In case you haven’t heard, AMD had a killer Q2. Its revenue of $3.85B represents a 99% growth year-over-year (YoY). Additionally, gross margin came in at 48% (up 4% YoY), and operating income at $831M.
In the datacenter group, the numbers were even more impressive. Enterprise, Embedded, and Semi-custom (EESC) reported $1.6B, up 183% YoY. Yes, 183%. Operating income came in at $398M, versus $33M a year ago.
These numbers come as Intel’s Datacenter Group saw an 8% drop in platform sales (YoY) and a 20% drop in cloud sales YoY.
Can AMD maintain this momentum through the second half of 2021? Is AMD’s growth directly leading to Intel’s dip in the cloud? And how are AMD’s OEM partners adding to this success? I’ll try to unpack all of this over the following few paragraphs.
First – Looking at the numbers
As mentioned, AMD’s EESC numbers were quite impressive. Let’s disclose a couple of significant non-server contributions before getting into the fun stuff. Baked into the EESC number is Tesla’s selection of AMD’s Embedded Ryzen and Radeon GPU for its Model S and Model X consoles.
In the semi-custom space, AMD benefited from a strong console market. Additionally, the company announced that Valve selected AMD to power its SteamDeck handheld consoles, which will launch in December.
These embedded and semi-custom wins tend to be lower-margin businesses while volumes can be very high.
Now, let’s pivot to EPYC. Even when backing out the success AMD saw in embedded and semi-custom, EPYC killed. In fact, during the earnings call, when asked by a financial analyst if EPYC’s growth outpaced the 19% quarter-over-quarter (QoQ) growth the EESC segment experienced, Lisa Su replied, “yes.” So, we know the EPYC’s growth wasn’t just tremendous YoY, but also sequential. Further, we also learned that Milan (3rd Generation EPYC) demonstrated the fastest ramp since AMD re-entered the server market. This is all excellent news for the company.
Second – Where is AMD winning?
Su was very clear on where AMD saw strong growth for EPYC in Q2: Cloud and high-performance computing (HPC). The company secured Google Cloud’s Tau instances on the cloud front and additional wins at Microsoft Azure and Alibaba Cloud. Also, EPYC is the processor of choice for Oracle Cloud’s HeatWave service (the industry-leading deep analytics engine for MySQL).
On the HPC front, AMD also showed significant growth. In the June listing of the Supercomputing Top500, EPYC increased its presence by a factor of five (as in five times more platforms). Equally significant, of the 58 new entries into the June 2021 listing, EPYC represents half. Additionally, an EPYC based system (HPE Cray) made the highest x86 placement on the list.
Third – Winning the enterprise
On the enterprise front, Su mentioned that AMD continues to find momentum in this segment and that she expects the second half of 2021 will show continued growth. Anecdotally, I’ve spoken with several IT professionals across various industries, and indeed, EPYC-based systems are finding traction.
I believe that AMD’s partners are doing the right things to find success with their respective portfolios. In addition to over 100 Milan-based platforms in the market, partners like HPE are investing for success. The company is working closely with AMD to drive channel programs and momentum–an absolute must for success in enterprise and below (for more information on HPE and AMD, read this report).
The key with the enterprise segment is patience. This segment requires a lot of front-side investing and time for those seeds to take root. And more time. And when you thought it was time–wait a little longer. For you Geoffrey Moore fans out there, think late majority to laggards in terms of adoption.
Enterprise customers look to the cloud and other hyperscalers to see whether technologies are worthy of adoption. This includes processors. And in this regard, Google Cloud’s recent selection of EPYC to power its Tau instances is perhaps the best endorsement the company could get. Seeing what the likes of GCP, Oracle Cloud, Azure and AWS are doing with EPYC indicates that this CPU has a performance and price-performance profile worthy of consideration.
The other influence on enterprise IT is other enterprises’ IT organizations. There is nothing more motivating for me as a CIO than to see my competition find great success with new technology. As AMD and its partners start touting wins in Fortune 500 customers, I believe we’ll see an acceleration of adoption.
Interestingly, AMD’s financial presentation showed the company securing multiple “high volume” Ryzen (client CPU) wins. This is a strong indicator of EPYC’s acceptance in the enterprise.
Fourth – What is AMD doing right?
As mentioned, I think AMD is doing many good things to drive broad adoption in the enterprise. First, it worked with its OEM partners (Cisco, Dell, HPE, Lenovo, Supermicro, etc.) to deliver differentiated platforms in the market, from a feature richness perspective to performance to pricing. Consider AMD’s “design in” efforts a huge success.
Second, the company is working with these partners to deliver differentiated solutions to the market, such as HCI, VDI, virtualized desktop, commercial HPC and the like. AMD and its partners have done a solid job developing easily consumed IT solutions that can be quickly deployed in the datacenter. Once again, I would consider AMD’s efforts to build “design-out” solutions a success.
Finally, I believe AMD and its partners are doing a solid job in managing the channel. By channel, I mean resellers that range from Connection to SHI to WWT. These are the real influencers of IT organizations of all sizes. Along this critical “last mile” to the customer, I see strong signals. Anecdotally, I’ve spoken with channel executives who speak to the enthusiasm and availability of AMD, along with its willingness to do what it takes to be successful. Anybody who has worked channel management knows how much of a personalized effort it takes to be successful. It takes sales trainings, tech trainings, lunch and learns, floor days, campaigns, weekly attainment calls, etc. But if you can be a champion with your channel partners, your company will reap significant benefits. By all indications, this is the case with AMD.
I would like to see the company and its partners lean more heavily into some of the more modern workloads for all of this goodness, as it applies to positioning and messaging. Containers and cloud-native enablement, multi-cloud and the edge are the workloads and initiatives occupying the thoughts and whiteboards of IT organizations across the globe, and EPYC-based servers can be an ideal building block. If AMD and its partners can succeed in one of these workloads, the rest of the datacenter will soon follow.
Fifth – Who is AMD’s biggest threat?
I believe EPYC will continue to eat into Xeon’s datacenter stranglehold. This is not a provocative or controversial statement. All of the indicators are there. But while Xeon is still well north of 80% market share, I think AMD needs to keep an eye on the Arm-based CPU companies and server players such as Ampere and Bamboo Systems. And when the Nvidia-Arm acquisition finalizes (which it will), I believe there’s a real threat to x86 in general that has to be given serious consideration.
Arm is popular in the cloud. Graviton2’s footprint continues to expand in AWS. Likewise, Oracle Cloud’s A1 Ampere instance is so popular, it is challenging to find available cycles.
But more importantly, Arm is also popular in the open-source community. It has first-class status with virtually all of the projects, and the commercial distributions optimize for it and x86. Compatibility? The cloud-native world doesn’t care, and cloud-native is where the enterprise is moving.
This is not to say Arm is overtaking x86 in the next couple of years, or even the next few years. But, it is to say that both AMD and Intel better keep an eye on these Arm players, lest we forget how “Wintel” overtook UNIX and RISC in the datacenter. Does anybody remember SPARC?
To quote the Cars (one of many awesome bands from Boston), “Let the good times roll.” AMD slaughtered the quarter and raised its guidance for the rest of the year, projecting 60% growth (versus a forecast of 37% at the beginning of the year). In my particular area of interest, EPYC also had a killer quarter.
I believe enterprise is now lining up for EPYC, and to be really cheesy, I think EPYC will also see an epic second half. I am looking forward to seeing what Lisa Su, Devinder Kumar and others have to say in October.