Chipmaker AMD has doubled its annual revenue growth outlook to 60% year-on-year, from the approximately 31% growth it projected at the beginning of the year, as CEO Dr Lisa Su pointed to what she described as a "significant inflection point" for the company driven by a doubling of enterprise segment revenues, expanded FPGA-as-a-Service and SmartNIC deployments at Tier 1 hyperscalers and synergies from its Xilinx acquisition.
(The company completed the industry's largest semiconductor company acquisition in February 2022 for an estimated $50 billion -- up from an initial $35 billion as AMD's share price had surged. Xilinx, a specialist in programmable and custom chips, is becoming the AMD Adaptive and Embedded Computing Group (AECG), led by former Xilinx CEO Victor Peng and is focussed on FPGA, Adaptive SoC and software roadmaps for the company.)
Asked by analysts to flesh out how meaningful data center was as a percentage of revenue in the quarter, vs server CPU, data center GPU and now FPGAs from Xilinx, Dr Su said that DC demand was "low 20s percentage of our overall revenue", adding that she was "excited" about the outlook ahead of new CPU launches in H2.
Follow The Stack today on LinkedIn
She said on the earnings call: "We're excited about the GPU portfolio as well. GPUs for us are a longer-term road map, similar to what we did on the CPU side. We had been more focused on supercomputing and HPC, so that was strong for us last year and that's why we're flattish year-on-year. We're very engaged on the AI front now, continuing our investments in our software stack and working with cloud guys to optimise our software stack.
AMD revenues hit a fresh quarterly record of $5.9 billion, up 71% year-over-year and 22% quarter-over-quarter in Q1, it reported late Tuesday May 3. Computing and graphics segment revenue was $2.8 billion. AMD's enterprise, embedded and semi-custom segment revenue surged 88% on-year to a record $2.5 billion. The sharp AMD outlook upgrade for 2022 came despite broader supply chain constrictions. Dr Su told analysts on an earnings call that China's Covid restrictions "haven't had any significant impact on our own shipments in our own supply chain."
It's an astonishing return from the "dead" for a "perpetually struggling" company that in 2013 was forced to sell its HQ to make some quick cash... (it then leased it back so it still had a home.)
2022 AMD outlook upgrade comes despite supply chain tightness
Pressed on broader wafer shortages, etc. she said: "We have been working with some customers that have had some customer build delays and that is contemplated in our second quarter guidance. We're going to continue to work on supply optimization with the addition of Xilinx, some of the, let's call it, more mature nodes. 16-nanometer and above wafer supply is still somewhat constrained. We're working... to try to bring more supply on board there, as well as continuing to ramp our overall capacity to support a very strong sort of next few quarters."
"Both the Xilinx portfolio and sort of the organic AMD portfolio do use TSMC, so we are complementary there as our primary wafer supplier. We use very similar substrate suppliers as well.
The Xilinx portfolio tends to be on more mature nodes. Although there is some 7-nanometer, the majority of the portfolio is on 16-nanometer and above. So I would say there's not that much fungibility there. However, on the back end, on the substrate side, there is very good fungibility across the portfolio. We have invested significantly over the last 18 months in securing supply and capacity and we're seeing it come online... we're continuing to dimension the company for just a much larger business. It's a lot of supply that we're bringing online. We're working very much with Victor and his team as well because he has a strong backlog and strong demand, and we're looking to use all of the AMD assets to also accelerate some of his builds," she added.
AMD Q1 earnings call highlights AI, Xilinx opportunities
The company has been building the hype for its next-generation Genoa server processors after impressive performance and sales for its Epyc range. Dr Su said AMD had expanded customer and partnering sampling for Genoa in the quarter and remains on track to launch Genoa in the second half of the year -- "we expect [its release] to continue our share gain trajectory based on expanded cloud, enterprise and HPC customer adoption. "
Atos, Cisco, Dell, HPE, Lenovo and Supermicro all launched servers featuring new AMD CPUs in the quarter while company development of its higher core count Bergamo processors optimised for high throughput cloud workloads continues to progress well, with shipments on track to begin in the first half of 2023.
Demand for compute powering AI inferencing continues to be a growth market for many IT vendors and AMD is no exception. Xilinx's Victor Peng noted: "We have this AI engine that is already deployed in production in a number of embedded applications and endpoints and also edge devices like in cars, that are doing a lot of image recognition, all kinds of inference applications, and that same architecture can be scaled and brought into the CPU product portfolio. That is exactly our plan. We're working on unified overall software to enable the broad portfolio but also especially in AI... we're definitely going to be leaning in, in AI both inference and training, and I would say, end-to-end because we have endpoints, we have edge devices, both computing and embedded devices and in the cloud and enterprise. So we're very excited about that revenue synergy opportunity."