Over the past year, AMD has often hinted that it would announce a major new semi-custom win in early 2016. Yesterday evening, it finally did so, and it’s a more significant announcement than investors or analysts were expecting. AMD has formed a new joint venture with THATIC (Tianjin Haiguang Advanced Technology Investment Co., Ltd). to build SoCs for the Chinese server market. AMD expects the total value of the deal to be $293 million, with $52 million in revenue recognized over the course of 2016.
Much of the company’s conference call was devoted to discussing AMD’s expectations for the coming year rather than focusing on the previous quarter. According to AMD, it is licensing technology to its Chinese partner, not market-ready products — but the company noted that its core expertise is in x86 processors with some additional valuable IP in server fabrics (likely acquired via SeaMicro). Asked point-blank whether AMD had licensed x86 IP to the THATIC joint venture, CEO Lisa Su confirmed that it had.
According to AMD’s CFO commentary, the joint venture “consists of two separate legal entities where AMD has a majority stake in one and a minority stake in the other. AMD’s total exposure to JV losses is limited to its investment in the JV. As of March 26, 2016 our investment in the JV was $0.”
This suggests that the JV is structured to bypass restrictions in AMD’s x86 license agreement with Intel that would otherwise prevent the company from signing any such agreement. Lisa Su also noted that the new joint venture would rely solely on AMD’s own intellectual property. This raises some significant questions about what, exactly, AMD can license. One of the problems with trying to talk about the x86 patent agreements between Intel and AMD is that what we refer to as “x86” is a collection of dozens of patents filed over decades by both companies. Intel still holds patents on major features like SSE4.x, AVX, and AVX2, while AMD has patents for AMD64.
This is where Su’s statement that AMD was licensing technology, not products, may come into play. It’s entirely possible that AMD wants to use the building blocks of Zen or AMD’s SeaMicro fabric and combine them with its own homegrown capabilities in ways that wouldn’t infringe on patents held by Intel or other players in the industry. In the short term, this joint venture gave AMD a modest shot of capital and reassured investors that the company had found new markets and options for its products — longer-term, AMD has the option to earn royalty revenue on product sales on top of the IP licensing agreements.
The other major question for AMD was whether the company was looking to console refreshes to dramatically change its semicustom revenue. Right now, the semicustom business is the only part of AMD turning a net profit, but we’ve known for a long time that the business was front-loaded — meaning AMD’s margins and net profit per console sale were highest at the beginning and would decrease over time.
News that the PS4 (and almost certainly the Xbox One) would receive major refreshes this cycle has prompted speculation that those refreshes would lead to commensurately higher revenue for the embattled chip manufacturer. CEO Lisa Su didn’t come right out and say this wouldn’t happen, but she didn’t endorse the idea, either. According to Su, AMD negotiated the contracts for its current semicustom design wins well in advance, and the current draw down in average selling price (ASPs) is in-line with that contract.
AMD did note, however, that it had won three major semicustom contracts that it expected to deliver in the back half of this year, with an estimated value of $1.5 billion over 3-4 years. The smart money here is on the Nintendo NX, the PS4 Neo, and (presumably) a next-generation Xbox One. AMD did not clarify if that $1.5 billion was additive to its existing game console business or represented the estimated value of its console segment over the relevant time period.
AMD didn’t say very much about Zen this time out, though Lisa Su noted that the chip is moving through internal validation, and meeting AMD’s internal goals. Asked when the chip would be available in market, she stated it would begin sampling to priority customers in Q2 2016 with expected data center availability in 2017. There were no updates on mainstream consumer availability.
AMD’s mainstream PC business remains, in a word, terrible. That’s a blunt way to summarize the company, but it’s a fair one, as illustrated by the following slide:
Compare these results with Q1 2014. Two years ago, AMD’s Computing Solutions (CPUs and APUs) revenue was $663 million, while its Graphics and Visual Solutions (consoles and graphics) revenue was $734 million. Net revenue in Q1 2014 was $1.397 billion.
Two years later, AMD’s total revenue is just $832 million. That’s partly due to increased competition from Nvidia, partly the result of weak performance from AMD’s own APUs, partly the general downturn in the PC market, and partly the revenue structure AMD accepted for its console division. All of these factors have played a role; the decline in the PC market alone probably cost AMD between $70-$110 million in quarterly revenue.
The next 12 months will decide whether AMD survives in anything like its current form. The company has lined up new technology partners, Zen appears to be ramping on schedule, Polaris is still on track for a Q2 launch, and we know console refreshes are on the way. Those are strong factors — but the margin for error between now and Zen’s mainstream availability for consumer and server hardware is just about nil.