Aivres at GTC 2026 Showcasing Solutions for Enabling AI Factories, Engineered for Intelligence at Scale
Aivres Empowers Limitless Possibilities for Enterprises and Hyperscalers with End-to-End AI Data Center Infrastructure Solutions
SAN JOSE, Calif.—March 16, 2025— Aivres, a data center servers and storage solutions provider, announced that, at GTC 2026 , the company will showcase its AI Factory solutions for maximizing value from data. Aivres is a Diamond Sponsor of the March 16-19 event at the San Jose Convention Center and will demonstrate its technologies, including multiple new products, in Booth 1313 under the theme “Empower Limitless Possibilities with End-to-End AI Data Center Infrastructure Solutions.”
Aivres at GTC 2026 is highlighting a range of its AI data center infrastructure solutions and capabilities:
Rack-Scale AI Supercomputer Based on the NVIDIA DGX Vera Rubin NVL72 platform—Aivres KRS8000V4 is purpose-engineered for giga-scale data centers running trillion-parameter training and long-context inference workloads. It offers extreme compute density, integrating 72 NVIDIA Rubin GPUs and 36 NVIDIA Vera CPUs in a single, fully liquid-cooled rack and delivering architectural peak performance (up to 2.5 EFLOPS for NVFP4 training and up to 3.6 EFLOPS for NVFP4 inference). KRS8000V4 provides up to 75TB of ultra-fast rack-local memory capacity, to accelerate data staging and checkpointing and support KV-cache expansion for long-context inference. The rack-scale solution offers full-stack networking and acceleration via NVIDIA NVLink 6 scale-up fabric (up to 3.6TB/s per GPU) combined with NVIDIA ConnectX-9 SuperNICS with RDMA for scale-out (up to 1.6Tb/s per GPU). NVIDIA BlueField-4 data processing unit (DPU) offloads storage and security functions to free CPU/GPU cycles for AI compute. Scalability is seamless from a single rack to multi-rack pods, and on to a full AI Factory. Utilizing NVIDIA Quantum-X800 InfiniBand and NVIDIA Spectrum-X Ethernet enables consistent peak performance and operational simplicity across deployments.
Flexible, Efficient AI Servers Based on NVIDIA Vera Rubin—To meet diverse deployment needs, Aivres is introducing the optimized KR2288 and KR5288 servers supporting the NVIDIA DGX Rubin NVL8 turnkey infrastructure solution, providing a blueprint for success in the age of agentic AI. Powered by eight NVIDIA Rubin GPUs, these systems deliver 400 PFLOPS of NVFP4 inference performance and 176TB/s of high-bandwidth memory (HBM) bandwidth, ensuring data is transported through infrastructure as quickly as possible. DGX Rubin NVL8 leverages sixth-generation NVIDIA NVLink to ensure seamless peer-to-peer communication for massive model parallelism. Aivres KR2288 is a 2U all-liquid-cooled, high-density system designed for large-scale deployments in modern, liquid-cooled data centers, pursuing ultimate PUE benefits. Based on a hybrid cooling solution, the 5U KR5288 offers deployment flexibility and cost savings. It is ideal for existing data centers, requiring no changes to power, cooling or rack architecture and enabling rapid adoption of the latest NVIDIA Vera Rubin platform. Both Aivres platforms support Intel and AMD CPU platforms, meeting enterprise compatibility and multi-task processing requirements.
Versatile AI Server—For professional workloads at the edge of AI and visual computing, Aivres will also introduce the KR6278, a new 6U PCIe system compliant with the NVIDIA MGX KR6278 delivers powerful NVIDIA RTX PRO™ 6000 Blackwell Server Edition and NVIDIA RTX PRO 4500 Blackwell Server Edition GPU support, merging unmatched AI compute with cutting-edge visual performance and enabling broad application reach. KR6278 redefines possibilities across agentic AI, physical AI, scientific computing, photorealistic rendering, 3D graphics and real-time video processing.
Solution for the KV Cache Challenge with BlueField-4—The explosive growth of the KV Cache as context length is increased presents a critical bottleneck in large language model (LLM) inference, burdening both limited GPU memory capacity and data movement bandwidth. The entire Aivres product line integrates NVIDIA BlueField-4, which act as data and storage processors for ultra-fast, pod-local context memory storage tier shared across GPUs in the NVL72 pod. This architecture offers a dramatic KV cache capacity boost, allowing each GPU to gain an effective additional 16TB of additional KV cache capacity per GPU. This solution reduces data movement and GPU stalls, delivering up to 5x higher tokens‑per‑second throughput and up to 5x better power efficiency compared with traditional storage approaches, so inference remains smooth and responsive even as context windows grow.
Aivres AI Factory solutions present a paradigm shift enabling the transformation of raw data into actionable intelligence at scale. Meticulously integrated with NVIDIA AI infrastructure, Aivres AI Factory solutions seamlessly integrate cutting-edge compute, networking and software into a specialized turnkey infrastructure for enterprises and hyperscalers to accelerate insights and enhance business value.
NVIDIA GTC is the premier global AI conference, where developers, researchers, and business leaders come together to explore the next wave of AI innovation. From physical AI and AI factories to agentic AI and inference, GTC 2026 will showcase the breakthroughs shaping every industry. For more information on the event, please visit https://www.nvidia.com/gtc/.
About Aivres
Aivres is a data center servers and storage solutions provider committed to delivering innovative technologies that propel the world’s leading industries to new frontiers. Based in Silicon Valley, Aivres widely delivers and deploys its cutting-edge hardware products and designs to major data centers around the globe, supporting critical modern applications that include cloud, AI, deep learning, big data, mass storage and edge. For more information on Aivres, please visit https://aivres.com/.