Source: Authority Magazine By Ken Claffey, CEO, VDURA , July 2, 2026 – Ken Claffey is CEO of VDURA, a modern data storage infrastructure software company purpose-built for AI and HPC workloads. With a track record of building and scaling businesses across the HPC and storage ecosystem, he has held senior executive roles at Seagate, Xyratex, Adaptec, and Eurologic. At Seagate, Ken led the Enterprise Storage division through a period of transformative growth, driving full-stack product innovation to address emerging customer needs across Cloud and Enterprise markets. At Xyratex, he built the ClusterStor HPC storage business from the ground up — a platform that went on to power 40% of the world’s top supercomputers before being acquired by Cray/HPE. He was also instrumental in the subsequent sale of Xyratex to Seagate, and held leadership roles spanning product, operations, sales, and engineering at Adaptec and Eurologic. As CEO of VDURA, Ken brings the strategic insight and hands-on expertise to translate vision into results.
Thank you so much for joining us in this interview series! Before we dig in, our readers would like to get to know you a bit more. Can you tell us a bit about your “backstory”? What led you to this particular career path?
I’ve spent my career in the storage and HPC world: Eurologic, Adaptec, Xyratex, Seagate and now VDURA. What drew me to this space was the same thing that keeps me here: it’s foundational infrastructure. The decisions made at the storage layer determine whether a system works or fails. At Xyratex I had the chance to build the ClusterStor HPC storage business from scratch, which ultimately powered 40% of the top supercomputers in the world. That experience of building something from nothing, with real consequences if it didn’t perform, shaped how I think about everything. When the opportunity came to lead VDURA — a company built on 25 years of parallel file system innovation, at precisely the moment AI was transforming what infrastructure needed to do — it felt like the right moment to take everything I’d learned and put it to work.
Can you tell our readers what it is about the work you’re doing that’s disruptive?
This industry spent a decade telling organizations that all-flash storage was the answer, that flash prices would keep falling and tiering was a legacy concern. We disagreed, and we were willing to say so publicly. VDURA is built on a mixed-fleet architecture: the right amount of NVMe flash to saturate GPU throughput, with high-density HDD for everything else, unified in a single namespace. That’s not a compromise — it’s the architecture the hyperscalers actually use. Google’s Colossus, Meta’s storage backbone, Microsoft Azure, none of them are all-flash. They run software-defined, mixed-fleet systems because they understand that flash is a performance medium, not a capacity medium. We’re bringing that architectural truth to enterprise AI and HPC customers who’ve been oversold on a model that was always economically fragile.
In today’s parlance, being disruptive is usually a positive adjective. But is disrupting always good? When do we say the converse, that a system or structure has ‘withstood the test of time’? Can you articulate to our readers when disrupting an industry is positive, and when disrupting an industry is ‘not so positive’? Can you share some examples of what you mean?
Disruption is positive when it solves a real problem that incumbents have stopped trying to solve, usually because their business model depends on the status quo. The shift from general-purpose to purpose-built storage for AI is a good example. Legacy vendors had strong incentives not to rebuild their platforms from scratch, so the gap between what customers needed and what was available kept growing. That’s the right moment to disrupt. Disruption becomes destructive when it prioritizes novelty over reliability, or speed to market over the needs of the people who depend on the system. In HPC and AI, downtime has real consequences — research delayed, clinical trials disrupted, national programs stalled. The parallel file system, for instance, is a structure that has genuinely stood the test of time precisely because the underlying problem — delivering consistent, high-throughput data access at scale — hasn’t changed. We didn’t abandon PanFS when we became VDURA; we rebuilt around its proven core. That’s a meaningful distinction.
Can you please share 5 ideas one needs to shake up their industry? If you can, please share a story or an example for each.
1. Treat storage as a first-class infrastructure decision, not an afterthought. AI projects are routinely scoped around compute and network, with storage bolted on at the end. The bottleneck is almost always the data layer. Organizations that design storage into the architecture from day one move faster and spend less.
2. Stop anchoring on flash price as a fixed variable. The all-flash era was built on an assumption that NAND costs would keep falling indefinitely. That assumption has broken. The organizations that built flexibility into their architectures are now in a much stronger position than those that standardised on a single media type.
3. Demand that infrastructure becomes more reliable at scale, not less. Most storage systems degrade in ways that are subtle but compounding as you add nodes. We designed VDURA to do the opposite, resilience should improve as a system grows. Hold your vendors to that standard.
4. Take TCO seriously, including energy and operational cost. A system’s cost isn’t just the purchase price. Power, cooling, management overhead, and the cost of unplanned downtime are all real. The right metric is cost per usable TB per workload, measured over five years, not sticker price per raw TB.
5. Don’t let the AI hype obscure the infrastructure reality. GPU clusters without the data infrastructure to feed them are expensive underutilisation. The question isn’t just “how many GPUs?” — it’s “how do you ensure those GPUs are never waiting for data?”
We are sure you aren’t done. How are you going to shake things up next?
The next phase is proving that mixed-fleet, software-defined storage isn’t just theoretically sound, it’s measurably better in production AI environments, at scale, over time. We’re building the reference architectures, the benchmarks, and the customer evidence to make that case conclusively. The Flash Relief Program is one expression of that: we’re willing to put a direct comparison in front of any customer running all-flash and let the economics speak.
Do you have a book, podcast, or talk that’s had a deep impact on your thinking? Can you share a story with us? Can you explain why it was so resonant with you?
Jensen Huang’s appearance on the BG2 Pod was one I found myself returning to. His framing of the end of general-purpose computing — the idea that domain-specific architecture is now the defining question for the entire industry — landed clearly. It aligned with something we were already building toward at VDURA, but hearing it articulated that directly by someone with his vantage point was clarifying. It sharpened the conviction that purpose-built infrastructure isn’t a niche bet; it’s where the industry is going.
Can you please give us your favorite “Life Lesson Quote”? Can you share how that was relevant to you in your life?
“Progress happens one rep at a time.” It’s something I’ve taken from powerlifting and applied everywhere. Compound improvement — in the gym, in a product, in a team — is rarely dramatic in the moment. The days when nothing feels like it’s moving are usually the days that matter most. The discipline to show up and do the work regardless of how it feels in the short term is what separates durable results from temporary ones.
You are a person of great influence. If you could inspire a movement that would bring the most amount of good to the most amount of people, what would that be? You never know what your idea can trigger. 🙂
I’d want to make the data infrastructure that powers scientific research genuinely accessible to institutions that can’t afford hyperscaler pricing. Some of the most important work in climate modelling, genomics, and materials science is being done at universities and research labs that are constrained not by ideas or talent, but by what they can afford to store and process. If we could close that gap — give researchers the same data infrastructure capabilities that large commercial operators take for granted — the impact on human knowledge could be extraordinary.