➀ The growing demand for AI infrastructure requires scalable interconnects; ➁ Ultra Ethernet and UALink address the challenges of scaling AI acceleration clusters; ➂ Synopsys' IP solutions offer efficient and power-efficient interconnects for data centers.

➀ Fujitsu unveils its Monaka processor, a 144-core Armv9-based chip designed for future data centers; ➁ The chip is built on TSMC's N2 process and features a CoWoS system-in-package with SRAM tiles and hybrid copper bonding; ➂ Fujitsu aims for superior energy efficiency by 2026-2027, using air cooling.

➀ Meta's plan to build a nuclear-powered data center was canceled due to the discovery of a rare bee species; ➁ The project faced environmental and regulatory issues; ➂ Despite the setback, Meta is looking for a new location and clean energy sources.

➀ Analog Bits showcased its on-die sensing IP and power management solutions at TSMC OIP; ➁ The company highlighted its progress in 3nm and 2nm technology nodes; ➂ Collaborations with Arm were discussed, focusing on power management and clocking IPs.

➀ SuperMicro unveils its new NVIDIA GB200 NVL72 SuperCluster, a liquid-cooled Exascale Compute system in a single rack; ➁ The system includes 72 NVIDIA Blackwell B200 GPUs and offers high-speed interconnects; ➂ Supermicro provides end-to-end liquid-cooling solutions and comprehensive deployment services.

➀ Dell has launched the new PowerEdge XE9712 with NVIDIA GB200 NVL72 AI servers, offering 30x faster real-time LLM performance over the H100 AI GPU; ➁ The system features 72 x B200 AI GPUs connected with NVLink technology, providing lightning-fast connectivity; ➂ Dell highlights the liquid-cooled system for maximizing datacenter power utilization and rapid deployment of AI clusters.

➀ AI is transforming data center infrastructure with innovations like advanced connectivity and chiplets; ➁ The shift to optical connectivity is enhancing performance and efficiency; ➂ Chiplet ecosystems are enabling scalable and cost-effective solutions for AI workloads.

➀ Rambus details its new HBM4 memory controller with speeds up to 10Gb/s and bandwidth of 2.56TB/sec; ➁ HBM4 offers 33% faster bandwidth than HBM3E and twice the speed of HBM3; ➂ SK hynix and Samsung are expected to mass produce HBM4 memory by the end of 2025.

➀ NVIDIA's new GB200 NVL72 AI server faces significant challenges due to its 132kW TDP; ➁ This makes it the highest-power-consuming server in history; ➂ The development of the server is experiencing delays due to technical issues.

➀ NVIDIA's dominance in AI GPU market extends to 2027 with Rubin Ultra AI GPU featuring 576GB of HBM4 memory; ➁ Competitors include AMD, Intel, Google, and others with their respective AI chip developments; ➂ TSMC's advanced CoWoS-L packaging technology and N3 process node support NVIDIA's next-gen AI GPUs.

➀ Kioxia demonstrated an optical interface SSD at FMS, developed with funding from NEDO to reduce data center power consumption by up to 40%. ➁ The optical interface allows for data transfer over long distances, enabling storage to be kept in separate rooms with minimal cooling requirements. ➂ The demonstration showed a slight loss in IOPS performance but significant latency advantages over traditional copper network links.

➀ The data center processing market shows robust growth, especially in processing, with a 20.3% increase. ➁ Nvidia dominates the market in terms of both revenue and profits, with a significant lead over competitors. ➂ The supply chain for AI GPUs has evolved, with memory supply becoming a critical component.

➀ AI data bottlenecks are addressed with PCIe 7.0, offering higher bandwidth and lower latency. ➁ PCIe 7.0 supports system-level optimization and reduces data center power consumption. ➂ The integration of photonics into PCIe ensures its relevance and benefits for future AI advancements.

1. Micron Technology introduces the Micron 9550 NVMe SSD series, boasting the world's fastest PCIe 5.0 data center SSDs; 2. These SSDs offer maximum sequential read speeds of 14GB/s and write speeds of 10GB/s, with up to 67% higher performance than competitors; 3. They are designed for AI workloads, with random read speeds of up to 3.3 million IOPS and write speeds of 400,000 IOPS, making them suitable for large language models and graph neural network training.

1. PAM4 SerDes technology significantly enhances data throughput and power efficiency for AI and data center applications; 2. The technology supports various reach requirements, from long to short, ensuring reliable and high-speed data transmission; 3. Cadence's advanced SerDes solutions and involvement in the Ultra Ethernet Consortium highlight ongoing innovations in Ethernet technology.

1. Samsung Electro-Mechanics announced it will supply high-performance chip boards for hyperscale data centers to AMD. 2. The chip boards are flip chip-ball grid array (FC-BGA). 3. This is the first official confirmation of the deal by Samsung Electro-Mechanics.

1. Micron introduces the 9550 datacenter SSD, utilizing PCIe gen 5×4 and 232-layer NAND. 2. The SSD is available in three formats: E1.S, E3.S, and U.2, designed for high performance in cloud and enterprise environments. 3. The 9550 offers significant performance improvements over its predecessors, with faster IOPS and bandwidth, and is more power-efficient in AI workloads.

1. Oxide Computer Company has shipped its first on-premises hyperscale 'cloud computers' targeting government, financial services, e-commerce, and service provider markets. 2. The company aims to revolutionize the private datacenter industry by combining cloud computing efficiency with on-premises security and control. 3. Oxide has shipped under 20 racks so far and hopes for customer scale-up based on positive outcomes.

1. A $2.4 million Texas property with no bedrooms but a 5786 sq ft built-in data center; 2. The data center features a full liquid cooling immersion system; 3. The property was previously used as a crypto mining base and is now being marketed for AI services, cloud hosting, traditional data centers, servers, or Bitcoin mining.

➀ Micron introduces the 6550 ION SSD, the first 60TB SSD with PCIe 5.0 x4 interface; ➁ The drive offers high performance with 12GB/s read and write speeds and low power consumption; ➂ Aimed at AI workloads, the SSD demonstrates improved energy efficiency and endurance.