 From theCUBE Studios in Palo Alto in Boston, bringing you data-driven insights from theCUBE and ETR. This is Breaking Analysis with Dave Vellante. AWS is pointing the way to a revolution in system architecture. Much in the same way that AWS defined the cloud operating model last decade, we believe it is once again leading in future systems design. The secret sauce underpinning these innovations is specialized designs that break the stranglehold of inefficient and bloated centralized processing and allows AWS to accommodate a diversity of workloads that span cloud, data center, as well as the near and far edge. Hello and welcome to this week's Wikibon Cube Insights powered by ETR. In this Breaking Analysis, we'll dig into the moves that AWS has been making, which we believe define the future of computing. We'll also project what this means for customers, partners, and AWS's many competitors. Now let's take a look at AWS's architectural journey. The IaaS revolution, it started by giving easy access, as we all know, to virtual machines that could be deployed and decommissioned on demand. Amazon at the time used a highly customized version of Zen that allowed multiple VMs to run on one physical machine. The hypervisor functions were controlled by x86. Now according to Warner Vogels, as much as 30% of the processing was wasted, meaning it was supporting hypervisor functions and managing other parts of the system, including the storage and networking. These overheads led to AWS developing custom ASICs that helped to accelerate workloads. Now in 2013, AWS began shipping custom chips and partnered with AMD to announce EC2 C3 instances. But as the AWS cloud started to scale, they really weren't satisfied with the performance gains that they were getting, and they were hitting architectural barriers. That prompted AWS to start a partnership with Anaperta Labs, this was back in 2014, and they launched then EC2 C4 instances in 2015. The ASIC and C4 optimized offload functions for storage and networking, but still relied on Intel's Xeon as the control point. AWS shelled out a reported $350 million to acquire Anaperta in 2015, which is a meager sum to acquire the secret sauce of its future system design. This acquisition led to a modern version of project Nitro in 2017. Nitro offload cards were first introduced in 2013. At this time, AWS introduced C5 instances and replaced Zen with KVM, and more tightly coupled the hypervisor with the ASIC. Vogels shared last year that this milestone offloaded the remaining components, including the control plane, the rest of the IO, and enabled nearly 100% of the processing to support customer workloads. It also enabled a bare metal version of the compute that spawned the partnership, the famous partnership with VMware to launch VMware cloud on AWS. Then in 2018, AWS took the next step and introduced Graviton, its custom designed ARM based chip. This broke the dependency on x86 and launched a new era of architecture which now supports a wide variety of configurations to support data intensive workloads. Now these moves proceeded other AWS innovations including new chips optimized for machine learning and training and inferencing and all kinds of AI. The bottom line is AWS has architected an approach that offloaded the work currently done by the central processing unit and most general purpose workloads like in the data center. It has set the stage in our view for the future allowing shared memory, memory disaggregation and independent resources that can be configured to support workloads from the cloud all the way to the edge. And Nitro is the key to this architecture. And to summarize, AWS Nitro, think of it as a set of custom hardware and software that runs on an ARM based platform from Annapurna. AWS has moved the hypervisor, the network, the storage virtualization to dedicated hardware that frees up the CPU to run more efficiently. This in our opinion is where the entire industry is headed. So let's take a look at that. This chart pulls data from the ETR data set and lays out key players competing for the future of cloud data center in the edge. Now we've superimposed NVIDIA up top and Intel. They don't show up directly in the ETR survey but they clearly are platform players in the mix. We covered NVIDIA extensively in previous breaking analysis and won't go too deep there today. But the data shows net scores on the vertical axis. That's a measure of spending velocity. And then it shows market share in the horizontal axis which is a measure of pervasiveness within the ETR data set. We're not going to dwell on the relative positions here. Rather let's comment on the players and start with AWS. We've laid out AWS, how they got here and we believe they are setting the direction for the future of the industry. AWS is really pushing migration to its ARM based platforms. Pat Morehead at the 6.5 summit spoke to Dave Brown who heads EC2 and AWS. And he talked extensively about migrating from x86 to AWS's ARM based Graviton 2. And he announced a new developer challenge to accelerate that migration to ARM instances, Graviton instances. And the end game for customers is a 40% better price performance. So a customer running 100 server instances can do the same work with 60 servers. Now there's some work involved by the customers to actually get there but the payoff, if they can get 40% improvement in price performance is quite large. Imagine this, AWS currently offers 400 different EC2 instances. Last year as we reported, sorry, last year as we reported earlier this year nearly 50% of the new EC2 instances. So nearly 50% of the new EC2 instances shipped in 2020 were ARM based. And AWS is working hard to accelerate this pace. It's very clear. Now let's talk about Intel. I'll just say it, Intel is finally responding in earnest and basically it's taking a page out of ARM's playbook. We're going to dig into that a bit today. In 2015, Intel paid $16.7 billion for Altera, a maker of FPGAs. Now also at the 6.5 summit, Naven Shenoy of Intel presented details of what Intel is calling an IPU, it's Infrastructure Processing Unit. This is a departure from Intel norms where everything is controlled by a central processing unit. IPUs are essentially smart NICs as our DPUs. They don't get caught up in all the acronym soup. As we've reported, it's all about offloading work and disaggregating memory and evolving SOCs, system on chip and SOPs, system on package. But just let this sink in a bit for a moment. Intel's moves this past week, it seems to us anyway, are designed to create a platform that is nitro-like. And the basis of that platform is a $16.7 billion acquisition. Just compare that to AWS's $350 million tuck-in of Annapurna. That is incredible. Now Shenoy said in his presentation, rough quote, we've already deployed IPUs using FPGAs in a very high volume at Microsoft Azure, and we've recently announced partnerships with Baidu, JD Cloud and VMware. So let's look at VMware. VMware is the other really big platform player in this race. In 2020, VMware announced Project Monterey, you might recall that, it's based on the aforementioned FPGAs from Intel. So VMware is in the mix and it chose to work with Intel, most likely for a variety of reasons. One of the obvious ones is all the software that's running on VMware. It's been built for x86 and there's a huge install base there. The other is Pat was heading VMware at the time and when Project Monterey was conceived. So I'll let you connect the dots if you like. Regardless, VMware has a nitro-like offering. In our view, its optionality, however, is limited by Intel, but at least it's in the game. It appears to be ahead of the competition in this space. AWS, notwithstanding, because AWS is clearly in the lead. Now, what about Microsoft and Google? Suffice it to say that we strongly believe that despite the comments that Intel made about shipping FPGAs in volume to Microsoft, that both Microsoft and Google, as well as Alibaba will follow AWS's lead and develop an ARM-based platform like Nitro. We think they have to in order to keep pace with AWS. Now, what about the rest of the data center pack? Well, Dell has VMware. So despite the split, we don't expect any real changes there. Dell is going to leverage whatever VMware does and do it better than anyone else. Cisco is interesting in that it just revamped its UCS, but we don't see any evidence that it has a nitro-like plans in its roadmap. Same with HPE. Now, both of these companies have history and capabilities around silicon. Cisco designs its own chips today for carrier-class use cases and HPE as we've reported, probably has some remnants of the machine hanging around, but both companies are very likely in our view to follow VMware's lead and go with an Intel-based design. What about IBM? Well, we really don't know. We think the best thing IBM could do would be to move the IBM cloud, of course, to an ARM-based nitro-like platform. So we think even the mainframe should move to ARM as well. I mean, it's just too expensive to build a specialized mainframe CPU these days. Now, Oracle, they're interesting. If we were running Oracle, we would build an ARM-based nitro-like database cloud where Oracle, the database, runs cheaper, faster, and consumes less energy than any other platform that would dare to run Oracle. And we'd go one step further and we would optimize for competitive databases in the Oracle cloud. So we would make OCI run the table on all databases and be essentially the database cloud. But you know, back to sort of FPGAs, we're not overly excited about the market. AMD is acquiring Xilinx for $35 billion, so I guess that's something to get excited about, I guess. But at least AMD is using its inflated stock price to do the deal. But we honestly, we think that the ARM ecosystem will obliterate the FPGA market by making it simpler and faster to move to SOC with far better performance, flexibility, integration, and mobility. So again, we're not too sanguine about Intel's acquisition of Altera and the moves that AMD is making in the long-term. Now, let's take a deeper look at Intel's vision of the data center of the future. Here's a chart that Intel showed depicting its vision of the future of the data center. What you see is the IPUs, which are intelligent NICs and they're embedded in the four blocks shown and they're communicating across a fabric. Now you have general purpose computer in the upper left and machine intelligent on the bottom left, machine intelligence apps. And up in the top right, you see stored services and then the bottom right, a variation of alternative processors. And this is Intel's view of how to share resources and go from a world where everything is controlled by a central processing unit to a more independent set of resources that can work in parallel. Now, Gelsinger has talked about all the cool tech that this will allow Intel to incorporate including PCI and Gen 5 and CXL memory interfaces and or CXL memory, which are interfaces that enable memory sharing and disaggregation and 5G and 6G connectivity and so forth. So that's Intel's view of the future of the data center. Let's look at ARM's vision of the future and compare them. Now there are definite similarities as you can see especially on the right hand side of this chart. You've got the blocks of different processor types. These of course are programmable and you notice the high bandwidth memory, the HBM3 plus the DDRS on the two sides kind of book ending the blocks. That's shared across the entire system and it's connected by PCIe Gen 5, CXL or CCIX multi-dye socket. So, you know, you may be looking at the same, okay, two sets of block diagrams, big deal. Well, while there are similarities around disaggregation and I guess implied shared memory in the Intel diagram and of course the use of advanced standards, there are also some notable differences. In particular, ARM is really already at the SOC level whereas Intel is talking about FPGAs. Neoverse, ARM's architecture is shipping in test mode and will have end market product by year end 2022. Intel is talking about maybe 2024, we think that's aspirational or 2025 at best. ARM's roadmap is much more clear. Now Intel said it will release more details in October so we'll pay attention then maybe we'll recalibrate at that point but it's clear to us that ARM is way further along. Now the other major difference is volume. Intel is coming at this from a high end data center perspective and you know, presumably plans to push down market or out to the edge. ARM is coming at this from the edge, it's low cost, low power, superior price performance. ARM is winning at the edge and based on the data that we shared earlier from AWS, it's clearly gaining ground in the enterprise. History strongly suggests that the volume approach will win not only at the low end but eventually at the high end. So we want to wrap by looking at what this means for customers and the partner ecosystem. The first point we'd like to make is follow the consumer apps. This capability, the capabilities that we see in consumer apps like image processing and natural language processing and facial recognition and voice translation, these inference capabilities that are going on today in mobile will find their way into the enterprise ecosystem. 90% of the cost associated with machine learning in the cloud is around inference. In the future, most AI in the enterprise and most certainly at the edge will be inference. It's not today because it's too expensive. This is why AWS is building custom chips for inferencing to drive costs down so it can increase adoption. Now the second point is we think that customers should start experimenting and see what you can do with ARM based platforms. Moore's law is accelerating. At least the outcome of Moore's law, the doubling of performance every 18 to 24 months. It's actually much higher than that now when you add up all the different components in these alternative processors. Just take a look at Apple's A15 chip. And ARM is in the lead in terms of performance, price performance, cost and energy consumption. By moving some workloads onto Graviton for example, you'll see what types of cost savings you can drive for which applications and possibly generate new applications that you can deliver to your business. Put a couple of engineers in the task and see what they can do in two or three weeks. You might be surprised or you might say, hey, it's too early for us, but you'll find out and you may strike gold. We would suggest that you talk to your hybrid cloud provider as well and find out if they have a nitro. We shared that VMware, they've got a clear path as does Dell because they're VMware cousins. What about your other strategic suppliers? What's their roadmap? What's the timeframe to move from where they are today to something that resembles nitro? Do they even think about that? How do they think about that? Do they think it's important to get there? So if so or if not, how are they thinking about reducing your costs and supporting your new workloads at scale? Now for ISVs, these consumer capabilities that we discussed earlier, all these mobile and automated systems and cars and things like that, biometrics, another example, they're going to find their way into your software and your competitors are porting to ARM. They're embedding these consumer-like capabilities into their apps. Are you? We would strongly recommend that you take a look at that, talk to your cloud suppliers and see what they can do to help you innovate, run faster and cut costs. Okay, that's it for now. Thanks to my collaborator, David Flore, who's been on this topic since early last decade. Thanks to the community for your comments and insights. And hey, thanks to Patrick Moorhead and Daniel Newman for some timely interviews from your event. Nice job, fellas. Remember, I publish each week on wikibon.com and siliconangle.com. These episodes are all available as podcasts. Just search for Breaking Analysis Podcast. You can always connect with me on Twitter at dvolonte or email me at david.volonte at siliconangle.com. Appreciate the comments on LinkedIn and Clubhouse. So follow us. If you see us in a room, jump in and let's riff on these topics. And don't forget to check out etr.plus for all the survey data. This is Dave Vellante for theCUBE Insights, powered by ETR. Be well and we'll see you next time.