CEO Lisa Su kicked off the event, which took place on Thursday afternoon, by recapping the company’s recent product and financial progress. Along the way, she asserted that AMD aims to be a “best-in-class growth franchise,” and -- with AMD’s recent revenue growth having helped it grow its R&D budget -- promised the company would have “a very strong cadence” in terms of the pace of its CPU and GPU innovation.
Not long afterwards, Su unveiled an upbeat set of 3-year financial targets: AMD is aiming for a compound annual revenue growth rate (CAGR) of about 20% through 2023, while also ultimately achieving a 50%-plus gross margin (GM), a mid-20s percentage operating margin and a 15%-plus free cash flow (FCF).
For comparison, AMD’s FactSet consensus revenue estimates prior to the Analyst Day event implied 30% 2020 revenue growth, 19% 2021 growth and 11% 2022 growth. In 2019, the company had a non-GAAP GM of 43%, a non-GAAP operating margin of 12% and an FCF margin of 4%.
AMD’s stock immediately jumped after the targets were released, and finished after-hours trading up 6% to $51.01. Ahead of the Analyst Day event, shares fell 4% in regular trading amid a sharp market selloff.
Following Su’s remarks, AMD execs Mark Papermaster, David Wang, Rick Bergman and Forrest Norrod made a slew of noteworthy product, technology and market share disclosures.
Papermaster, AMD’s CTO, disclosed that AMD is working on 2.5D and 3D packaging approaches for chiplets within a larger chip package (AMD’s PC and server CPUs containing chiplets currently rely on 2D packaging). X3D, a packaging approach in development that will rely on both 2.5D and 3D packaging, is declared to deliver more than a 10x increase in bandwidth density. The disclosures come as Intel (INTC) - Get Report works on 2.5D and 3D packaging technologies of its own (respectively called EMIB and Foveros), along with a solution (known as Co-EMIB) that combines the two.
Papermaster also shared details about the third-gen version of AMD’s Infinity high-speed interconnect technology (previously called Infinity Fabric), which was briefly mentioned on Wednesday in the press release for a major supercomputer win. Whereas the second-gen version of Infinity (launched in 2019) could separately connect CPUs and GPUs, the third-gen version will be able to connect CPUs and GPUs with each other, while supporting a unified memory space across all of the chips.
Wang, SVP of AMD’s Radeon Technologies Group, asserted that AMD’s next-gen gaming GPU architecture (known as RDNA 2) will deliver a 50% increase in performance per watt relative to the current RDNA architecture. RDNA 2 will also (like Nvidia’s high-end gaming GPUs) support variable rate shading and real-time ray tracing, with AMD promising a common ray-tracing architecture across PCs and game consoles. A line of RDNA 2-based PC gaming GPUs (codenamed Navi 2X) are due by the end of the year.
In addition, Wang provided details about CDNA, a next-gen architecture for server GPUs that’s meant to succeed the current GCN architecture and has been optimized for high-performance computing (HPC) and AI/deep learning workloads. Like some of Nvidia’s (NVDA) - Get Report server GPUs, CDNA-based GPUs will feature circuitry dedicated to handling deep learning operations.
A CDNA-based GPU relying on Taiwan Semiconductor’s (TSM) - Get Report 7-nanometer (7nm) manufacturing process node is due later this year. CDNA 2, a successor architecture that will support 3rd-gen Infinity and rely on a more advanced process node (possibly 5nm), will arrive by 2022. Nvidia is expected by many to unveil a server GPU based on its next-gen Ampere architecture later this month.
Bergman, AMD’s EVP of Computing and Graphics, confirmed that AMD’s first 4th-gen Ryzen PC CPUs (most likely desktop products) will launch later this year. He also disclosed that AMD shipped 43 million PC CPU units in 2019, up from 31 million in 2018, that its PC CPU average selling price (ASP) rose by 14% and that AMD estimates its PC CPU share rose to 17% from 13%.
Also disclosed: AMD estimates its share of the commercial (business) PC market rose to 11% last year from 8%, and that its notebook CPU share rose to 16% from 12%.
Norrod, GM of AMD’s Data Center and Embedded Solutions Group, confirmed that AMD’s third-gen Epyc server CPUs (codenamed Milan) will ship in late 2020. He asserted that, whereas second-gen Epyc CPUs (codenamed Rome) can address 80% of workloads across cloud, enterprise IT and HPC environments, Milan will be able to address all of them.
Norrod also predicted AMD would reach a double-digit server CPU in Q2 (the company previously said it would happen in mid-2020), and forecast that 150-plus, Epyc-powered, cloud computing instances would be available this year, up from 106 in 2019.
CFO Devinder Kumar, meanwhile, shared that AMD is aiming to receive more than 30% of its 2023 revenue from data center products -- up from 15% in 2019 -- even though it’s forecasting its PC/gaming revenue will grow at a mid-teens annual clip. Kumar also indicated AMD is open to using stock buybacks to offset equity plan dilution, as well as additional, unnamed “shareholder return vehicles.”
Following Kumar’s presentation, Su provided some comments about the impact of the COVID-19 outbreak on AMD. She declared AMD is back to “near-normal” supply chain capacity following recent disruptions, and while noting Chinese demand has been hurt, added that cloud-related demand has been stronger than expected in Q1.
Su said AMD is maintaining its Q1 revenue guidance of $1.75 billion to $1.85 billion, while adding that revenue could come in at the low end of this range. The company is also reiterating full-year guidance for 28% to 30% revenue growth.