Shared Everything

On this episode of the Shared Everything podcast from VAST, Nicole talks with Chris Powell, Quantum Information Science Lead at SAIC, about what “data” actually means in a quantum system. We move past storage and into representation, where problems have to be mapped into mathematical forms like Hamiltonian matrices and expressed directly into quantum hardware with no traditional memory or registers. We also delve into how hybrid architectures tie CPUs, GPUs, and QPUs together, why classical data infrastructure becomes more important rather than less so, and how quantum computing shifts the model from retrieving answers to shaping systems where answers emerge from probabilities.

In this episode of the Shared Everything podcast, Nicole Hemsoth Prickett speaks with Norm Marks, VP of Automotive at NVIDIA, about how autonomous systems are evolving from detection to prediction and now reasoning-driven AI. Marks explains how that shift is driving massive increases in GPU scale, synthetic data generation, and simulation, forcing companies to rethink infrastructure as training pipelines expand across hybrid datacenter and cloud environments. The result is a new class of AI factories built to train and operate autonomy at industrial scale.

In this episode of Shared Everything, we sit down with Jeff Denworth, co-founder of VAST Data, and Scott Shadley of Solidigm to unpack what’s actually driving the current flash supply crunch. The conversation moves from NAND physics and fab constraints to hyperscaler buying behavior and the sudden surge in AI-driven demand, explaining why this cycle is fundamentally different from past downturns. At the center is a clear reality check for anyone building AI infrastructure right now: flash has become a limiting resource, and software efficiency and architecture now determine how much usable capacity the industry really has.

On today's episode of the Shared Everything podcast, Nicole is live at SC25 with Dan Stanzione, Executive Director of the Texas Advanced Computing Center (TACC), for a look at why Horizon required a fundamental architectural reset. Stanzione explains how rising GPU power densities, liquid cooled 20 megawatt racks, and an increasingly irregular IO profile forced TACC to abandon long held assumptions about parallel filesystems. Years of watching billions of tiny files, unpredictable 4k and 64k reads, and metadata stalls slow entire machines led them to an all solid state tier and a VAST global namespace built for resilience, consistency, and shared access at scale. He describes how this model simplifies AI and hybrid scientific workflows, why the file system has always been the real point of failure, and how Horizon’s architecture reflects a world where IO, not FLOPS, determines what large scale science can do next.

In this episode Nicole talks to Jeff Denworth, Co-Founder of VAST Data, about the deep architectural shifts behind the 5.4 release, delving into how a new distributed runtime, native vector database, and event-driven compute layer transform VAST from a storage platform into a fully programmable AI operating system. Jeff explains how real-time vector inserts, parallelism without inter-node communication, and disaggregated shared-everything design make it possible to reason over data as it arrives, powering applications from Smart City analytics to trillion-scale AI pipelines.

On today’s episode of the Shared Everything podcast, Nicole talks to Jason Vallery, who just joined VAST after a 13-year career at Microsoft where he helped build the Azure cloud from the ground up. Jason reflects on the early days of object storage and cloud-native computing, when scaling from petabytes to exabytes redefined what infrastructure meant, and explains how lessons from Azure’s hyperscale era now shape his vision for VAST’s role in the AI age. He talks about the convergence of file and object systems, the evolution of AI storage built for thousands of GPUs, and the industry’s pivot from “data gravity” to a world where compute follows power and data must follow compute. Together, they trace how public cloud principles birthed the AI supercomputers of today and how the next wave of disaggregated, multi-cloud “neo clouds” will demand architectures that look a lot like what VAST is building

In this episode Nicole talks to Chris Powell, Chief Scientist at SAIC, and Kartik, Chief Scientist at VAST Data, about how the foundations of supercomputing are being rewritten by quantum advances, new architectures, and the collapse of distance between data and compute. Together they explore what happens when data becomes the environment of computation itself, how proximity and randomness define the next frontier, and why the systems of the future will think exactly where the information lives.

In this episode Nicole talks to Laurent Sifre, co-founder and CTO of H Company and former DeepMind scientist behind AlphaGo, AlphaFold, and Chinchilla. They explore how his discoveries in model scaling shaped the design of H Company’s computer-use agents, including the Surfer H platform that learns to navigate software through perception and action instead of APIs. The conversation dives into data infrastructure, distributed KV caching, sovereign compute, and why the future of automation depends on smaller, specialized models that act in the real digital world.

In this episode of the Shared Everything, reasoning models take center stage. No longer just text predictors, they now loop, branch, and drag in outside data, which blows open context windows and GPU limits. Alon Horev, CTO of VAST Data, unpacks how this shift strains infrastructure, while Kevin Deierling, SVP of Networking at NVIDIA, explains how NVIDIA Dynamo moves KV caches and workloads across GPUs, networks, and storage to keep agentic workflows moving. Data platforms become an extension of memory, enabling longer chains of thought, real-time agents, and secure, observable data paths. The result is a vivid picture of the AI datacenter as the nervous system for reasoning at scale.

On this episode, Nicole sits down with Danny McGinniss, VP of Product Management for Cisco Compute, Jacob Liberman, Director of Enterprise Product at NVIDIA, and John Mao, VP of Business Development and Alliances at VAST, to pull apart what it really means when three of the biggest forces in infrastructure line up behind the Cisco Secure AI Factory with NVIDIA, an architecture that brings together Cisco’s compute and networking, NVIDIA AI Data Platform, and VAST InsightEngine. The episode walks through the reimagining the datacenter as an AI factory where security, storage, speed, and data gravity collide to make enterprise AI real.

In this episode, Nicole speaks with Aaron Chaisson and Blake Golliher of VAST Data about how the company is reframing its mission for the AI era, centering on the idea of becoming the Operating System for AI. Aaron lays out the strategy behind this shift, while Blake—drawing on his deep background in building large-scale data platforms explains how the VAST SyncEngine enables customers to move and manage massive volumes of data across sites, clouds, and AI pipelines in real time. The discussion highlights why the ability to synchronize data at scale is critical for enterprise AI adoption, and how VAST’s approach marries technical architecture with business strategy to help organizations operationalize intelligence in ways traditional storage platforms never could.

In this episode, we explore how SK Telecom, South Korea’s largest wireless carrier and now a major force in AI infrastructure, joined forces with VAST Data to tackle one of the most difficult problems in large-scale computing: building a sovereign AI cloud that doesn’t compromise on speed, security, or scalability. Facing the nation’s mandate to keep AI models, data, and infrastructure fully under domestic control, SK Telecom had to rethink GPU virtualization from the ground up. The result is a platform that delivers near-bare-metal performance, strict multi-tenancy, and instant provisioning, setting a new standard for how sovereign AI infrastructure can be designed and operated.

In this episode, Nicole talks with Ken Patchett, VP of Datacenter Infrastructure at Lambda, about how hyperscale AI and sovereign LLMs are redefining datacenter and data management strategies. Ken highlights the challenge of data gravity, emphasizing the critical role of co-locating extensive storage infrastructure alongside ultra-high-density compute to support increasingly data-hungry workloads. He outlines Lambda’s "aggregated edge" model, designed for regional deployment of inference and enterprise workloads, enabling localized data processing and compliance with global sovereignty and privacy regulations. The conversation also addresses how these changes demand adaptive multi-density infrastructure, integrating flexible compute-storage designs that accommodate shifting hardware requirements and evolving regulatory landscapes.

Stacks of federal reports tell countless stories of IT investments gone sideways, yet the stakes have never been higher as artificial intelligence reshapes government. David Hinchman, Director of IT and Cybersecurity at the Government Accountability Office (GAO), joins Shared Everything to dissect why federal technology initiatives often falter and how these invisible fault lines could dangerously widen in the age of AI.  From planning pitfalls to hidden infrastructure challenges, Hinchman reveals the critical decisions that determine whether AI becomes government’s greatest tool...or its most costly failure.

In this episode of Shared Everything, Glenn Lockwood, just named Principal Technical Strategist at VAST Data, shares what decades at the bleeding edge of large-scale systems design have taught him about architecting for an AI future that refuses to stay put. From building the first all-NVMe 30PB Lustre file system to designing Azure’s training clusters, Glenn walks us through why performance alone is no longer enough, why inferencing shattered traditional supercomputing data patterns, and why today’s infrastructure decisions must be guided not by legacy conservatism, but by intrinsic flexibility. With characteristic clarity and conviction, Glenn lays out the case for treating adaptability as a first-class citizen in architecture...and why VAST is where he’s chosen to do just that.

00:00 – 02:00Intro and Kartik’s background in physics and industry evolution; early days of personalized medicine and genomic research.02:00 – 04:30Breakdown of three main advances in life sciences: genomics, gene editing (CRISPR), and long-read sequencing technologies like Oxford Nanopore and PacBio.04:30 – 06:30Deep technical dive into nanopore sequencing: how it works, why it matters, and why it requires GPU acceleration.06:30 – 08:30The computational bottleneck: memory mapping, random I/O, why short-read sequencers are now limiting, and why SSDs are necessary.08:30 – 10:00Parallel file systems break under modern life sciences loads; shift toward storage architectures that can handle random I/O at scale.10:00 – 12:30How AlphaFold reshaped structural biology and compute expectations; protein folding as a graph neural network challenge.12:30 – 15:00LLMs in pharma, managing clinical trial data, and the rise of mixed, hybrid workloads in research computing.15:00 – 17:00Microscopy at scale (cryo-EM, light sheet imaging) and the data tsunami—petabytes per microscope, per year.17:00 – 19:30Shifting away from HPC-era assumptions: new workloads, new storage expectations, and lessons from vendors like Oxford Nanopore.19:30 – 20:36What’s next: generative AI models trained on molecular sequences and protein structures; a vision of disease-free future.

In this episode from GTC Paris, Nicole chats with Andy Pernsteiner, Global Field CTO at VAST Data, for an insider's view of Europe's quickly evolving AI landscape. From the sidelines of Nvidia’s packed event—strategically co-located with Viva Tech—Andy shares candid insights on the shift from heavy-lifting AI infrastructure builds toward practical, profitable services built on those investments. Sovereign clouds take center stage, as Andy unpacks Europe's increasing emphasis on secure, traceable, and auditable data architectures designed around tight regulatory frameworks and national boundaries. He offers a close look at how European Neo-cloud providers are pushing toward innovative services like inference-as-a-service, making AI genuinely consumable and not just an expensive science project.

Podcast Timeline: Dan Stanzione (TACC) & Don Schulte00:00–02:07Introduction by Nicole; guests Dan Stanzione (Executive Director, TACC) and Don Schulte (VAST Data).02:08–03:51Reflections on TACC’s history, reputation for innovation, and pioneering adoption of new technologies.03:52–05:57Discussing dramatic shifts in HPC due to increased emphasis on power consumption, driven by the end of Dennard scaling.05:58–08:37Recent explosion of AI workload demands; increased costs and shortages (GPUs, skilled personnel, power infrastructure).08:38–12:53Speculation on future HPC developments: potential impacts of photonics, quantum computing, carbon-free energy sources, and changes in AI scaling strategies.12:54–18:20Dan emphasizes the importance of foundational HPC research historically done at national labs and universities, highlighting that current AI and infrastructure innovations rely heavily on these early HPC breakthroughs.18:21–21:49Introduction of Horizon, TACC’s upcoming NSF-funded supercomputer, replacing the Frontera system, focusing on scientific throughput, GPU optimization, and extensive solid-state storage.21:50–27:57Detailed discussion on the NSF’s Leadership Class Computing Facility (LCCF) award that supports Horizon, emphasizing scientific outcomes over raw computing power.Horizon system designed specifically for real-time data assimilation, persistent interactive services, and complex scientific workflows, enabling significant improvements in science productivity.27:58–30:36Shift from batch-oriented computing to interactive, real-time workflows and persistent data management.Importance of new data platforms (like VAST) providing consistent, high-performance data access across diverse computing tasks.30:37–34:47Stanzione emphasizes new data access patterns: smaller, random, constant I/O operations, challenging traditional HPC storage assumptions. Highlights VAST’s platform role in addressing these new storage needs effectively.34:48–36:33Closing remarks on the dramatic evolution in HPC data management over the past decade, noting fundamental shifts that were not anticipated even ten years ago.

In this conversation with host Nicole Hemsoth Prickett, experts Wes Brewer (Oak Ridge National Lab) and Adrian Jackson (EPCC) discuss the evolving landscape of digital twins, particularly in the context of supercomputing and AI. They explore the distinctions between digital twins and traditional simulations, the real-world applications of digital twins, and the infrastructure challenges faced in their implementation. The discussion also delves into the potential of distributed digital twins and the role of AI in enhancing digital twin workflows, concluding with insights on future workloads and applications.

00:00 – IntroductionNicole introduces Jeff Denworth, reminiscing about the Big Data era (~2010–2014).01:15 – Big Data to Big MetadataJeff reflects on the Big Data era (Hadoop, analytics, NoSQL).Today's valuations (Snowflake, Databricks) suggest Big Data's continued relevance.02:11 – The Rise of Big MetadataJeff describes the shift from Big Data to Big Metadata.AI creates new categories and applications, rapidly driving data infrastructure demands.Example: Nvidia’s rapid growth due to deep learning-driven workloads.05:01 – Synthetic Data and Metadata ExplosionJeff notes social networks using synthetic data to circumvent privacy regulations.Metadata types include large-scale data catalogs to manage exabytes of data (e.g., OpenAI).08:14 – Dynamic Data CatalogsVAST Database as an example of transactional and analytical infrastructure.Benefits of SQL queries replacing traditional file operations for faster data handling.09:50 – Metadata Evolves with VectorsExplanation of embeddings, vector databases, and similarity search.AI-driven understanding of unstructured data via vectors.11:56 – Massive Scale of Vector DatabasesRough estimate: ~40 trillion vectors per 100 petabytes of data.Challenges with conventional vector databases at massive scale (cost, memory, speed).13:22 – Future Scale Problems and AI-driven Data EngineeringRetrieval-Augmented Generation (RAG) increases vector database scale needs.Nvidia's data flywheel concept accelerates embedding and data engineering automation.15:48 – Predicting Infrastructure Needs (Two-Year Outlook)Jeff predicts AI models will significantly improve data engineering within two years.Enterprises need vector databases capable of transactional, real-time performance.18:10 – Future-Proofing Infrastructure (Five-Year Outlook)Jeff expects AI-driven automation to impact all business processes (factories, back-office).Businesses must be prepared for rapid scaling and foundational AI-driven changes.21:14 – Industries Leading the AI Infrastructure RaceAI adoption speed varies by industry—highest "fury" is in software development.Banks and trading firms leverage AI differently: profit efficiency vs. alpha-seeking.23:55 – Cloud vs. On-Premises Infrastructure ChoicesJeff sees hybrid approaches prevailing; decision-making depends on enterprise-specific needs.Introduces idea of "agentic workforce" prompted by Jensen Huang's statement (100M AI agents).24:31 – Agent Ownership and Future ConsequencesRaises profound questions about ownership and management of AI agents in business.Jeff notes limited current customer recognition of these deeper implications.25:56 – Closing RemarksNicole and Jeff conclude by noting broad societal implications of AI-driven changes.Emphasis on importance of continued discussions around big metadata.

Jonathan Koomey—renowned researcher and the mind behind Koomey’s Law—joins to explore how AI is testing the limits of power infrastructure and computing efficiency. We dig into the history of data center energy use, the real implications of synthetic data, and why renewables—not nuclear—may be the only scalable path forward. Koomey offers a candid look at the rebound effect, cost allocation in the grid, and why assuming infinite AI demand is a dangerous bet. If you care about the future of energy in the age of AI, this one’s for you.

This isn’t about faster servers or more storage; it’s about designing infrastructure for a new species of intelligent entities. In this episode, we explore how VAST Data is setting the stage for the AI-native cloud, where inference workloads take center stage, and agents form the nervous system of a new computational world.