturing deals, trading high margins for long design cycles in some cases (PS4, Xbox One), and signing IP licensing deals in other cases. The IP licensing deals are great for margin and can help smooth quarterly results, while the long console product cycles were critical to helping AMD survive into the Ryzen era in the first place.
We don’t expect to hear much about the Xbox Next and PS5 until 2020 rolls around, but AMD could still announce new semicustom wins in 2019 and we’ll hear at least general discussion of the expected shape of the console business as cloud gaming services continue to ramp.
Console and semicustom margins tend to be low, in other words, though the margins for PS5 and Xbox Next will be highest at launch, assuming AMD uses a similar agreement for these systems that it used for Xbox One / PS4. The impact of IP licensing is generally quite good for margins. Epyc sales and launching Ryzen into higher price points is good for margin. How much additional profit AMD makes on 7nm is a complicated question that’ll depend on how good its 7nm yields are, how much its paying for wafers at TSMC, any associated penalties or costs that might be part of its 14nm WSA with GF, any additional costs associated with combining the TSMC / GF silicon on the same package, and, of course, just how good 7nm Ryzen and Epyc actually are compared to Intel.
AMD isn’t just depending on its 7nm hardware to drive profits, however, and this point is often missed in discussions of how companies price products at launch. According to Lisa Su during AMD’s Q3 2018 conference call, “2018 remains an inflection point for AMD, as we expect to exit the year with well over 50% of our revenue coming from new products, driving significant margin expansion.”
“New products,” in this context, refers to the Ryzen, Epyc, Vega, and Polaris families in their various refreshes and designs. It may or may not include the Xbox One X and PS4 Pro (as compared to the 2013 original designs). What’s striking about this paragraph is that it shows how long it takes for older hardware to transition out of the market. Based on how thoroughly Ryzen annihilated the old Piledriver cores, you might think Bulldozer and Carrizo had been completely driven from the market. This won’t actually happen until after Ryzen’s second birthday.
AMD’s Largest Challenges
AMD’s biggest challenges will be to demonstrate the strength and maturity of its 7nm deployments. CPU-side, that means Ryzen and Epyc and continued goals for increased market share in desktop, consumer, and mobile. GPU-side, consumer Vega didn’t demonstrate the kind of performance leadership that GPU fans were hoping for, leaving all eyes on Navi. AMD’s ability to build solid silicon for consoles is not in doubt, but the company hasn’t robustly competed in the high-end GPU market for years.
If AMD is serious about taking the fight to Nvidia and Intel in AI, ML, and GPU virtualization markets, it’s going to have to do a better job at both fielding and communicating about its competitive hardware solutions. When I’ve asked acquaintances and contacts who work in these fields whether they’ve considered AMD hardware, the response is always some variant of “I think Nvidia is required.”
Navi doesn’t need to revolutionize the entire space, top-to-bottom, but wherever it lands, it needs to land well, and AMD needs to demonstrate a sustained commitment to building the infrastructure and software development teams to support whatever play it intends to make in the GPU market outside of consoles, APUs, and their obvious adjacencies. 7nm Vega got the ball rolling in 2018, but AMD needs to demonstrate some momentum around its own parts.
2017 was the year AMD proved it could still build a high-end microprocessor. 2018 demonstrated it could build demand for those solutions and keep its 12nm refresh cycle on-track. In 2019, AMD will be focused on pulling off a difficult node-and-architecture transition, building new console SoCs, and working to break into markets like deep learning, machine learning, and artificial intelligence.