At the ongoing SC19 AMD’s SVP Forrest Norrod gave a nice interview to Eric Jhonsa of TheStreet/Real Money, summarized here. Norrod went over a whole bunch of new supercomputer deals and expanded partnership with AWS. What I find most intriguing was what he said about Zen 3, arriving in 2020. From the article:
” Norrod observed that — unlike Zen 2, which was more of an evolution of the Zen microarchitecture that powers first-gen Epyc CPUs — Zen 3 will be based on a completely new architecture.
Norrod did qualify his remarks by pointing out that Zen 2 delivered a bigger IPC gain than what’s normal for an evolutionary upgrade — AMD has said it’s about 15% on average — since it implemented some ideas that AMD originally had for Zen but had to leave on the cutting board. However, he also asserted that Zen 3 will deliver performance gains “right in line with what you would expect from an entirely new architecture.”
So Zen 3 not only will be using TSMC’s 7 nm+ process, but it is actually a NEW architecture! Zen 1, being a completely new architecture, achieved nearly 50% IPC improvement over Bulldozer. Even if Zen 3 achieves an IPC improvement of 15%, when you combine that with 7nm+ (EUV process), what kind of gains are we looking at?!
Now I am really looking forward to seeing what Milan can do. Norrod also mentioned that the 64 cores Epyc is surprisingly popular, the demand is higher than they had anticipated. While he gave reassurance that TSMC supply is not a problem, this may explain why highly binned 8 cores chiplets are flying out the window.
Meanwhile, AMD put out a nice press release, here.
First, AWS is announcing two new Elastic Compute Cloud (Amazon EC2) instances, using Epyc Rome.
France’s Joliet-Curie supercomputer is now using the high powered Epyc 7H12 CPU.
Cray/HPE had announced ARCHER2 and Vulcan will be using Epyc for their supercomputers.
San Diego Supercomputer Center will be using Epyc, supplied by Dell/EMC, detailed by insideHPC, here. ETH Zurich will also be using Epyc.
Even more importantly, AMD introduced ROCm 3.0. This is an open source software used for GPU compute. Nvidia’s proprietary CUDA is already widely used, but ROCm is open sourced and converts CUDA codes effortlessly so that they can be run on AMD’s Radeon GPUs. ROCm is crucial for AMD to break Nvidia’s monopoly in datacenter GPU. It’s good to see AMD keep plugging away on the software side as well.
Cowan’s Matthew Ramsay spoke to Lisa Su and came away so impressed that he upped the PT from 40 to 47. This StreetInsider article quoted Ramsay:
“Ramsay notes the topic of AMD’s long-term model and share targets remains of keen interest to investors. Management only commented they plan to update these targets soon and we believe an analyst day in early 2020 is likely. While the analyst leaves out any guidance speculation regarding the long-term model, they believe 20%+ dollar share of AMD’s x86 target markets of desktop, notebook and server, along with gross margins approaching 50% are reasonable long-term targets.Should AMD execute against this type of long-term model, they analyst lays out a bull case revenue, margin and EPS scenario that yields $3/EPS for discussion.”
Well, I am anticipating AMD to earn $0.60+ for 2019, $1.30+ for 2020. If AMD makes $2.0+ for 2021, what kind of PE ratio will investors apply to AMD? I am guessing PE of 30 to 50, depending on the macro environment at the time. I am very much looking forward to Analyst Day 2020. Talks of $3 earnings is music to my ears!
Quanta, a huge OEM, is now selling some very well designed and differentiated Rome servers, detailed in this Servethehome article. For Epyc to do well, support from OEMs such as Quanta is crucial.
Viking Enterprise launched a NVMe storage platform, using Epyc 7002, detailed here, also by Servethehome.
Last but not least, Rosenblatt Securities’ Hans Mosesmann published yet another great research note, after he spoke to Ruth Cotter, SVP of Marketing at AMD, detailed in this Reddit post.
First Hans reiterated his PT of 52.
Then he went through why AMD will be gaining market shares in both CPU and GPU.
What I find most fascinating is:
“This past August, AMD’s BoD gave Lisa Su a Special Value Creation Award based on CAGR
milestones which include stock price target levels of $65 and $90, which is incremental to the existing long term equity incentive plan which expires on 8/9/22.”
I feel very encouraged by this news, previously not known to the public. Of course AMD would love to keep Lisa Su as CEO indefinitely, what with IBM rumors popping up constantly. AMD would love to reward her as much as possible. What better way to give her a big reward, once AMD share price reaches 65, then 90! Whatever amount Lisa Su gets when AMD reaches such lofty valuation, it would be worth every penny!
On the other hand, it also shows internally that AMD feels that these CAGR milestones and such price targets are possible and even likely to be achieved. I love that optimism. Go AMD!
AMDinvestorblog 