TechPrufer
Get started
Coreweave
Coreweave logoCoreweave · Feature trail

Data Center Expansion and Network Infrastructure

3 posts · May 2026 – Sep 2026

What this trail covers

CoreWeave enhances its AI infrastructure by expanding its data center footprint with the opening of LGA1 in NYC, improving network infrastructure with ultra-low latency and high-bandwidth connectivity, and offering regional availability of GPU and CPU instances with high-performance storage. This expansion serves compute-intensive workloads across North America, leveraging purpose-built data centers for AI with superior networking services using NVIDIA Quantum-2 InfiniBand and Spectrum switches. The deployment of thousands of NVIDIA A40 GPUs further strengthens their offering for visual computing and AI workloads.

All Coreweave trails
  1. Multi-Rack NVIDIA Vera Rubin NVL72 Bring-Up | CoreWeaveThis post details the engineering challenges and solutions involved in bringing up multi-rack NVIDIA Vera Rubin NVL72 clusters. It highlights the automation of hardware lifecycle management from initial detection to production readiness, including firmware updates and configuration. The post emphasizes the importance of rack-level performance validation, identifying and troubleshooting outliers before they become stragglers in distributed workloads. It also describes the construction of a non-blocking RoCE network fabric across racks, utilizing ConnectX-9 SuperNICs and a two-tier, multi-rail, multi-plane architecture to achieve 1.6 Tb/s of backend network connectivity per GPU and scale to approximately 128,000 GPUs.
  2. CoreWeave Brings Liquid Cooled Switching to AI Networking | CoreWeave BlogThis post details the deployment of NVIDIA Spectrum-X SN6600-LD liquid-cooled Ethernet switches, doubling per-rack network bandwidth to 1.64 Pb/s. It highlights the switch's 102.4 Tb/s capacity, 200G SerDes, and compact 2U form factor, enabling a non-blocking, multi-plane, multi-rail spine and leaf fabric for Vera Rubin NVL72 GPUs. The post also describes the integration of these switches with CoreWeave Mission Control's Valvey (liquid cooling valve assembly) and Racky (rack control manager) for unified thermal management and operational control, and the Rack Lifecycle Controller (RLCC) for automated rack-level infrastructure management. Measurable gains include a 100% increase in total switching performance per rack, a 62.5% reduction in rack footprint, and a 30% improvement in power efficiency compared to air-cooled systems.
  3. CoreWeave announces the deployment of thousands of NVIDIA A40 GPUs, the largest fleet in North America. This addition enhances their infrastructure for visual computing, data science, and AI training, offering 2x rendering and 3x AI training performance improvements over previous generations. The A40 GPUs feature 48 GB of graphics memory, RT Cores, Tensor Cores, and CUDA Cores, accelerating photorealistic rendering, professional graphics applications, and large-scale AI/data science model training without code changes. This expansion complements their existing network infrastructure and data center capabilities.
  4. CoreWeave Data Center Operations: Built for AI | CoreWeaveThis post details CoreWeave's data center operations built for AI, focusing on superior networking services with NVIDIA Quantum-2 InfiniBand, forward-thinking sustainability through direct-to-chip liquid cooling, and top-tier security with ISO 27001, SOC2, and GDPR compliance. It also highlights the human element with dedicated support teams (FleetOps, CloudOps, DCTs) ensuring reliability and performance.
  5. Introducing LGA1: CoreWeave’s Cutting-Edge Data Center in NYCCoreWeave opens LGA1, a new data center in NYC, providing ultra-low latency access (sub-1ms to Manhattan) to its GPU-accelerated cloud instances and CPU-only instances. The facility is ISO 27001 certified, SSAE 18 SOC 2 compliant, and Energy Star Certified, featuring a Tier 1 internet backbone for low latency and high reliability. This expansion, along with existing data centers, offers connected compute across North America with redundant 200Gbps+ public internet and 400Gbps+ dark fiber transport between facilities.
  7. Our Networking Services For GenAI | CoreWeaveThis post details CoreWeave's networking services for GenAI, focusing on the use of NVIDIA Quantum-2 InfiniBand for ultra-low latency and high throughput interconnects between GPUs, enabling GPU superclusters and faster data exchanges. It highlights SHARP for collective operations and SHIELD for link failure recovery, achieving up to 3.2Tbps on 8NDR. The post also discusses direct connections for multi-cloud needs with no egress fees and the use of upgraded Ethernet with NVIDIA Spectrum switches and BlueField DPUs for predictable performance between GPU nodes and storage clusters. It quantifies InfiniBand link reliability at 99.9975% with 95%+ goodput.
  8. The Redesign of the Data Center Has Already Started | CoreWeaveThis post details the fundamental redesign of data centers to meet the demands of AI workloads, contrasting them with traditional data centers. It highlights the architectural shifts required in power density (from 5-10 KW/rack to 60+ KW/rack), cooling (transitioning from air cooling to liquid cooling due to increased heat generation from multi-GPU servers), and networking (reconfiguring backbones for higher bandwidth to facilitate efficient communication between densely packed GPU racks). It also discusses the need for purpose-built software stacks for AI workloads, emphasizing resilience and optimization. The post mentions CoreWeave's commitment to building supercomputer-scale GPU clusters using technologies that optimize performance and reliability for tens of thousands of GPUs simultaneously, and introduces SUNK (Slurm on Kubernetes) and Tensorizer as examples of their purpose-built software tools.