On this week’s episode of the MI&S Datacenter Podcast, hosts Matt and Paul analyze the week’s top datacenter and datacenter edge news. This week they are talking Microsoft, Supercomputing 2024, Atom Computing, and more!
Watch the video here:
Listen to the audio here:
2:41 Little AI Guys
7:10 Supercomputing Goes Mainstream
13:20 2 Dozen Logical Qubits
22:43 Chips & Chips & Chips At Microsoft Ignite
32:18 Getting To Know Us – The Thanksgiving Edition
Little AI Guys
https://www.microsoft.com/en-us/research/articles/magentic-one-a-generalist-multi-agent-system-for-solving-complex-tasks/
Supercomputing Goes Mainstream
https://www.linkedin.com/feed/update/urn:li:activity:7264651754347085825/
https://www.linkedin.com/feed/update/urn:li:activity:7265018456003997696/
2 Dozen Logical Qubits
https://atom-computing.com/high-fidelity-gates-and-the-worlds-largest-entangled-logical-qubit-state/
https://www.forbes.com/sites/moorinsights/2024/01/25/microsoft-uses-ai-and-hpc-to-analyze-32-million-new-materials/lion-new-materials/
Chips & Chips & Chips At Microsoft Ignite
https://www.datacenterknowledge.com/cloud/microsoft-ignite-2024-new-azure-data-center-chips-unveiled
Disclaimer: This show is for information and entertainment purposes only. While we will discuss publicly traded companies on this show. The contents of this show should not be taken as investment advice.
Matt Kimball: We are now live, Paul. We are now live. I didn’t have my makeup treatment done this morning, so we’re in trouble. Hey, welcome to another edition of the Moor Insights & Strategy DataCenter podcast. My name is Matt Kimball, and I am joined by the smartest guy I know, Paul Smith-Goodson, who covers AI and quantum computing for Moor Insights & Strategy. We are here to talk about all things DataCenter. Before we get going, a couple of things. This is not to be taken as financial advice, or any other kind of life advice. We do this for entertainment purposes only. Some fun discussion happens, but don’t take anything we say and go to your broker and place bets based on our conversation. If you do, shame on you.
Paul, we have hit Episode 33 of the DataCenter Podcast.
Paul Smith-Goodson: Yeah.
Matt Kimball: You want some fun facts on the number 33?
Paul Smith-Goodson: Yeah, let’s do that.
Matt Kimball: Because we love numbers. All right, 33 is a palindrome both in decimal and binary. Pretty cool, right?
Paul Smith-Goodson: Yeah.
Matt Kimball: Yeah, maybe not. All right, you don’t seem that excited. I thought it was kind of cool. It’s also the atomic number of arsenic. It also represents the number of vertebrae in the human spine. If you got something wrong with you, it might be the number of teeth in your head. You should have 32, but you never know, right? A couple more things. It is the number of deities in the Vedic religion, an old Indian religion, and it’s also the number of the greatest basketball player to ever live, Mr. Larry Bird. This is the Larry Bird edition of the DataCenter podcast.
Paul Smith-Goodson: Yeah, I can agree with that.
Matt Kimball: I knew I liked you. For an Arkansas boy, you get some common sense. Hey, so we got a lot going on today. We’re going to talk about, so Will’s not with us. We kicked him off, we suspended him from the podcast for a week for some inappropriate behavior. Maybe he’ll join us on the next one if we allow him to. Just kidding, he’s in Japan doing some stuff in Japan for NTT. It’s Paul and I today and we’re going to talk a few things. We’re going to talk about Microsoft, Supercomputing, Atom Computing, oh, we have a couple of Microsofts, and then Ignite 2024.
Paul Smith-Goodson: Yeah.
Matt Kimball: I think you’re going to be kicking us off, right?
Paul Smith-Goodson: I guess so.
Matt Kimball: Yeah. What’s this little AI guys thing you’ve been talking about?
Paul Smith-Goodson: I don’t have my notes or anything, but little AI guy, Microsoft came out with a new AI orchestration system. We’ve all heard about agents and that type of thing. Well, these are agents, they’ve got a specialized agents called Magentic-One. I kept thinking it was magnetic one, but Magentic-One, that’s that agent in there and there’s four of them. It’s also got a commander, it’s not called a commander, it’s called an orchestrator that manages these four agents and they’re autonomous. They can plan, they can execute, they can do strategies and demo stuff, and they can do all kinds of things. The way I explained it to my wife is she got used Alexa a lot for facts and everything. Imagine that she would program one of these Alexa agents every time a song came on the TV that was written by Roy Rogers, it would turn it over to the other agent who would turn on the recorder and then notify her that it was being recorded. In that case, she would be the orchestrator and she’d have two agents working for her.
Matt Kimball: Mm-hmm.
Paul Smith-Goodson: In a way. We do have agents already with Alexa.
Matt Kimball: Mm-hmm.
Paul Smith-Goodson: Anyway, the program is open source, so they’re not going to be charging anything. It’s released alongside one called AutoGenBench, I believe, a testing tool to evaluate agents. I’m looking forward to trying it out. Just imagine you have your own team of agents and send them off to do a lot of things. You can go off and have a cup of coffee and watch TV.
Matt Kimball: Do you know if you can build your own agents and add it into the library, if you will, of agents?
Paul Smith-Goodson: I think you can program these agents. I haven’t seen the paper on it yet.
Matt Kimball: Okay. Okay.
Paul Smith-Goodson: Looking forward to paper on it, but I think it’s good news that then it’s released as open source so we can play around with it.
Matt Kimball: This is really Microsoft trying to get ahead of and establish itself with kind of the coming wave of agentic AI, right? I mean, is that the takeaway from this?
Paul Smith-Goodson: Yeah, I mean it’s an important component for leveraging tools to do complex work in the real world.
Matt Kimball: Mm-hmm.
Paul Smith-Goodson: I think the fact that it’s open source is going to help level up whatever’s coming next.
Matt Kimball: Yeah, we’ll get developer using it, getting used to and building their agentic environments around Microsoft, right? Even if they’re just playing around, it’s familiarizing with that and kind of it becomes part of the organization’s tool.
Paul Smith-Goodson: Yeah, it’ll be fun for a while until it gets boring.
Matt Kimball: Yeah. Yeah.
Paul Smith-Goodson: Everyday.
Matt Kimball: As is all technology. Oh, that’s cool stuff. Do you know, like you talk about Microsoft, are there competing, let’s call them orchestration, agent orchestration tools that are kind of out in the market that are also attracting enterprise and also very easy to download?
Paul Smith-Goodson: Actually they’ve got the benchmarks. One is G-A-I-A, I forgot what it stands for. AssistantBench and WebArena.
Matt Kimball: Okay.
Paul Smith-Goodson: There’s benchmarks that you can run these against.
Matt Kimball: NVIDIA doesn’t have one?
Paul Smith-Goodson: I have not seen anything yet from NVIDIA.
Matt Kimball: Interesting. Okay.
Paul Smith-Goodson: There are other agents out there on the market.
Matt Kimball: Yeah. Okay. Excellent. Very cool. This is already downloadable from GitHub or wherever. Okay.
Paul Smith-Goodson: Yeah.
Matt Kimball: Excellent. Very good. Let’s move on. Next one, Supercomputing goes mainstream. This is mine. Supercomputing 24 was this past week in Atlanta, or this week in Atlanta. It’s crazy, you used to go to Supercomputing shows and you almost felt like you were walking through a chapter of a science fiction book, right? There was so much going on in the line from what was happening on the show floor to enterprise IT, it was long and crooked and it went through many kind of layers before what you saw on the show floor actually made it to the enterprise. It was almost more like looking at a bunch of concept cards than anything else.
That’s all changed. I guess we could thank AI, but really we can thank big data analytics, EDA tools and a lot of commercial HPC for helping kind of bring supercomputing to the enterprise and the enterprise to supercomputing. It’s been kind of symbiotic. I went this year. I got back on Wednesday and man, there’s a lot going on. My big takeaways, if you think that liquid cooling was interesting or a passing fad, it’s not, it’s here to stay and it’s only going to grow. There were 50 cooling companies listed in the show exhibitor pamphlet from Supercomputing. That’s 550, not one five. Liquid cooling, I think there were 28, 22, I’m sorry, 22, and that doesn’t even include the Schneiders and the Emersons and the Deltas that are jumping into cooling as well, but they weren’t listed in that.
Lots going on. Crazy. Silicon is hitting a new level. This is another one where silicon used to be mythical and magical to folks and it has changed. We forever have lived in a world where you try and take really complex functions and stuff, if you will, and move it onto discrete silicon to free up the CPU or the CPU to do other things. That has hit a whole new level. Related to that, if you think that NVIDIA is going to be challenged by AMD, Intel, Qualcomm, Marvell, Broadcom, you might be right, but you might equally be right if you walk along the fringes of those Supercomputing shows and see those tiny booths with these silicon startups. There is so much innovation going on in the silicon world. I’m not entirely sure that the big guys are the ones that are going to ultimately challenge the dominance of NVIDIA because of the ability for these smaller companies to innovate so freely and not worry about that kind of innovator’s dilemma, right? Where you build something and you have to support that market that you’ve built.
There’s so much going on. NeuReality, Asiato, there are so many different companies out there that are just doing so much in silicon. It’s going to be fun to watch over the next few years. Tenstorrent is another one, based here in Austin. The last thing is, yep, El Capitan went live. Yep. The Blackwell, Grace Blackwell went live in the 9,712 from Dell. Both really interesting. Probably the most interesting in that to me is the fact that El Capitan is not being driven by discrete GPUs. It’s actually an AMD APU. It’s where you take a CPU and GPU and infuse them onto the same die. I was actually super impressed by that. Did not realize that they did not rely on discrete GPUs for that.
Paul Smith-Goodson: Yeah. The compute density is getting so high, it’s going to get higher. I mean that’s one reason that we’ve got liquid cooling. I know Dell always got some racks with new equipment on it and the GB200s and they’ve got liquid cooling in their racks and models are just going to get bigger and meaner.
Matt Kimball: Bigger and smaller, right?
Paul Smith-Goodson: Bigger and smaller. Yeah, that’s right.
Matt Kimball: Yeah, it’s fun to watch though Paul, I mean you, you’ve been in the technology business for a little bit too, right? I mean it’s crazy to see how quickly all of this innovation is making its way into the enterprise, how short that line and direct that line has become.
Paul Smith-Goodson: It’s accelerating. I mean, look at everything in the AI market. I mean even power, I mean, you know, who would thought that companies would be buying nuclear power plants because the power requirements are so great.
Matt Kimball: Yeah. Literally, you’re building kind of what would outfit a nuclear submarine. You’re buying that and putting it next to your data center. I mean that is just incredible because these requirements, so lots of fun. Anyway, let’s move on because, because this is a lot easier without Will taking all the air out of the room, am I right?
Paul Smith-Goodson: Yeah. Yeah, he needs to stay in Japan for a while.
Matt Kimball: He just might. All right, here we go. Two dozen logical qubits. What are we talking about here? I’m guessing that’s something to do with Atom Computing.
Paul Smith-Goodson: Yeah, Atom Computing and Microsoft are teaming up and doing some really neat things. Logical qubits are made out of physical qubits and in this case that’s four, I think four physical qubits to a logical qubit. And it’s a way of correcting errors when they happen in qubits. That’s one of the big obstacles there is for error correction. It’s preventing us from building really monster quantum computers. There’s been a lot of progress with logical qubits. Microsoft had worked with Quantium on one project and they’re coming along pretty fast.
Matt Kimball: Mm-hmm.
Paul Smith-Goodson: One project a couple months ago, and I don’t remember, I think it was three logical qubits or four logical qubits. They have another project with Atom Computing and it was 12 logical qubits. Now they’ve got 28 logical qubits, Microsoft and Atom Computing. It’s going to be able to go up to like 50 logical qubits. I mean, even now with 50 regular qubits, you can do a little bit of interesting work, but with 50 logical qubits, you should be really able to do a lot of advanced work with a computer and it won’t be up until it gets up to like maybe a hundred or 200 logical qubits, that you’ll be able to do advanced stuff and take on supercomputers. It’s going along really good. Atom Computing has got neural atom processor and they can pack like a thousand qubits into something the size of a grain of salt.
Matt Kimball: Oh my gosh.
Paul Smith-Goodson: There’s not a lot of room to move those atoms around, but they’ve got enough room to move them between. It’s in an array, so they’ve got enough room to move those atoms around. Fidelity is pretty good on this two qubit gate, logical gate, like almost 99.6, which is really good.
Matt Kimball: Yeah.
Paul Smith-Goodson: They’re planning to release a thousand qubit machine for commercial customers next year and it’ll be available either on-premise or it’ll be the Azure Cloud and get it through Microsoft.
Matt Kimball: Wow.
Paul Smith-Goodson: They’re moving right along.
Matt Kimball: When you say commercially available, this is for like companies to start experimenting?
Paul Smith-Goodson: Yeah. Companies can buy to experiment with and thousand qubits, you can do some interesting thing. It shouldn’t be too long before it’s up there a couple of thousand qubits.
Matt Kimball: What’s that power?
Paul Smith-Goodson: That’s physical, not logical.
Matt Kimball: Yeah, yeah. What’s that power footprint look like?
Paul Smith-Goodson: The what?
Matt Kimball: The power footprint on a machine that has a thousand.
Paul Smith-Goodson: I knew it one time what they all were. Now, I don’t remember. It’s a lot less than AI of course.
Matt Kimball: Like a traditional, got it, yeah.
Paul Smith-Goodson: Much less than that.
Matt Kimball: Much more capability. It’s so crazy and it’s interesting when I hear that Microsoft is entering into a space, it tells me that space is reaching some level of maturation in the not too too distant future. I mean, they do a lot of experimenting at Microsoft, but you know, they’re very pragmatic. This is exciting to watch. Super exciting.
Paul Smith-Goodson: It’s exciting the fact of teaming up with Quantium and Atom Computing. Neutral Atom is a good modality. They ran a couple of benchmarks on this 20 logical qubits and they use an algorithm called Bernstein-Vazirani, and what that is they’ve got like a hidden 20 digits and a string of bits and the machine has to find them. A classical computer takes, you have to run it each time to find one bit at a time. With a quantum computer, you can find them all with one shot.
Matt Kimball: Wow.
Paul Smith-Goodson: You can find it with one running it one time.
Matt Kimball: Do you think the market will start waking up to quantum more and talking about it more? I mean, I can count on just a couple of hands the number of companies that actually talk about quantum, and probably one hand the number of analysts that talk about it, and maybe one finger being you, analysts that talk about it and talk about it with kind of an understanding that’s kind of real and based in the real world. Do you think this starts getting more coverage soon?
Paul Smith-Goodson: Yeah, I mean there’s value in just quantum computer computing by itself. I think the real value is going to come when you match up HPC and AI and quantum computing.
Matt Kimball: Yeah.
Paul Smith-Goodson: Microsoft is going to be doing that.
Matt Kimball: Yeah.
Paul Smith-Goodson: I mean they ran, I’ve got an article on the show notes, but a while back they used AI and HPC and what they did was they created 32 million new materials and they went through and sorted through it and they found one that they could use as a potential ion battery substitute. Going from 32 million to one is so impressive.
Matt Kimball: Yeah. Yeah, that is, wow, like that. Yeah. We’re hitting a whole new kind of, it’s almost unfair to call it a new wave of computing.
Paul Smith-Goodson: Yeah.
Matt Kimball: It feels like a brave new world entirely that we’re about to enter into.
Paul Smith-Goodson: Yeah, yeah. Quantum is moving along. I mean IonQ is doing good, getting the market. Quantium is doing really good stuff, technical stuff. Then now Atom is coming to play. The next big one is quantum sensors are going to take another bump. I’m waiting for that. That’s the field I’m really interested in, quantum sensors, Quantum sensor can replace an antenna and a receiver.
Matt Kimball: Oh.
Paul Smith-Goodson: They can receive from basically zero to terahertz or gigahertz.
Matt Kimball: Oh wow.
Paul Smith-Goodson: I know your background. That’s going to be interesting.
Matt Kimball: Yeah, my head is kind of spinning around that. Now I’m going to be spending the afternoon kind of looking through this.
Paul Smith-Goodson: Yeah. We’ve got a little bit of time here. They create what they call Redbird atoms. Redbird, you take the outermost atom, electron, you put a little energy into the atom and you force that outer electron a far distance from its nucleus so it’s more sensitive to electromagnetic fields.
Matt Kimball: Yep. Wow. I mean there’s so many uses for that and so many applications for that.
Paul Smith-Goodson: Quantum sensors are already a commercial item, not that particular application.
Matt Kimball: Thank you for giving me something to spend my afternoon, on quantum sensors.
Paul Smith-Goodson: Yeah. There’s an article in Forbes, I wrote to them a couple of years ago when the Air Force labs were experimenting with quantum sensors and then there’s a company called Redbird Technologies and there’s an article in Forbes on Redbird Technologies that talk about that. It’s also got a picture of my loop antennas on that.
Matt Kimball: We’ll throw that into the show notes so people can go reference it as well. Yeah, that’d be great. Hey, let’s close out with one more thing. Chips, chips, and more chips at Ignite, Microsoft Ignite. Staying on the Microsoft theme, it seems like they’re getting a lot of love. They had their Ignite conference this week and they had a whole bunch of announcements across the stack from Copilot stuff to traditional software to stuff they’re doing with Oracle, Oracle database to chips. I’m going to focus on the chips for a second and then touch on Oracle real quickly. So, we all know that Microsoft launched, I think it was at Ignite last year, they announced Cobalt, the CPU, and Maya, the accelerator, AI accelerator for training and for inference. Those have been kind of, you know, chugging along, getting broader adoption. Cobalt is out there, Maya is out there. This year they announced three new chips.
First they announced an HSM, which is a hardware security module, and it offloads. Talking about offloading discrete functions, it offloads a lot of the crypto and digital signature kind of work from the CPU onto or from a key manager onto a specific chip. Kind of like bring your own security to the cloud function and functions within Azure. It offloads all of that making for a better experience for customers. Very cool. It complements a lot of what they’ve already done with their, they have a silicon-based root of trust and kind of a full security stack. This kind of brings it all together, if you will. That’s part one. Part two, they introduced the Boost DPU, data processing unit. For those that are in the know, it is exactly what you would think. Network functions, IO functions, storage management, all offloaded onto discrete silicon to make that storage environment run faster. Data-centric applications are really going to benefit from this.
The third piece of news coming out of that is a partnership with AMD whereby AMD developed a custom piece of silicon, a custom epic CPU specifically for Azure. It’s called the 9V64H. It’s actually an EPYC 4, so it’s an N minus one core running at four gigahertz with HBM, high-bandwidth memory, attached to it, and InfiniBand, NVIDIA’s InfiniBand coming out of it for faster connectivity to NVIDIA GPUs. This is designed specifically for HPC, virtualized HPC. It’s an Azure only part. This thing is going to scream. It’s a pretty cool part.
Here’s what is maybe even more interesting than that. When you look at custom silicon for the CSP’s, AMD has been doing custom silicon for gaming boxes and all that for some time. But when it came time to do custom silicon for CSP’s, typically that was the domain of Intel. This to me signals that AMD really does have the dominant position in the cloud. I mean not only from a market share position, I think they’re about 60% market share in the cloud, but more importantly, just as being viewed as the innovator of the group, if you will, relative to other silicon providers out there. Really cool stuff. They did a lot in silicon. They really are building that end-to-end compute environment for traditional compute, for AI, for HPC, for all of those workloads that you deploy. Microsoft is building out custom silicon and their own silicon from end to end. Really cool story following in the paths of AWS and GCP.
Last thing I want to talk about, I’ll go quickly on this, is they also, Microsoft also made some announcements with Oracle. Oracle Database at Microsoft Azure, three things came out of that. The first is, well, four things actually. They’ve expanded the number of regions that it’s offered in. This is where Oracle Cloud infrastructure, like the hardware, the exit data hardware is being installed within Azure data centers and connected to the Azure backbone. This is native integration into Azure and it’s running obviously Oracle’s database service. They’ve expanded, they’ve added 12 regions, they’re going to add a whole bunch more in 2025. That’s part one. Part two, they have made this fully manageable through Azure Resource Manager. Terraform, all of those forms you use to kind of configure your infrastructure. Oracle Database appears as a native part as your Azure infrastructure to go ahead and manage and provision. Very cool. Three, it now resides on Azure Fabric, which is the underlying high-speed interconnect for Azure. Very cool.
Four, Oracle database environments or Oracle Database at Azure is now natively integrated with Purview, which is a, when you think about like regulated industries, insurance, financial, legal, government, so on and so forth, there’s a lot of auditing and regulatory compliance that’s required. Purview is a tool that allows you to look across your data environment and look for where you may have vulnerabilities. Either you’re not masking a social security number appropriately, or you’re not encrypting to the degree that you should. They have all of these kind of audit trails that are built in that allow you to look at what your compliance looks like based on the type of company you are. There are third-party companies out there like CyberSaint that do it for like the really large scale and go really deep, but Purview’s been a very cool tool for Azure and now Oracle is part of that environment as well. I just rambled on for four minutes without taking a breath, so I’m going to stop and let you add any of your 2 cents you want to add in Paul.
Paul Smith-Goodson: No, I’m just curious, is that a trend for hyperscale cloud providers to develop their own silicon?
Matt Kimball: Yeah, it’s funny because some of this stuff, this stuff, some of the silicon, like the security module, that’s born out of the open compute platform standards and setting up standards for creating silicon to perform security functions. It’s built to that spec. Google’s been doing custom silicon since 2012, I think it is, or maybe it’s ’24, I think 2012. AWS has been doing it forever. You know, everything from Nitro to Graviton to Trainium and Inferentia. Microsoft’s been doing custom silicon for a long time, but as it applies to the cloud, they’ve really accelerated their embracing of custom silicon. You know this as well as I do, right? Every data center environment is unique. Every single one’s a snowflake. It sounds corny, but it really is. The requirements are super unique from CSP to CSP to CSP.
Paul Smith-Goodson: Mm-hmm.
Matt Kimball: Often the littlest differences you can make in silicon. It might be from a power perspective, it could be from a functionality perspective. When you look at that at scale, it delivers a lot of value back to the company. This is why CSPs have jumped into that custom silicon space to begin with. This is something that, it is a trend, it’s a trend that’s been going on for a long time. Frankly, I expect more and more of this out of these companies, out of these CSPs as they move forward.
Paul Smith-Goodson: Mm-hmm.
Matt Kimball: It doesn’t mean they’re moving away from AMD, Intel, Qualcomm, Broadcom, Marvell. It doesn’t mean that at all. It just means where they can’t get their needs fully met, even through a custom chip, like what EPYC just delivered, then they have the ability to do that in-house and really truly, you save two watts on one server. I mean when you look at that across hundreds of thousands of servers, it’s a big deal.
Paul Smith-Goodson: Yeah. Yeah.
Matt Kimball: Yeah. Anyway, that’s all I got on Ignite. There’s a lot more good news and you should check out the Moor Insights site. Our research notes will cover all of the stuff that’s going on. I know that Jason Anderson’s going to be doing some stuff on AI and Will will be doing some security stuff. There’s a lot of good stuff to read up on around that. With that said, and we’re cruising along man, we’re doing good. Want to throw one more thing out that kind of getting to know us section or a kind of topic that we hit on every week, keep it a little light this week and just kind of Thanksgiving’s coming up next week and kind of curious as to what you’re going to be doing for Thanksgiving, Paul.
Paul Smith-Goodson: We are going to have a family get-together.
Matt Kimball: Right on. Big get-together?
Paul Smith-Goodson: Yeah, probably 30 people maybe.
Matt Kimball: Oh wow. Okay. Do you do any of the cooking?
Paul Smith-Goodson: Not me. Firearms out and shoot up the place too.
Matt Kimball: You know your place. I love it. I love it. We have my two boys, Jasper and Ryan are coming in next week from school and my sister-in-law, Pam is here. It’ll be a small five of us getting together, enjoying time together, and making the most of a few days together. It’ll be a nice time, nice time to take a break and we’ll reflect on all the good things that have been going on in our lives.
Paul Smith-Goodson: Yeah, we’re having all three of the daughters and their kids and my wife and I, so it’ll be fun.
Matt Kimball: Fantastic man. I hope you get a chance to truly unwind and decompress for a few days before the last spurt, before end of the year, and you get to enjoy your time.
Paul Smith-Goodson: Trim up a tree and you feel better.
Matt Kimball: I’m going to have to try that.
Paul Smith-Goodson: Yeah.
Matt Kimball: Hey, listen, before we go, I want to let you all know one more time, what we talked about is entertainment purposes only. Might hear a good tidbit here and there, usually from Paul, not from me, do not take what we say as financial advice. Please, please, please. Also, if you like us or please hit the like, subscribe button and be sure to tune in a couple of weeks where we’ll do this all over again. We’ll talk about what we ate for Thanksgiving and all the cool stuff that we see out at AWS re:Invent and elsewhere. Until then, y’all have a wonderful week and a wonderful Thanksgiving.


