Research Finder
Find by Keyword
Is Hammerspace Correct That Data Gravity No Longer Matters in Today’s AI Architecture?
New performance, locality-aware data paths, and multi-cloud integration position Hammerspace to challenge long-held assumptions about AI data pipelines.
Key Highlights:
Hammerspace v5.2 delivers major AI and HPC performance gains, including a 33.7 percent higher IO500 score, doubled bandwidth and an 800 percent improvement in IOR-Hard-Read.
New Tier 0 Affinitization adds automatic locality awareness, directing I/O to the requesting GPU node’s local NVMe to reduce east-west latency and improve predictability.
Continued upstream Linux/NFS enhancements provide performance improvements without proprietary client software or changes to underlying storage.
Share Referrals extend metadata scale-out for linear namespace growth across extremely large AI and HPC file counts.
New Oracle Cloud Infrastructure support, along with Kerberos and Labeled NFS, expands multi-cloud reach and strengthens enterprise-grade security.
The News
At SC25 in St. Louis, Hammerspace introduced version 5.2 of its Data Platform, delivering performance, scalability, and multi-cloud enhancements aimed squarely at modern AI and HPC environments. The company published updated IO500 benchmark results showing a 33.7 percent improvement over the prior version, including doubled bandwidth and an 800-plus percent gain in IOR-Hard-Read. These results continue the company’s trajectory of optimizing parallel NFS for AI-centric throughput.
A central driver of this improvement remains Hammerspace’s ongoing contribution to upstream Linux kernel and NFS client enhancements, allowing organizations to benefit from performance gains without proprietary client software or new storage infrastructure.
Version 5.2 introduces Tier 0 Affinitization, which automatically detects when a GPU or compute node hosts local NVMe and arranges pNFS layouts so reads and writes land on that node’s Tier 0 volume. This reduces east-west saturation, improves performance consistency and eliminates the manual configuration steps required in earlier iterations.
To support extreme-scale AI and HPC datasets, v5.2 adds Share Referrals, enabling the global namespace to distribute across multiple metadata servers. This ensures linear scalability even as file counts reach into the hundreds of billions. Security enhancements include Kerberos authentication and Labeled NFS, allowing SELinux and other MAC systems to carry and enforce security labels across NFS workloads.
The release also expands Hammerspace’s cloud integration story with new support for Oracle Cloud Infrastructure (OCI), including new shapes and planned Dedicated Region compatibility. Together with existing integrations across AWS and Azure, Hammerspace v5.2 presents a consistent, multi-cloud namespace without copying data into new silos or disrupting existing workflows. For more details, read the Hammerspace press release.
Analyst Take
Hammerspace is making a credible argument that the industry’s long-standing assumptions about data gravity deserve to be re-examined. In AI environments dominated by GPU availability, power density, and distributed compute footprints, the locus of gravity is shifting toward the compute, not the data. Version 5.2 strengthens this position by combining standards-based parallelism with intelligent data placement and transparent multi-cloud orchestration.
AI performance is increasingly defined not just by FLOPS, but by dataflow efficiency and the ability to keep GPUs fed without waiting on storage. Hammerspace’s approach to data locality directly targets this challenge by minimizing time-to-first-token, a metric that now defines competitiveness for both training and inference. By aligning data paths dynamically with GPU topology, the platform enables model pipelines to sustain higher throughput without re-architecting applications or introducing proprietary middleware.
Tier 0 Affinitization is the most strategic addition in this release. As GPU clusters grow more geographically distributed and operationally constrained, data locality becomes one of the most important determinants of AI performance. Automating locality-aware data routing so GPUs consistently read and write to their own NVMe storage is a meaningful step toward predictable throughput at scale.
Hammerspace’s continued investment in upstream kernel innovation reinforces its core differentiation. By improving performance through the Linux kernel rather than through proprietary agents, the company minimizes operational drag and maximizes compatibility. Enterprises will increasingly value these two traits as they operationalize AI pipelines. Support for OCI, alongside AWS and Azure, signals that multi-cloud GPU sourcing is becoming standard practice for AI-first organizations. A data platform that can follow compute resources across facilities, clouds, and availability domains is well-positioned for the next phase of AI infrastructure expansion.
The rise of agentic AI workloads intensifies the need for a platform that can route intelligence to data in real time. Hammerspace’s assertion that “data gravity no longer matters” captures this inversion: the data layer is becoming fluid, moving with the compute wherever intelligent processes are deployed. If proven at scale, this could mark a genuine architectural shift where storage ceases to be the bottleneck and becomes the enabler of continuous, location-independent AI.
In our opinion, Hammerspace is no longer a traditional file system player. It is evolving into a broader AI data platform that unifies namespaces, orchestrates data movement without copies, and establishes consistent governance across complex hybrid environments. This evolution aligns with what enterprises are now asking for: outcomes, not tools.
Looking Ahead
We believe the next phase of AI infrastructure will be defined by a set of architectural choices that are already creating clear lines of separation between hyperscalers, neocloud providers, and enterprise IT teams. Each group is approaching the problem from a different angle, but the underlying challenge is the same: how to build an AI environment where data is available with the performance, locality, and governance required for continuous AI operations.
Hyperscalers are optimizing for GPU density and global reach, betting that customers will bring data to their clouds and live inside their managed ecosystems. Neoclouds, by contrast, prioritize raw performance, local NVMe, and east-west efficiency to keep GPUs fully utilized, often sidestepping the deeper, layered network hierarchies of traditional cloud regions. Enterprises sit between these worlds, facing the growing complexity of hybrid estates while operating within cost, sovereignty, and staffing constraints.
In our opinion, these approaches will collide as organizations mature beyond early AI experiments. Future AI workloads will not reside in a single place, nor will they tolerate copies, silos, or multi-hop data paths. They will require a data platform that can follow the compute, whether that compute lives in hyperscale regions, neocloud GPU clusters, on-premise datacenters, or all three.
This is where Hammerspace’s strategy becomes relevant. The company’s emphasis on standards-based parallelism, accelerating time-to-first-token, locality-aware performance, and transparent multi-cloud namespace mobility positions it as a potential enabler in this emerging landscape. Enterprises evaluating their next moves will increasingly need to decide whether they adopt a vertically integrated AI stack from a single provider or deploy a data platform that can fluidly span the heterogeneous environments where their GPUs will live.
We find that Hammerspace can improve its overall competitiveness and ecosystem influence by doubling down on its strategic alignment with the compute-first needs of distributed AI and HPC environments. Over the next 12 months, the company should aggressively market the Tier 0 Affinitization feature as the critical answer to data locality, emphasizing its ability to automate predictable, low-latency performance essential for maximizing GPU utilization across multi-cloud and hybrid environments.
By showcasing the massive performance gains (e.g., the 33.7% higher IO500 score and 800% IOR-Hard-Read improvement) achieved through standards-based parallelism and upstream Linux/NFS enhancements, Hammerspace can minimize the perceived operational risk for enterprises. The company must position itself not merely as a file system, but as the AI Data Platform that unifies namespaces and orchestrates data movement without copying data, directly addressing the demand for consistent governance across heterogeneous estates.
To further augment its ecosystem influence, we advocate that Hammerspace must continue its multi-cloud expansion and strategic integration to become the default data mobility layer for AI-first organizations. The recent support for OCI should be leveraged to target organizations diversifying their GPU sourcing, demonstrating that the platform can fluidly follow compute resources wherever they reside - be it AWS, Azure, OCI, neoclouds, or on-premises data centers.
Furthermore, Hammerspace should solidify its influence by increasing its contribution to upstream kernel development to reinforce its differentiation and expand compatibility, simultaneously reducing the operational drag for customers. By strategically engaging enterprises that are evaluating whether to adopt a vertically integrated AI stack or a data platform that spans diverse environments, Hammerspace can position its solution as the necessary, flexible layer that breaks the constraints of traditional data gravity and enables complex, continuous AI operations.
HyperFRAME will be watching how Hammerspace navigates these architectural battles. In our view, the choices vendors and customers make over the next 12 to 18 months will determine which platforms become central to AI operations, and which ones remain tied to the constraints of yesterday’s data gravity assumptions.
Don Gentile | Analyst-in-Residence -- Storage & Data Resiliency
Don Gentile brings three decades of experience turning complex enterprise technologies into clear, differentiated narratives that drive competitive relevance and market leadership. He has helped shape iconic infrastructure platforms including IBM z16 and z17 mainframes, HPE ProLiant servers, and HPE GreenLake — guiding strategies that connect technology innovation with customer needs and fast-moving market dynamics.
His current focus spans flash storage, storage area networking, hyperconverged infrastructure (HCI), software-defined storage (SDS), hybrid cloud storage, Ceph/open source, cyber resiliency, and emerging models for integrating AI workloads across storage and compute. By applying deep knowledge of infrastructure technologies with proven skills in positioning, content strategy, and thought leadership, Don helps vendors sharpen their story, differentiate their offerings, and achieve stronger competitive standing across business, media, and technical audiences.
Stephanie Walter | Practice Leader - AI Stack
Stephanie Walter is a results-driven technology executive and analyst in residence with over 20 years leading innovation in Cloud, SaaS, Middleware, Data, and AI. She has guided product life cycles from concept to go-to-market in both senior roles at IBM and fractional executive capacities, blending engineering expertise with business strategy and market insights. From software engineering and architecture to executive product management, Stephanie has driven large-scale transformations, developed technical talent, and solved complex challenges across startup, growth-stage, and enterprise environments.
Ron Westfall | VP and Practice Leader for Infrastructure and Networking
Ron Westfall is a prominent analyst figure in technology and business transformation. Recognized as a Top 20 Analyst by AR Insights and a Tech Target contributor, his insights are featured in major media such as CNBC, Schwab Network, and NMG Media.
His expertise covers transformative fields such as Hybrid Cloud, AI Networking, Security Infrastructure, Edge Cloud Computing, Wireline/Wireless Connectivity, and 5G-IoT. Ron bridges the gap between C-suite strategic goals and the practical needs of end users and partners, driving technology ROI for leading organizations.