Co-Founder, Chief Executive Officer, President & Director, NVIDIA
Supermicro CEO Keynote 2025 with Japanese Subtitles | ft. NVIDIA CEO Jensen Huang | During COMPUTEX
Watch the full CEO Keynote from Supermicro Innovate! 2025 held during COMPUTEX week, where Charles Liang, Founder, President & CEO of Supermicro will be joined by Jensen Huang, Founder & CEO of @NVIDIA.
Together, they will unveil Supermicro’s revolutionary Data Center Building Block Solutions® (DCBBS) capable of:
Fully outfitting liquid-cooled AI data centers and factories with all critical computing and cooling infrastructure components
Offering unparalleled design flexibility at the system-level, rack-level, and data center-level
Fastest time-to-market and time-to-online advantage, deplo...
About Jensen Huang
Jensen Huang, CEO of Nvidia, has been active in public appearances and keynote speeches in mid-2026, focusing on the rollout of new hardware and his views on the societal and economic impact of artificial intelligence. At Computex 2026 in Taipei, Huang announced the Vera Rubin architecture for data centers, which he described as being in full production and designed to run AI agents, as well as the RTX Spark superchip for Windows PCs, developed in partnership with Microsoft. He characterized the RTX Spark as a reinvention of the personal computer for the era of local AI agents. Huang also discussed Nvidia's partnership with Marvell on MVLink Fusion technology and highlighted Taiwan's role as "the epicenter of the world's technology ecosystem" and a strategic partner for the United States in diversifying supply chains. In interviews, Huang addressed the broader implications of AI. He stated that AI's impact will be "largely wonderful" but acknowledged areas requiring caution, comparing the need for new social norms to the advent of automobiles. He argued that AI is creating more software engineering jobs, citing a near-tripling of GitHub commits in early 2026, and described the notion of AI reducing jobs as "complete nonsense." Huang also called for the United States to become "pro-energy growth" again, asserting the country is "woefully behind in energy production" and that energy is foundational to AI infrastructure. He described the current period as the largest infrastructure buildout in history, with a trillion dollars being invested annually across energy, chips, data centers, and AI models.
Transcript (57 segments)
✨ AI-enhanced transcript with speaker attribution
N
Narrator0:04
History rewards the innovator, the pioneer, the trailblazer. For over 30 years, Super Micro has been fueling innovators and visionaries with the industry's best computing resources. We push technology forward so our customers can be first. We have built the industry's widest portfolio of servers deployed in the world's largest AI data centers with exactly the best solutions. So whether it's the first of its kind, the fastest ever built, the newest innovation, the never seen or done before, Super Micro is there so you can be first and the best. Ladies and gentlemen, I'd like to introduce Super Micro's chairman, president, and CEO, Charles Leang.
C
Charles Leang1:05
Thank you. Thank you. Welcome everyone to our innovator 2025. I'm very excited, very happy to see you all, especially once a year in computer. As the AI boom continues to grow, I'm super excited to share our experience, innovations, and solutions with you. Most importantly, I invite you to grow alongside us at our speed, at our pace, or even faster than us. Over the past three years, our revenue has grown about five times. Looking ahead, I expect that in the next three years, we should be able to continue this strong growth. The strong growth doesn't rely too much on the AI boom continuing at the pace of the last three years or on strong macroeconomic conditions, because of our strong foundation, truly optimized strategies, and together with you, the ability to offer a better total solution than others, and thus expand our market share and mark 10. That's why I invite you to grow with us or at least fully leverage our optimized product solutions.
For the past 32 years, we have been pioneers, consistently delivering many new technologies to the market well ahead of others, particularly in CPU and GPU AI system and bra solutions. Back in the early 1990s, we introduced the industry's first dual-processor solution based on the Intel Pentium Processor, paving the way for our journey to enterprise computing. In the early 2000s, with our system building block solution, we created the first dual-processor twin server system, which became the most popular system architecture of today's large data centers. A few years later, our ship brought unparalleled computing density, flexibility, and power efficiency to data centers. About the same time, we committed to green computing technology earlier than anyone else in the industry, emphasizing energy-efficient designs that reduce total cost to our environment and make our earth greener, more beautiful, and save money for customers without compromising performance.
In 2012, with Mr. Jensen's personal advice, we were the first company to design the industry's first GPU server using Nvidia Tesla GPU, which many people call the world's first GPU supercomputer. It was not easy to squeeze two GPUs into one new system and air-cool it at that time, and make it available in just a few weeks. Time to market has always been our advantage. This system set new benchmarks for high-performance computing that eventually became the foundation of today's artificial intelligence system.
Fast forward to 2023, we were again the first to deploy high-volume H100 HGX solutions that dramatically propelled today's AI boom. Many of our customers have since become today's most successful and valuable AI leaders. Last year, we once again led the industry by being the first company to design and deliver a high-volume direct liquid cooling (DLC) solution to the market. We deployed over 4,000 DLC H100 and H200 AI racks, including as you may know X AI's 1500 racks within 122 days, breaking data center history by building such a huge data center in such a short time. Our DLC solution not only enables up to 30% power savings for customers but also significantly reduces total cost of ownership (TCO) while providing a more reliable and efficient alternative to traditional air-cooled systems.
Now in 2025, what's new and what's the big thing that can change our business? To grow our business even faster and provide much more value to our customers and industries, as an engineering company with green computing DNA, today I'm excited to introduce our latest DLC2, a revolutionary data center building block solution, or we call it DCBPS, to you, our value partners. We know DLC is a great solution, especially for GPU racks and AI data centers. It has become a mature and very reliable solution, at least from Super Micro. I'm not quite sure about other companies, you can try, but at least from Super Micro, it's been very reliable, we prove it. Today we aim to make it even better.
Let me introduce our new DLC2. It will deliver up to 40% savings in most power and water consumption. People say it's water cooling, yes, water cooling, but we can save most customers up to 40% on water consumption. It can operate efficiently with natural room temperature water up to 45°C. That means water or chiller compressor is just an option; you can remove it, it's not a must-have device anymore. The DLC2 can capture system thermal energy up to 98% or even higher and does not need any faster fan speed. In auto configuration, you can even totally remove all fans, so no fan or very low-speed fan. Thus it can reduce data center noise level to about 50 dB. Let's think about how quiet 50 dB is; it's as quiet as our home here or as quiet as most libraries. Can you imagine a data center can be as quiet as this environment? It's happening.
Additionally, it lowers both data center capital and operating expenses, both lower by up to 20%. Save up to 20% of money. So DLC liquid cooling is cheaper; don't feel DLC cooling will charge you more, no, less money, while offering even greater reliability than traditional air-cooled data centers or common industry DLC solutions. Starting today, our DLC2 solution also comes with another big bonus. What kind of big bonus? Can our PM show up to show what kind of extra bonus?
Hi Charles. So, watch this. Hi Charles, this is our latest and greatest cooling innovation just from our lab. Our team of scientists has been developing this PG DLC2 coolant for the past year, and we are bringing the best out of our DLC solution with this new cooling solution. What's the difference? So traditional PG25 works, but we have developed for the past year this optimized formulation with ultra-high electrical impedance. So we are so proud today to announce this new Super Micro PG DLC2. So when can we ship? Well, Charles, most importantly, you know shipment is very important here. In fact, that's right, with PG DLC2 we are continuing our commitment to on-time, at-scale rack deliveries by Kangabai. Oh, days or hours? We're going to go quick. I'm telling you.
So, she delivered 4,000 DLC racks last year, and this year now we are going to deliver DLC2 by day. Days. Absolutely. Charles, you forgot hour. Wait, we're going to go hours if needed. Yes. Yes. Thank you. Thank you guys. Thank you. Great. All right guys, thank you. Thank you. Thank you. Good. Our people work very hard. Thank you a lot. So our DLC2 now is much better than last year's DLC, saves more power, is safer, and has lots of other advantages. Anyway, after 10 months of hard work, research, study, and real-world validation again and again, today I'm very excited to announce that we have created a brand new cooling that can make your CPU and GPU system much more protected, more reliable, safer, and save money.
So good, huh? In the past 30 years, people keep using PG25 as data center common cooling. Today, with our new DLC2 solution, we will ship with PG DLC2, a new coolant for your data center. The new coolant has 1,000 times higher electrical impedance while with better thermal conductivity. That's the way to be more reliable, safer, and save money. This is Super Micro. We always look and search for the best solution for you. We are a delivery company. Okay. Now let's focus on something really big and bigger than DLC2. The big thing for today and for this year, maybe for the next many years to come, a big change to change how you grow your business easier, grow your business faster, and be more profitable.
32 years of data center experience with our green computing DNA, with our building block solution practice. Today, a new way to build a new data center, cloud, and AI infrastructure is born. A new way is available now. Our DCBPS. With our DCBPS, we make your job to design a data center, to build an AI data center as easy as building a Lego castle. You know kids can build a castle with Lego easily, beautifully, quickly. To build a data center as quickly as in a few months, not as long as many quarters or many years. Because of our easy design and easy build DCBPS solution, your job will become easier, quicker, and your data center and cloud performance become higher, better quality, faster time to online, and lower in cap expense, especially lower in operating cost, I mean in TCO.
With our DCBPS total solution, we deliver a force-to-market AI platform, whether at system level. You can buy our system, server system, AI system, storage, or rack-level or rack-scale plug-and-play solution. We ensure seamless integration by pairing you with the best match storage solution. We offer more than 10 storage platforms including E1S, E3S, traditional storage, more than 10 platforms for you to select from, and a full set of all other components. This includes a variety of switches. We offer more than 10 different kinds of switches: high-data switches, management switches, or InfiniBand switches. CDU is an in-row CDU or end-row CDU at the main DLC system or DLC2 water tower. Water tower? Which one is water tower? The smaller one is water tower. What's the bigger one? The bigger one doesn't need water at all, cold dry tower. So we have both. You want a water tower, we have for you. Many different skills for you. You want a tower without water, we have a dry tower for you very soon in the next few weeks.
Power shelf. Power shelf. Chiller door. You build a data center or you transfer a warehouse to a data center without air conditioning, you just use a chiller. And we offer you a variety of chillers in the next few months. Power shelf again, we provide you a variety of DC power shelves. Battery backup unit (BBU). Our battery backup comes with many different skills, many different aspects, and comprehensive networking design. You want to build your data center, we can work with you to design networking together, depending on your scale, depending on your application. Our DCBPS is to provide you the option that you like the most.
The goal... you miss something... and a variety of service packages, right? Although I just landed in that yesterday early morning and met almost every hour, but I'm not this year. So anyway, onsite deployment, right? Customers just buy our devices and deploy by themselves, welcome. Or you want Super Micro to do that deployment for you, we offer that. And cabling, right? You know a data center, especially a big data center with 100,000 GPUs, for example, the cabling system is very complicated, very sophisticated. We offer that service as well and end-to-end management software. We have more than 1,000 solution engineers now, I believe in USA, Taiwan, and Europe. So our networking engineer, data center management team, very happy to work with you shoulder to shoulder to build a whole data center, design the whole data center, deploy the whole data center, and offer different kinds of services, a broad range of services. You know, whatever selection you want, right? That's brainstorming work together for your most efficient and profitable business. Or if you prefer, we can just simply support you with whatever subsystem we have and leave the design, deployment, and service for yourself. So we are very flexible. Whatever the best you like, we are happy to work with you. Our data center building block solution is to provide you the option that you like the most. You are the boss. You pay, you are the boss. You decide what kind of service you want us to service you.
The goal of the data center building block solution (DCBPS) is to save your time for time to deployment, time to online, and improve data center reliability while reducing the cost for you. We know new technology time to market, time to deployment, time to online is critical to you. We know minimizing power usage is important to you. We know your data center quality and reliability are very critical to your business. With our DCBPS solution, we aim to shorten the time to delivery (TTD) and time to online (TTO) for new data centers from years to just quarters. For customers upgrading existing facilities or data centers, DCBPS can accelerate your time to online from quarters to months or even just a few weeks because everything has been well designed. We manufacture them and keep safety inventory. When you need to extend your data center, just a phone call, we ship the total solution to you and deploy with you to speed up your time to online.
With our data center building block solution, especially with DLC2, we help you and your customer achieve up to again 40% savings in power and water while enhancing the system and data center availability and performance. Again, additionally, the DLC2 further reduces data center noise as low as 50 dB, ensuring a quieter and more efficient environment. And then better features, better advantages. Yes, we aim to save extra cost 10%, 20%, or up to 30% cost as a big bonus to you.
Our DCBPS solutions are fully optimized and ready for high-volume deployment about now. So first, composer service now. So with global production capacity of 5,000 racks per month now, with our new Malaysia campus fully operational for large-scale production today, alongside continual capacity growth in the US, Europe, and Taiwan. Super Micro is well established not just as the best brand name data center total solution company, we are also one of the highest volume ODM partners for global data centers. With our DCBPS, now we have set a stage to welcome our good friend and AI genius, Mr. Jensen.
Is it Jensen here? That's it. Not yet. Traffic. Jensen, anybody waiting for you? I think that picture is too big. Take it down. Yes, crazy man. Taj, you predict AI boom will continue at a super fast speed. And we saw last year the AI market grow, and how about next year and the next many years to come?
J
Jensen Huang28:04
You know if you look at the last 10 years, AI evolved. Where are you going? Same thing. Oh, okay. I need a water. I just started talking. He walked away.
C
Charles Leang28:27
This makes me seasick. This is all your technology. Okay. Yes. We sell it by the tons now. Yeah. This indeed is the DLC solution. We shipped last year. Yeah. 4,000 racks last year. Yeah. And this one, can you shift a little bit? This is a new one. Oh, no. Oh, here. Here a new one. Our new DLC solution we call DLC2. Uh-huh. Save even more power. Uh-huh. DLC2. That's clever.
J
Jensen Huang29:04
So, AI listen. So, 12 years ago when AI first came into the scene, it was really focused on perception, right? Computer vision, speech recognition. And then five years ago, everybody started to learn about generative AI. AI can now not only understand but it can generate content. And so text-to-text is summarization. Maybe text generation, maybe story generation. Image-to-text is computer vision. Text-to-image is image generation. So there's a whole lot of different ways you now have AI that can take any modality, any content of any modality input, and it can generate content of any modality output. It does the translation. So that's the second phase. Yeah. But the real breakthrough now is reasoning AI, and most people should understand this. The fact of the matter is generative AI learned a great deal of knowledge by recognizing patterns and relationships in pre-training, and then you ask it a question. Generative AI does a one-shot generate text. You think you're getting an answer. Okay. And so, but most of that knowledge was compressed into the large language model. Yeah. As you know, intelligence is about thinking. Intelligence is about reasoning. And by reasoning you can solve problems that you never learned about. So just by understanding some basic rules and theorems and maybe some prior knowledge, you could decompose a new problem step by step and solve a problem you've never seen before. Right? Reasoning that's in problem solving.
Yeah. Now there are many different ways to do reasoning. You could break a problem down step by step, or you could simulate the future and choose the best outcome. Yeah, you simulate the future, you evaluate its consequence, you simulate the future, you weigh the trade-offs, you know, which one is better and which one is less good, and then you decide that to be the action plan and then you execute the plan. Yeah. Maybe when you execute the plan, you need to read some reports. Maybe you need to use a calculator. Maybe you need to use a charting system, you know, like a drawing system, whatever it is. Maybe it's even using SAP or Oracle, use tools, right? And so now this is the era of Agentic AI. It's essentially a digital robot inside our system. Yeah. And it can go off and solve problems.
Now, why is Agentic AI such a big deal? Agentic AI is a big deal because today most of everything that we do is codified into business processes. Business processes we figured out, we reasoned about what is the best way to solve a problem, and we are very prescriptive. We specify very clearly input and we specify very clearly the output, and we take step by step and we implement a workflow. Okay. Well, the problem is most of the interesting problems in the world require unstructured data. Most of the business processes are implemented with structured data. That's why we use SQL. That's why we use Oracle. That's why so on and so forth, relational databases. Most of the interesting problems are unstructured data, meaning we have to read something from the internet. Maybe we read a PDF. Maybe we have to take information from multiple sources and we have to decide how to solve the problem. Yeah. Well, now based on everything I've said, Agentic AI has the opportunity to understand unstructured data, images, text, videos, combinations of all of that. Read it, understand it, reason about how best to solve a problem given a goal. Maybe even look to another agent to help. And so for the very first time we now have the opportunity to use AI to solve a whole bunch of problems we didn't have. Digital customer service, digital researcher, software programmer, supply chain manager, marketing campaign manager, all of these types of problems are very difficult to put into a business process. Yeah, it requires intelligence. Now we have the software to do that.
Now the challenge, of course, is the computational load is very high for thinking. If you ask a question, somebody says you know what's 8 times 8? Very simple, right? Yeah, everybody knows the answer. 64, right? It's one shot, you effectively memorized it. But if you have to go and reason, you have to think for a long time and maybe you're analyzing a lot of different paths and then finally you have the answer. So that's the reason why when you use deep research on ChatGPT, it thinks for a long time, goes on and on, or you use Perplexity and you do a deep research search. I use it every day and it breaks it down step by step. I'm going to this website, I'm reading that, I'm thinking about it some more, I'm reading that, and it eventually comes back and self-analyzes: did I give a good answer? Maybe it's not a very good answer, try again. And so this kind of AI requires a lot of computation.
So now you have these two circumstances coming together. One, you have an application, you have a capability that's very useful. Yeah. And two, so hopefully the volume is high. Two, the computation demand is very high. Therefore, the volume is high. The combination of these two things, my sense is that for the next decade, we're going to see a lot of demand. Wow. Yeah. Great. So in conclusion, the good news is that deep learning also needs computing power, but the AI application needs even more computing power, way more. I'm hoping the goal of AI is not to train the model; the goal of AI is to use the intelligence, right? So training the model hopefully is 0.5% of the total consumption. Yeah, inference time scaling is 1,000 times more load. So they are double, triple, quadruple, 10 times growth opportunity in the next few years. The IT industry is $1 trillion large. The AI industry is above the IT industry. It's likely trillions of dollars large. Yeah, much bigger. So lots of chances to make money.
C
Charles Leang35:55
Great. Appreciate the AI wave. Can you come with me? GB2, I mean GB200, GB300 again. Your smart engineers suggest sharing the same solution, the same rack, right? Good idea. Yeah, so save lots of resources. And this is our last year's NVL B200. We start to ship stuff from this March, and then we have DLC one, DLC traditional, and then we have DLC2 here. DLC classic. Yeah. And this is DLC2. New DLC available. That's beautiful. Yeah. We'll be ready to ship this will be ready for B200 next week. Next week. And ready for B300. Yeah. Next few months.
J
Jensen Huang36:56
Going from DLC. What do you call your GB200? What's the system code name? Product name? Uh, that's DLC. Oh, which one is HGX? HGX. Uh, DLC. HGX. That's DLC.GX. And what's MGX? MGX is over there. And what is the name of that? Over there. Over there. Okay. See, not technology. Okay. So, this is called over there. Over there, ladies and gentlemen. Over there, generation one. Yeah.
C
Charles Leang37:34
This B200, BLC2. Uh-huh. Okay. Yeah. 40% power saving, water saving, and the system can run at a 50 dB noise level. Well, nobody's been working on liquid cooling data centers and servers longer than you have. Yes. Yeah. Yes. Thank you for your support. Yeah. Yeah. Last year we shipped about 4,000. Charles is 85% computer scientist, 15% plumber. Yes. Yes. Maybe more percentage water-related technology increasingly. Yeah. Yeah. And this is the air cooler. You know that, right? B2, B300.
J
Jensen Huang38:23
And Charles, I'm really glad that you're gearing up all these different options and that you're gearing up for large capacity. And the reason for that is this. This is the transition of our industry from building servers for data centers. Yeah. Now building systems for AI factories. Yeah. And these AI factories are intended to generate tokens to make money. Yeah. That's the goal. That's why I call it a factory. Yeah. A data center you hold data, you know when you hold data it doesn't make any money. So AI factory, yeah. AI factory. And so the world still needs data centers. We need data centers to host our storage and so on and so forth for our employees, right? Our ERP system, run our company. But we also need to have an AI factory to generate intelligence for agents and for robots and self-driving cars and you know all these different use cases.
C
Charles Leang39:21
Yeah, we have both. Yeah. Yeah. We service your traditional data center customer also. Factory. Yes, that's right. And this is your MGX. I guess this one MGX, too many hard to remember.
J
Jensen Huang39:39
Yeah, this new system is really terrific. This is the RTX Pro. Yeah. The difference between this architecture and the Blackwell architectures for AI systems for AI factories. This is really designed for AI focus. Yeah. For training and for inference. But this has the benefit of being able to do computer graphics ray tracing as well as AI. Yeah. And the reason why that's so important is because of digital twins. Yeah. Running Omniverse. Inside Omniverse, you need sensor simulation or ray tracing. You also need AI to do generative content. You also need AI to run the robotic system inside the digital twin. And so this one machine is designed for that future, the digital twin future. Robotic systems, factory digital twins. This is going to be so successful.
C
Charles Leang40:34
Yeah, we already shipped for a few years and this is a new model. Yeah, new model is going to be great. And don't forget to put CX8 in there. CX8. Yeah, I see it. Oh, waiting for your CX8. You just showed me. Okay, I'll get it to you as fast as I can. Your CS team needs to ship as quick.
J
Jensen Huang40:53
All right, guys. Get to it. I wish I didn't bring it up.
C
Charles Leang41:01
No, you're right. CX8 is in such hot demand. Yeah, CX8 is a revolutionary chip and the reason for that is it does two things, of course. First, it's a very smart NIC. Yeah. And it has nearly all of the benefits of InfiniBand, but it's Ethernet. So, it's Ethernet super super. Yeah. You know, we need 100,000 pieces ASAP. Okay. All right. Did you place a PO yet?
J
Jensen Huang41:31
Hey, press PO right away.
C
Charles Leang41:35
Okay. So, let me finish my story. And the second thing, the second thing is usually the NIC sits at the end of the PCI Express bridge. Yeah. We now turned it around. So, CX8 is also a PCI Express bridge. Yes. So, it's a switch first. Yes. It's a switch first and a NIC second. We have that design already. It's so good, right? Yeah. So good. I know. So smart. Yeah. Doesn't need your switch in your chip. Okay.
J
Jensen Huang42:03
And this is your storage solution based on Grace. Yeah. One terabyte, one petabyte. Okay. Well, in the future even storage is going to be smart storage, right? AI storage. Yeah. Yeah. You could use SQL query for structured data. Yeah. And then you could use semantic query, AI query for unstructured data. Yeah. And so instead of usually in SQL queries you tell it exactly what you're looking for. Maybe you'll do some joins and some combines and some processing, but eventually you are very specific about what you want to get back out of a SQL query. In the case of a semantic query, you just have a question. You just ask it a question and the storage will tell you what the answer is. It's so strange. And so in the future you don't have to retrieve files. Uh-huh. The purpose of storage is not to retrieve the file. The purpose of storage is I just want the answer. Storage. That's right. You just ask the storage the question. So easy, right? Yeah. Just ask the storage. And so imagine your enterprise database, right? Yeah. And it's got unstructured data, all these PDFs and PowerPoint slides and Word documents and videos and things like that. In the future, you just talk to it. Yeah, you ask it whatever question you want and maybe the customer service all night long they're responding to the customers' requests and now they have a bunch of data, customer requests, and maybe it's voice, maybe it's somebody typing it in, maybe it's right, whatever complaints are, and then you just ask what happened last night. Yeah. And the AI will tell you what happened.
C
Charles Leang43:47
Customers are super excited about the new DLC generation 2. Yeah. And everywhere over there, generation one. They're super excited about that. And the POs keep flying in and they're waiting for their delivery. That's what happened last night. I am certain that's what happened to your customer service last night. Yeah. Where is my supercomputer? Where's my... Yeah. Place order is all here. I'm ready to receive orders. Yeah. So again, like J said before, storage usually interfaces with CPU and then CPU handles the data to GPU. But now the new fashion, data storage connects to GPU directly to speed up, save money, and speed up. Yep. That's right. And get smart. Much faster. And get smart. Right. Good.
So thank you so much for your strong support, very rich. So from Jensen says next few years, the growth can be more than... is it thank you, the cue for me to leave? Oh, no, not yet. Okay, not yet. Unless you... no, I got all day. I was just trying to figure out when I was supposed to get off stage. You people told me usually people don't say 'Jensen, thank you for coming' until the end. Thank you. Is it me? I don't think it's me.
J
Jensen Huang45:28
No. What? Go ahead. It's your show.
C
Charles Leang45:33
We didn't talk about any of those things. Jensen is much more generous than his people. His people call me Charles only 10 minutes means. Okay. Our Jensen is very tired. He needed some noise. I said, 'Okay, 20 minutes.' As long as you wish.
J
Jensen Huang45:49
Thank you. If we can sell more supercomputers, I'll...
C
Charles Leang45:53
Great. So anyway, I mean from B21A, B3A, GP21A, GP300, all the products are ready to ship, especially B3 air cool. Yeah. Will be ready in the next 30 days. And then I dwell, I'm still waiting for your engineering support. So hopefully that comes very soon too. But this is GB3 coming very soon. He invites me to a keynote to lodge all these grievances on me. I'm late on engineering support. I'm late on CX8.
J
Jensen Huang46:37
It's easier to ask something. Of course. No problem. Absolutely. As fast as possible. Well, hey Charles, listen. You know, I've loved working with you and I've loved working with you for a very long time. Super Micro is exceptional at the most extreme engineering. The craft, the engineering craft is second to none. And your execution speed is incredible.
C
Charles Leang46:58
Oh yeah, Jason, maybe you come a little bit late, right? So I just introduced. Did I say the wrong things? I just introduced. Did I say the wrong things?
J
Jensen Huang47:06
No, everything is perfect. Usually he's supposed to... I think that... Let me show you how conversations go. I say that and then you say 'thank you' and 'Jensen, you're incredible at this and that.'
C
Charles Leang47:21
You know, do you know what? Is it me? I just want to share a really good news with you. Okay. The big thing I just shared with our people here. I gave him a compliment. And he's supposed to give me a compliment like 'Jensen, your new leather jacket is so great.'
J
Jensen Huang47:39
I just announced to the wall. Super Micro just officially announced our data center building block solution. With that, we aim to build all the key components for the data center for our AI factory, especially including the cooling, including water tower, dry tower, water tower, chiller, everything, indeed including transformer, including the source. Yeah, that's a great idea because these days, you know, one little piece of something late, the whole project, and the project, you know, a lot of capital is sitting waiting. Tens of billions of dollars of capital is waiting for a water tower or power generator or CDU or... Yeah. You know. All kinds of power racks. Yeah. And a power shelf and BBU, battery backup unit. That's right. Now we have everything ready. Whenever a customer needs your GPU power, we have everything ready to support. Yeah, that's a great idea to support their AI infrastructure. Yeah. So to make sure the world can move AI even faster pace.
C
Charles Leang48:56
Yeah. Well, speed matters a lot. You know right now the technology is moving very fast and the people who move fast are establishing their position. Yeah. Yeah. And so speed is everything in the beginning of a new industry. Yeah. And right now speed matters. And I promise our audience here with our data center building block solution, now we can have a customer shrink their time to build a data center. Yeah. From idea to online data center in a couple of quarters or a couple of months instead of a couple of years. Yeah. Couple of years. So good friend, right? Really supporting your incredibly good friend.
J
Jensen Huang49:41
Thank you. All right, I think that's my cue.
C
Charles Leang49:45
When you get thanked three times in a minute, that means time for you to leave. Thank you, Charles. Thank you. I appreciate it.
J
Jensen Huang49:56
All right. Thanks, everybody.
C
Charles Leang50:08
You know, our marketing guy next time she told me Jensen can be here in 1 hour. Okay. Anyway, so lots of good products. So start from B200 HGX 10U air core system. Super Micro is one of the first companies to start shipping B2 HGX systems since early March this year, and also our B200 HGX for you. Again, Super Micro is the first company to ship two B200 HGX for you starting from March this year, and we already shipped many thousand systems to the market. And GP200 MVL fully ready for production, whenever you place order we can ship now, especially with our promise. And GB300 MV72 is coming very soon.
And other than that, we have B21 HGX 4U with our DLC2, kind of 50 dB noise level. This skill will be ready in small volume in one month, high volume in two months. And MI325X air cool, MI325 4U DLC, B300 HGX 4U air cooler solution should be ready in small volume hopefully in one month, high volume in two months. And then B200 DLC1 in that has been shipping. B200 DLC2 again can be ready to ship within the next few weeks, high volume in one or two months. And then B300 DLC2 hopefully within a quarter. And then MS3 350 air cooling shipping in two months. For example, MI355 for you and E3S, old flash AI storage for you, 90-bay scale-out storage, 400G high-speed data switch. So a variety of products, whenever you need, Super Micro pretty much has a solution ready for you to speed up your time to market, time to online.
So, kind of wrong time, that's why I'm a little bit nervous. So thank you everyone for joining us this year's Computex. To conclude, Super Micro is ready to ship DLC2 solutions and fully ready to offer DCBPS data center building block solutions to our market today. And Super Micro today is not just a brand name AI server system or rack company; we are also one of the largest CSP ODM manufacturers, especially with our Malaysia capacity. Now we can ship many more than 5,000 racks per month. Super Micro today is not just a system and rack-scale company; we are a data center AI factory infrastructure company. And especially in green computing, green computing saves you money and makes our planet more beautiful and greener. That's for green computing. And thank you so much for your support, for your trust. Super Micro will continue to grow at a fast pace to support you, to support our industry. Thank you.
N
Narrator54:38
Thank you. Thank you so much, Charles, for the very inspiring CEO keynote session.