Hewlett Packard Enterprise announced today that it has shipped its first NVIDIA Blackwell family-based solution, the NVIDIA GB200 NVL72. This rack-scale system by HPE is designed to help service providers and large enterprises quickly deploy very large, complex AI clusters with advanced, direct liquid cooling solutions to optimize efficiency and performance. "AI service providers and large enterprise model builders are under tremendous pressure to offer scalability, extreme performance, and fast time-to-deployment," said Trish Damkroger, senior vice president and general manager of HPC & AI Infrastructure Solutions, HPE. "As builders of the world's top three fastest systems with direct liquid cooling, HPE offers customers lower cost per token training and best-in-class performance with industry-leading services expertise."

The NVIDIA GB200 NVL72 features shared-memory, low-latency architecture with the latest GPU technology designed for extremely large AI models of over a trillion parameters, in one memory space. GB200 NVL72 offers seamless integration of NVIDIA CPUs, GPUs, compute and switch trays, networking, and software, bringing together extreme performance to address heavily parallelizable workloads, like generative AI (GenAI) model training and inferencing, along with NVIDIA software applications. "Engineers, scientists and researchers need cutting-edge liquid cooling technology to keep up with increasing power and compute requirements," said Bob Pette, vice president of enterprise platforms at NVIDIA. "Building on continued collaboration between HPE and NVIDIA, HPE's first shipment of NVIDIA GB200 NVL72 will help service providers and large enterprises efficiently build, deploy and scale large AI clusters."

AI reasoning models and agents are set to transform industries, but delivering their full potential at scale requires massive compute and optimized software. The "reasoning" process involves multiple models, generating many additional tokens, and demands infrastructure with a combination of high-speed communication, memory and compute to ensure real-time, high-quality results. To meet this demand, CoreWeave has launched NVIDIA GB200 NVL72-based instances, becoming the first cloud service provider to make the NVIDIA Blackwell platform generally available. With rack-scale NVIDIA NVLink across 72 NVIDIA Blackwell GPUs and 36 NVIDIA Grace CPUs, scaling to up to 110,000 GPUs with NVIDIA Quantum-2 InfiniBand networking, these instances provide the scale and performance needed to build and deploy the next generation of AI reasoning models and agents.

NVIDIA GB200 NVL72 on CoreWeave
NVIDIA GB200 NVL72 is a liquid-cooled, rack-scale solution with a 72-GPU NVLink domain, which enables the six dozen GPUs to act as a single massive GPU. NVIDIA Blackwell features many technological breakthroughs that accelerate inference token generation, boosting performance while reducing service costs. For example, fifth-generation NVLink enables 130 TB/s of GPU bandwidth in one 72-GPU NVLink domain, and the second-generation Transformer Engine enables FP4 for faster AI performance while maintaining high accuracy. CoreWeave's portfolio of managed cloud services is purpose-built for Blackwell. CoreWeave Kubernetes Service optimizes workload orchestration by exposing NVLink domain IDs, ensuring efficient scheduling within the same rack. Slurm on Kubernetes (SUNK) supports the topology block plug-in, enabling intelligent workload distribution across GB200 NVL72 racks. In addition, CoreWeave's Observability Platform provides real-time insights into NVLink performance, GPU utilization and temperatures.

Supermicro, Inc., a Total IT Solution Provider for AI/ML, HPC, Cloud, Storage, and 5G/Edge, is announcing full production availability of its end-to-end AI data center Building Block Solutions accelerated by the NVIDIA Blackwell platform. The Supermicro Building Block portfolio provides the core infrastructure elements necessary to scale Blackwell solutions with exceptional time to deployment. The portfolio includes a broad range of air-cooled and liquid-cooled systems with multiple CPU options. These include superior thermal design supporting traditional air cooling, liquid-to-liquid (L2L) and liquid-to-air (L2A) cooling. In addition, a full data center management software suite, rack-level integration, including full network switching and cabling and cluster-level L12 solution validation can be delivered as turn-key offering with global delivery, professional support, and service.

"In this transformative moment of AI, where scaling laws are pushing the limits of data center capabilities, our latest NVIDIA Blackwell-powered solutions, developed through close collaboration with NVIDIA, deliver outstanding computational power," said Charles Liang, president and CEO of Supermicro. "Supermicro's NVIDIA Blackwell GPU offerings in plug-and-play scalable units with advanced liquid cooling and air cooling are empowering customers to deploy an infrastructure that supports increasingly complex AI workloads while maintaining exceptional efficiency. This reinforces our commitment to providing sustainable, cutting-edge solutions that accelerate AI innovation."

ASUS is proud to announce that ASUS AI POD, featuring the NVIDIA GB200 NVL72 platform, is ready to ramp-up production for a scheduled shipping date of March 2025. ASUS remains dedicated to providing comprehensive end-to-end solutions and software services, encompassing everything from AI supercomputing to cloud services. With a strong focus on fostering AI adoption across industries, ASUS is positioned to empower clients in accelerating their time to market by offering a full spectrum of solutions.

Proof of concept, funded by ASUS
Honoring the commitment to delivering exceptional value to clients, ASUS is set to launch a proof of concept (POC) for the groundbreaking ASUS AI POD, powered by the NVIDIA Blackwell platform. This exclusive opportunity is now open to a select group of innovators who are eager to harness the full potential of AI computing. Innovators and enterprises can experience firsthand the full potential of AI and deep learning solutions at exceptional scale. To take advantage of this limited-time offer, please complete this surveyi at: forms.office.com/r/FrAbm5BfH2. The expert ASUS team of NVIDIA GB200 specialists will guide users through the next steps.

According to a well-known industry analyst, Ming-Chi Kuo, NVIDIA has restructured its "Blackwell" architecture roadmap, emphasizing dual-die designs using CoWoS-L packaging technology. The new roadmap eliminates several single-die products that would have used CoWoS-S packaging, changing NVIDIA's manufacturing strategy. The 200 Series will exclusively use dual-die designs with CoWoS-L packaging, featuring the GB200 NVL72 and HGX B200 systems. Notably absent is the previously expected B200A single-die variant. The 300 Series will include both dual-die and single-die options, though NVIDIA and cloud providers are prioritizing the GB200 NVL72 dual-die system. Starting Q1 2025, NVIDIA will reduce H series production, which uses CoWoS-S packaging, while ramping up 200 Series production. This transition indicates significantly decreased demand for CoWoS-S capacity through 2025.

While B300 systems using single-die CoWoS-S are planned for 2026 mass production, the current focus remains on dual-die CoWoS-L products. From TSMC's perspective, the transition between Blackwell generations requires minimal process adjustments, as both use similar front-end-of-line processes with only back-end-of-line modifications needed. Supply chain partners heavily dependent on CoWoS-S production face significant impact, reflected in recent stock price corrections. However, NVIDIA maintains this change reflects product strategy evolution rather than market demand weakness. TSMC continues expanding CoWoS-R capacity while slowing CoWoS-S expansion, viewing AI and high-performance computing as sustained growth drivers despite these packaging technology transitions.

Giga Computing, a subsidiary of GIGABYTE and an industry leader in generative AI servers and advanced cooling technologies, announced new GIGABYTE G893 series servers using the NVIDIA HGX B200 platform. The launch of these flagship 8U air-cooled servers, the G893-SD1-AAX5 and G893-ZD1-AAX5, signifies a new architecture and platform change for GIGABYTE in the demanding world of high-performance computing and AI, setting new standards for speed, scalability, and versatility.

These servers join GIGABYTE's accelerated computing portfolio alongside the NVIDIA GB200 NVL72 platform, which is a rack-scale design that connects 36 NVIDIA Grace CPUs and 72 NVIDIA Blackwell GPUs. At CES 2025 (January 7-10), the GIGABYTE booth will display the NVIDIA GB200 NVL72, and attendees can engage in discussions about the benefits of GIGABYTE platforms with the NVIDIA Blackwell architecture.

NVIDIA "Blackwell" series is barely out with B100, B200, and GB200 chips shipping to OEMs and hyperscalers, but the company is already setting in its upgraded "Blackwell Ultra" plans with its upcoming GB300 AI server. According to UDN, the next generation NVIDIA system will be powered by the B300 GPU chip, operating at 1400 W and delivering a remarkable 1.5x improvement in FP4 performance per card compared to its B200 predecessor. One of the most notable upgrades is the memory configuration, with each GPU now sporting 288 GB of HBM3e memory, a substantial increase from the previous 192 GB of GB200. The new design implements a 12-layer stack architecture, advancing from the GB200's 8-layer configuration. The system's cooling infrastructure has been completely reimagined, incorporating advanced water cooling plates and enhanced quick disconnects in the liquid cooling system.

Networking capabilities have also seen a substantial upgrade, with the implementation of ConnectX 8 network cards replacing the previous ConnectX 7 generation, while optical modules have been upgraded from 800G to 1.6T, ensuring faster data transmission. Regarding power management and reliability, the GB300 NVL72 cabinet will standardize capacitor tray implementation, with an optional Battery Backup Unit (BBU) system. Each BBU module costs approximately $300 to manufacture, with a complete GB300 system's BBU configuration totaling around $1,500. The system's supercapacitor requirements are equally substantial, with each NVL72 rack requiring over 300 units, priced between $20-25 per unit during production due to its high-power nature. The GB300, carrying Grace CPU and Blackwell Ultra GPU, also introduces the implementation of LPCAMM on its computing boards, indicating that the LPCAMM memory standard is about to take over servers, not just laptops and desktops. We have to wait for the official launch before seeing LPCAMM memory configurations.

Renowned for its groundbreaking innovations in computing technology, GIGABYTE Technology continues to spearhead advancements that redefine the industry. From January 7 to 10, 2025, GIGABYTE will participate in CES, showcasing its latest breakthroughs in high-performance computing and personal computing solutions. Under the theme "ACCEVOLUTION," GIGABYTE's booth will present a diverse range of cutting-edge technologies. Highlights include:

• Rack-scale cluster computing solution for large-scale data centers
• AI servers supporting next-generation "superchips"
• Advanced cooling technologies that optimize energy efficiency and performance
• Workstations tailored for smaller-scale workloads
• Mini-PCs, embedded systems, and in-vehicle platforms designed for endpoint deployment and commercial applications

NVIDIA and Microsoft today unveiled product integrations designed to advance full-stack NVIDIA AI development on Microsoft platforms and applications. At Microsoft Ignite, Microsoft announced the launch of the first cloud private preview of the Azure ND GB200 V6 VM series, based on the NVIDIA Blackwell platform. The Azure ND GB200 v6 will be a new AI-optimized virtual machine (VM) series and combines the NVIDIA GB200 NVL72 rack design with NVIDIA Quantum InfiniBand networking.

In addition, Microsoft revealed that Azure Container Apps now supports NVIDIA GPUs, enabling simplified and scalable AI deployment. Plus, the NVIDIA AI platform on Azure includes new reference workflows for industrial AI and an NVIDIA Omniverse Blueprint for creating immersive, AI-powered visuals. At Ignite, NVIDIA also announced multimodal small language models (SLMs) for RTX AI PCs and workstations, enhancing digital human interactions and virtual assistants with greater realism.

ASRock Rack Inc., a leading innovative server company, today announces its presence at SC24, held at the Georgia World Congress Center in Atlanta from November 18-21. At booth #3609, ASRock Rack will showcase a comprehensive high-performance portfolio of server boards, systems, and rack solutions with NVIDIA accelerated computing platforms, helping address the needs of enterprises, organizations, and data centers.

Artificial intelligence (AI) and high-performance computing (HPC) continue to reshape technology. ASRock Rack is presenting a complete suite of solutions spanning edge, on-premise, and cloud environments, engineered to meet the demand of AI and HPC. The 2U short-depth MECAI, incorporating the NVIDIA GH200 Grace Hopper Superchip, is developed to supercharge accelerated computing and generative AI in space-constrained environments. The 4U10G-TURIN2 and 4UXGM-GNR2, supporting ten and eight NVIDIA H200 NVL PCIe GPUs respectively, are aiming to help enterprises and researchers tackle every AI and HPC challenge with enhanced performance and greater energy efficiency. NVIDIA H200 NVL is ideal for lower-power, air-cooled enterprise rack designs that require flexible configurations, delivering acceleration for AI and HPC workloads regardless of size.

CoolIT Systems (CoolIT), the world leader in liquid cooling systems for AI and high-performance computing, introduces the CHx1000, the world's highest-density liquid-to-liquid coolant distribution unit (CDU). Designed for mission-critical applications, the CHx1000 is purpose-built to cool the NVIDIA Blackwell platform and other demanding AI workloads where liquid cooling is now necessary.

"CoolIT created the CHx1000 to provide the high capacity and pressure delivery required to direct liquid cool NVIDIA Blackwell and future generations of high-performance AI accelerators," said Patrick McGinn, CoolIT's COO. "Besides exceptional performance, serviceability and reliability are central to the CHx1000's design. The single rack-sized unit is fully front and back serviceable with hot-swappable critical components. Precision coolant controls and multiple levels of redundancy provide for steady, uninterrupted operation."

According to The Information, NVIDIA's latest "Blackwell" processors are reportedly encountering significant thermal management issues in high-density server configurations, potentially affecting deployment timelines for major tech companies. The challenges emerge specifically in NVL72 GB200 racks housing 72 GB200 processors, which can consume up to 120 kilowatts of power per rack, weighting a "mere" 3,000 pounds (or about 1.5 tons). These thermal concerns have prompted NVIDIA to revisit and modify its server rack designs multiple times to prevent performance degradation and potential hardware damage. Hyperscalers like Google, Meta, and Microsoft, who rely heavily on NVIDIA GPUs for training their advanced language models, have allegedly expressed concerns about possible delays in their data center deployment schedules.

The thermal management issues follow earlier setbacks related to a design flaw in the Blackwell production process. The problem stemmed from the complex CoWoS-L packaging technology, which connects dual chiplets using RDL interposer and LSI bridges. Thermal expansion mismatches between various components led to warping issues, requiring modifications to the GPU's metal layers and bump structures. A company spokesperson characterized these modifications as part of the standard development process, noting that a new photomask resolved this issue. The Information states that mass production of the revised Blackwell GPUs began in late October, with shipments expected to commence in late January. However, these timelines are unconfirmed by NVIDIA, and some server makers like Dell confirmed that these GB200 NVL72 liquid-cooled systems are shipping now, not in January, with CoreWave GPU cloud provider as a customer. The original report could be using older information, as Dell is one of NVIDIA's most significant partners and among the first in the supply chain to gain access to new GPU batches.

We know that NVIDIA's latest "Blackwell" GPUs are fast, but how much faster are they over the previous generation "Hopper"? Thanks to the latest MLPerf Training v4.1 results, NVIDIA's HGX B200 Blackwell platform has demonstrated massive performance gains, measuring up to 2.2x improvement per GPU compared to its HGX H200 Hopper. The latest results, verified by MLCommons, reveal impressive achievements in large language model (LLM) training. The Blackwell architecture, featuring HBM3e high-bandwidth memory and fifth-generation NVLink interconnect technology, achieved double the performance per GPU for GPT-3 pre-training and a 2.2x boost for Llama 2 70B fine-tuning compared to the previous Hopper generation. Each benchmark system incorporated eight Blackwell GPUs operating at a 1,000 W TDP, connected via NVLink Switch for scale-up.

The network infrastructure utilized NVIDIA ConnectX-7 SuperNICs and Quantum-2 InfiniBand switches, enabling high-speed node-to-node communication for distributed training workloads. While previous Hopper-based systems required 256 GPUs to optimize performance for the GPT-3 175B benchmark, Blackwell accomplished the same task with just 64 GPUs, leveraging its larger HBM3e memory capacity and bandwidth. One thing to look out for is the upcoming GB200 NVL72 system, which promises even more significant gains past the 2.2x. It features expanded NVLink domains, higher memory bandwidth, and tight integration with NVIDIA Grace CPUs, complemented by ConnectX-8 SuperNIC and Quantum-X800 switch technologies. With faster switching and better data movement with Grace-Blackwell integration, we could see even more software optimization from NVIDIA to push the performance envelope.

Last week, we got a preview of Microsoft's Azure production-ready NVIDIA "Blackwell" GB200 system, showing that only a third of the rack that goes in the data center is actually holding the compute elements, with the other two-thirds holding the cooling compartment to cool down the immense heat output from tens of GB200 GPUs. Today, Google is showing off a part of its own infrastructure ahead of the Google Cloud App Dev & Infrastructure Summit, taking place on October 30, digitally as an event. Shown below are two racks standing side by side, connecting NVIDIA "Blackwell" GB200 NVL cards with the rest of the Google infrastructure. Unlike Microsoft Azure, Google Cloud uses a different data center design in its facilities.

There is one rack with power distribution units, networking switches, and cooling distribution units, all connected to the compute rack, which houses power supplies, GPUs, and CPU servers. Networking equipment is present, and it connects to Google's "global" data center network, which is Google's own data center fabric. We are not sure what is the fabric connection of choice between these racks; as for optimal performance, NVIDIA recommends InfiniBand (Mellanox acquisition). However, given that Google's infrastructure is set up differently, there may be Ethernet switches present. Interestingly, Google's design of GB200 racks differs from Azure's, as it uses additional rack space to distribute the coolant to its local heat exchangers, i.e., coolers. We are curious to see if Google releases more information on infrastructure, as it has been known as the infrastructure king because of its ability to scale and keep everything organized.

To drive the development of open, efficient and scalable data center technologies, NVIDIA today announced that it has contributed foundational elements of its NVIDIA Blackwell accelerated computing platform design to the Open Compute Project (OCP) and broadened NVIDIA Spectrum-X support for OCP standards.

At this year's OCP Global Summit, NVIDIA will be sharing key portions of the NVIDIA GB200 NVL72 system electro-mechanical design with the OCP community — including the rack architecture, compute and switch tray mechanicals, liquid-cooling and thermal environment specifications, and NVIDIA NVLink cable cartridge volumetrics — to support higher compute density and networking bandwidth.

NVIDIA's "Blackwell" series of GPUs, including B100, B200, and GB200, are reportedly sold out for 12 months or an entire year. This directly means that if a new customer is willing to order a new Blackwell GPU now, there is a 12-month waitlist to get that GPU. Analyst from Morgan Stanley Joe Moore confirmed that in a meeting with NVIDIA and its investors, NVIDIA executives confirmed that the demand for "Blackwell" is so great that there is a 12-month backlog to fulfill first before shipping to anyone else. We expect that this includes customers like Amazon, META, Microsoft, Google, Oracle, and others, who are ordering GPUs in insane quantities to keep up with the demand from their customers.

The previous generation of "Hopper" GPUs was ordered in 10s of thousands of GPUs, while this "Blackwell" generation was ordered in 100s of thousands of GPUs simultaneously. For NVIDIA, that is excellent news, as that demand is expected to continue. The only one standing in the way of customers is TSMC, which manufactures these GPUs as fast as possible to meet demand. NVIDIA is one of TSMC's largest customers, so wafer allocation at TSMC's facilities is only expected to grow. We are now officially in the era of the million-GPU data centers, and we can only question at what point this massive growth stops or if it will stop at all in the near future.

NVIDIA is reportedly considering a significant design change for its GPU products, shifting from the current on-board solution to an independent GPU socket design following the GB200 shipment in Q4, according to reports from MoneyDJ and the Economic Daily News quoted by TrendForce. This move is not new in the industry, AMD has already introduced socket design in 2023 with their MI300A series via Supermicro dedicated servers. The B300 series, expected to become NVIDIA's mainstream product in the second half of 2025, is rumored to be the main beneficiary of this design change that could improve yield rates, though it may come with some performance trade-offs.

According to the Economic Daily News, the socket design will simplify after-sales service and server board maintenance, allowing users to replace or upgrade the GPUs quickly. The report also pointed out that based on the slot design, boards will contain up to four NVIDIA GPUs and a CPU, with each GPU having its dedicated slot. This will bring benefits for Taiwanese manufacturers like Foxconn and LOTES, who will supply different components and connectors. The move seems logical since with the current on-board design, once a GPU becomes faulty, the entire motherboard needs to be replaced, leading to significant downtime and high operational and maintenance costs.

Late Tuesday, Microsoft Azure shared an interesting picture on its social media platform X, showcasing the pinnacle of GPU-accelerated servers—NVIDIA "Blackwell" GB200-powered AI systems. Microsoft is one of NVIDIA's largest customers, and the company often receives products first to integrate into its cloud and company infrastructure. Even NVIDIA listens to feedback from companies like Microsoft about designing future products, especially those like the now-canceled NVL36x2 system. The picture below shows a massive cluster that roughly divides the compute area into a single-third of the entire system, with a gigantic two-thirds of the system dedicated to closed-loop liquid cooling.

The entire system is connected using Infiniband networking, a standard for GPU-accelerated systems due to its lower latency in packet transfer. While the details of the system are scarce, we can see that the integrated closed-loop liquid cooling allows the GPU racks to be in a 1U form for increased density. Given that these systems will go into the wider Microsoft Azure data centers, a system needs to be easily maintained and cooled. There are indeed limits in power and heat output that Microsoft's data centers can handle, so these types of systems often fit inside internal specifications that Microsoft designs. There are more compute-dense systems, of course, like NVIDIA's NVL72, but hyperscalers should usually opt for other custom solutions that fit into their data center specifications. Finally, Microsoft noted that we can expect to see more details at the upcoming Microsoft Ignite conference in November and learn more about its GB200-powered AI systems.

NVIDIA has reportedly discontinued its dual-rack GB200 NVL36x2 GPU model, opting to focus on the single-rack GB200 NVL72 and NVL36 models. This shift, revealed by industry analyst Ming-Chi Kuo, aims to simplify NVIDIA's offerings in the AI and HPC markets. The decision was influenced by major clients like Microsoft, who prefer the NVL72's improved space efficiency and potential for enhanced inference performance. While both models perform similarly in AI large language model (LLM) training, the NVL72 is expected to excel in non-parallelizable inference tasks. As a reminder, the NVL72 features 36 Grace CPUs, delivering 2,592 Arm Neoverse V2 cores with 17 TB LPDDR5X memory with 18.4 TB/s aggregate bandwidth. Additionally, it includes 72 Blackwell GB200 SXM GPUs that have a massive 13.5 TB of HBM3e combined, running at 576 TB/s aggregate bandwidth.

However, this shift presents significant challenges. The NVL72's power consumption of around 120kW far exceeds typical data center capabilities, potentially limiting its immediate widespread adoption. The discontinuation of the NVL36x2 has also sparked concerns about NVIDIA's execution capabilities and may disrupt the supply chain for assembly and cooling solutions. Despite these hurdles, industry experts view this as a pragmatic approach to product planning in the dynamic AI landscape. While some customers may be disappointed by the dual-rack model's cancellation, NVIDIA's long-term outlook in the AI technology market remains strong. The company continues to work with clients and listen to their needs, to position itself as a leader in high-performance computing solutions.

During its Q2 2024 earnings call, NVIDIA confirmed that its upcoming Blackwell-based products are facing low-yield challenges. However, the company announced that it has implemented design changes to improve the production yields of its B100 and B200 processors. Despite these setbacks, NVIDIA remains optimistic about its production timeline. The tech giant plans to commence the production ramp of Blackwell GPUs in Q4 2024, with expected shipments worth several billion dollars by the end of the year. In an official statement, NVIDIA explained, "We executed a change to the Blackwell GPU mask to improve production yield." The company also reaffirmed that it had successfully sampled Blackwell GPUs with customers in the second quarter.

However, NVIDIA acknowledged that meeting demand required producing "low-yielding Blackwell material," which impacted its gross margins. During an earnings call, NVIDIA's CEO Jensen Huang assured investors that the supply of B100 and B200 GPUs will be there. He expressed confidence in the company's ability to mass-produce these chips starting in the fourth quarter. The Blackwell B100 and B200 GPUs use TSMC's CoWoS-L packaging technology and a complex design, which prompted rumors about the company facing yield issues with its designs. Reports suggest that initial challenges arose from mismatched thermal expansion coefficients among various components, leading to warping and system failures. However, now the company claims that the fix that solved these problems was a new GPU photomask, which bumped yields back to normal levels.

Despite recent rumors speculating on NVIDIA's supposed cancellation of the B100 in favor of the B200A, TrendForce reports that NVIDIA is still on track to launch both the B100 and B200 in the 2H24 as it aims to target CSP customers. Additionally, a scaled-down B200A is planned for other enterprise clients, focusing on edge AI applications.

TrendForce reports that NVIDIA will prioritize the B100 and B200 for CSP customers with higher demand due to the tight production capacity of CoWoS-L. Shipments are expected to commence after 3Q24. In light of yield and mass production challenges with CoWoS-L, NVIDIA is also planning the B200A for other enterprise clients, utilizing CoWoS-S packaging technology.

With the growing demand for high-speed computing, more effective cooling solutions for AI servers are gaining significant attention. TrendForce's latest report on AI servers reveals that NVIDIA is set to launch its next-generation Blackwell platform by the end of 2024. Major CSPs are expected to start building AI server data centers based on this new platform, potentially driving the penetration rate of liquid cooling solutions to 10%.

Air and liquid cooling systems to meet higher cooling demands
TrendForce reports that the NVIDIA Blackwell platform will officially launch in 2025, replacing the current Hopper platform and becoming the dominant solution for NVIDIA's high-end GPUs, accounting for nearly 83% of all high-end products. High-performance AI server models like the B200 and GB200 are designed for maximum efficiency, with individual GPUs consuming over 1,000 W. HGX models will house 8 GPUs each, while NVL models will support 36 or 72 GPUs per rack, significantly boosting the growth of the liquid cooling supply chain for AI servers.

TrendForce's latest industry report on AI servers reveals that high demand for advanced AI servers from major CSPs and brand clients is expected to continue in 2024. Meanwhile, TSMC, SK hynix, Samsung, and Micron's gradual production expansion has significantly eased shortages in 2Q24. Consequently, the lead time for NVIDIA's flagship H100 solution has decreased from the previous 40-50 weeks to less than 16 weeks.

TrendForce estimates that AI server shipments in the second quarter will increase by nearly 20% QoQ, and has revised the annual shipment forecast up to 1.67 million units—marking a 41.5% YoY growth.

A new startup emerged out of stealth mode today to power the next generation of generative AI. Etched is a company that makes an application-specific integrated circuit (ASIC) to process "Transformers." The transformer is an architecture for designing deep learning models developed by Google and is now the powerhouse behind models like OpenAI's GPT-4o in ChatGPT, Anthropic Claude, Google Gemini, and Meta's Llama family. Etched wanted to create an ASIC for processing only the transformer models, making a chip called Sohu. The claim is Sohu outperforms NVIDIA's latest and greatest by an entire order of magnitude. Where a server configuration with eight NVIDIA H100 GPU clusters pushes Llama-3 70B models at 25,000 tokens per second, and the latest eight B200 "Blackwell" GPU cluster pushes 43,000 tokens/s, the eight Sohu clusters manage to output 500,000 tokens per second.

Why is this important? Not only does the ASIC outperform Hopper by 20x and Blackwell by 10x, but it also serves so many tokens per second that it enables an entirely new fleet of AI applications requiring real-time output. The Sohu architecture is so efficient that 90% of the FLOPS can be used, while traditional GPUs boast a 30-40% FLOP utilization rate. This translates into inefficiency and waste of power, which Etched hopes to solve by building an accelerator dedicated to power transformers (the "T" in GPT) at massive scales. Given that the frontier model development costs more than one billion US dollars, and hardware costs are measured in tens of billions of US Dollars, having an accelerator dedicated to powering a specific application can help advance AI faster. AI researchers often say that "scale is all you need" (resembling the legendary "attention is all you need" paper), and Etched wants to build on that.

TrendForce reports that NVIDIA's Hopper H100 began to see a reduction in shortages in 1Q24. The new H200 from the same platform is expected to gradually ramp in Q2, with the Blackwell platform entering the market in Q3 and expanding to data center customers in Q4. However, this year will still primarily focus on the Hopper platform, which includes the H100 and H200 product lines. The Blackwell platform—based on how far supply chain integration has progressed—is expected to start ramping up in Q4, accounting for less than 10% of the total high-end GPU market.

The die size of Blackwell platform chips like the B100 is twice that of the H100. As Blackwell becomes mainstream in 2025, the total capacity of TSMC's CoWoS is projected to grow by 150% in 2024 and by over 70% in 2025, with NVIDIA's demand occupying nearly half of this capacity. For HBM, the NVIDIA GPU platform's evolution sees the H100 primarily using 80 GB of HBM3, while the 2025 B200 will feature 288 GB of HBM3e—a 3-4 fold increase in capacity per chip. The three major manufacturers' expansion plans indicate that HBM production volume will likely double by 2025.