Eviden unveils the BullSequana XH3500, a supercomputer that combines density, modularity, and energy efficiency

The BullSequana XH3500 is a next-generation supercomputing platform that merges power, hardware modularity, and energy efficiency. This generation marks a significant milestone for organizations seeking to integrate High-Performance Computing (HPC) with Artificial Intelligence (AI) without expanding their data center footprint. Designed as a sustainable technological foundation, the XH3500 addresses the growing demands of AI factories and rigorous scientific projects.

<p>The gradual unification of high-performance computing and AI workloads is transforming data center architectures. While the demand is surging, the available space and energy capacities are evolving at a much slower pace. In this constrained landscape, every advancement in density, cooling, or modularity becomes crucial to keeping up with innovation.</p>

<p>Eviden aims to tackle this challenge with the BullSequana XH3500, a system engineered to maximize computing capacity without inflating energy costs or multiplying infrastructure needs. This evolution reflects a broader trend where organizations are looking for platforms capable of accommodating AI accelerators, future HPC processors, and emerging technologies such as quantum computing, all while managing costs and carbon footprints without increasing the number of server rooms.</p>

<h4><strong>Notable Progress in Density and Energy Efficiency</strong></h4>

<p>The BullSequana XH3500 boasts a remarkable uptick in overall performance. Compared to the previous XH3000 platform, it offers over 80% more electrical power per square meter and 30% more cooling capacity per kilowatt, all without expanding its rack footprint. This design choice reflects a strong trend in the industry, where organizations favor denser environments to handle the demands of AI models and complex simulations.</p>

<p>Utilizing Eviden's fifth-generation hot water cooling system further reinforces this direction. Completely fanless, the system allows water circulation at 40°C, significantly reducing cooling needs and supporting heat reuse strategies. This combination contributes to sustainable total cost of ownership and improved energy resilience, essential for data centers heavily leveraged by large-scale AI.</p>

<p>This efficiency-focused approach also represents a quest for balance between power and frugality—two factors long deemed antagonistic. AI workloads, particularly the training of advanced models, require stable power and significant thermal management, often incompatible with aging infrastructure. By offering an exascale-ready platform, Eviden directly addresses this requirement. According to HPC division leaders, this generation prepares critical environments for uses that evolve faster than infrastructure cycles. The expected benefits span both raw performance and operational stability, bolstered by intelligent energy and resource management enabled through a fully integrated software ecosystem.</p>

<h4><strong>Modularity Designed to Embrace Technological Cycles</strong></h4>

<p>The strength of the BullSequana XH3500 lies in its open, modular architecture. Organizations can integrate different generations of processors, AI accelerators, and interconnects. This technological freedom, heralded as a cornerstone of Eviden's roadmap, mitigates dependency on a single vendor, securing investments across multiple technological cycles. HPC leaders emphasize that the platform is equipped to support both traditional simulation workloads and foundational model training, along with emerging workloads that blend HPC with quantum algorithms. Modularity also facilitates the diversification of use cases, a critical challenge as AI workloads proliferate faster than the available specialized human resources.</p>

<p>To support these varied applications, Eviden is investing in a comprehensive ecosystem of software and services covering installation, optimization, energy monitoring, and team support. This service dimension has become strategic in a market where technical leaders express a strong demand for out-of-the-box solutions. As Addison Snell, CEO of Intersect360 Research, points out, “users are now looking for integrated AI-HPC environments capable of scaling AI workloads while maintaining scientific rigor.” Organizations thus find a coherent framework to test, adjust, and industrialize their applications without layering additional complexity or relying on vendor-imposed roadmaps. This integration flexibility represents a crucial competitive advantage in an industry characterized by accelerating technological volatility.</p>

<h4><strong>A Platform Aligned with AI Factory Needs</strong></h4>

<p>The design of the BullSequana XH3500 directly addresses the new imperatives of AI factories—systems that integrate models, data, orchestration, and energy supervision into a cohesive whole. These environments demand not merely power, but also thermal stability, integration flexibility, and continuous evolution capability. The XH3500 serves as a pivot for deploying diverse environments ranging from foundational model training to digital simulation, including hybrid pipelines that combine scientific experimentation with generative AI. Eviden representatives emphasize that this convergence is redefining how critical workloads are orchestrated, enabling organizations to modularize their architecture according to industrial or regulatory needs.</p>

<p>The combination of a fully liquid cooling system, an open architecture, and an integrated software ecosystem creates a configuration particularly suited for environments facing high energy and operational pressures. Organizations gain a foundational platform capable of hosting, optimizing, and ensuring the reliability of AI applications, even as they evolve rapidly. This positioning also stands out in its consideration of future use cases, particularly the potential coupling of HPC with quantum technologies, which will require hybrid and agile infrastructures. The BullSequana XH3500 therefore solidifies its role as a strategic component for data centers tasked with balancing innovation, performance, and material efficiency.</p>
<div class="td-a-rec td-a-rec-id-content_bottom  td-rec-hide-on-p tdi_3_444 td_block_template_1">
<style>
/* custom css */
.tdi_3_444.td-a-rec-img{
                text-align: left;
            }.tdi_3_444.td-a-rec-img img{
                margin: 0 auto 0 0;
            }
</style></div>
Scroll to Top