70th IT Press Tour: Hammerspace says AI's real bottleneck is data, not GPUs. At the 70th edition of The IT Press Tour in Palo Alto, Hammerspace laid out new customer wins, a plain-spoken case against copy-and-migrate AI storage, and a teaser for a batch of announcements coming at Supercomputing in November. If you walked into Hammerspace's session at the 70th IT Press Tour expecting a polished pitch about yet another "AI storage" box, you walked into the wrong room. Floyd Christofferson, the company's VP of Product Marketing, spent his time explaining the opposite: that the last thing most enterprises need is a new pile of storage. What they need, he said, is to stop copying data around long enough to put the data they already own to work. And he came with receipts. Christofferson walked us through a planned ~30 PB Hammerspace deployment in AWS at a major hyperscale AI research lab, a European healthcare AI project headed from 128 GPUs to more than 1,000, a 144-node GPU cluster on Oracle Cloud Infrastructure (OCI) for a global communications company, and a university that cut storage costs by roughly 48% by reclaiming NVMe it didn't realize it was wasting. He also made it clear that more is on the way: Hammerspace will be in attendance at SC26, the Supercomputing event taking place in Chicago this November. The 70 percent problem nobody puts on a slide. Christofferson opened with a story from the Storage Developers Conference in Santa Clara the week before. The CTO of the Berkshire Hathaway companies, he said, told the room something most vendors avoid saying out loud: in the enterprise, roughly 70% of AI projects fail. Not at the proof of concept. Those go great, because the data is curated, the scope is small, and everyone is in a good mood. They fail on the way to production. Why? He gave three reasons, and anyone reading this who has run a data center will nod along. Companies can't justify the cost of duplicating infrastructure. They can't stomach the risk and complexity of migrating or copying data. And they have existing operations that can't be interrupted just because someone in the business wants to try an agent. That CTO, by his account, also delivered a lecture to the vendors in the audience: don't bring me a net new storage system for AI when I have a business to run and you can't tell me the return on investment. Here's the thing: the industry's default answer is exactly that. Buy a shiny AI-specific platform, copy the data you want to expose, and you're good. Christofferson ran through what that does to data sovereignty, to cost, and to the investments already made in existing storage. His presentation put it more bluntly: AI isn't compute-constrained, it's data-constrained. The file system is the trap. To understand what Hammerspace does differently, you have to accept a premise that sounds almost too simple. When you buy storage, you are also buying a file system, and that file system is welded to the hardware. Fill it up and you get a second one, then a third, then ten. Add a cloud region and you've got another. The industry has spent thirty years building tiering tools, hierarchical storage managers, gateways, and migration utilities to paper over that one fact. Hammerspace's answer is to lift the file system out of the infrastructure entirely. Users and applications see one orchestrated global namespace through NFS, SMB, S3, or a CSI driver for containers, while the data underneath sits wherever it makes sense on whatever storage is already in place. Nothing gets installed on the storage itself, and no agents run on it. Standard Linux, no special client. This is where Hammerspace's story gets technical, and where I think it gets interesting. For roughly eight or nine years, the company has been pushing parallel file system capabilities into the mainline Linux kernel: NFSv4.2 with Flex Files (going back to 2019), attribute delegations, fast failover, LOCALIO, NFS Direct, and more. Christofferson noted that Hammerspace's CTO is the Linux NFS client kernel maintainer, and that other maintainers are on staff as well. The payoff is a line he delivered with a notable smile on his face: Hammerspace's client "is already installed in every data center on the planet." It's a little tongue-in-cheek, but the point is real. Any standard Linux distribution already has the pieces, so there's nothing proprietary to roll out to GPU servers, whether they sit in your data center, a neocloud, or a hyperscaler. He was also honest about the hard parts. The pNFS N-to-1 capability, the thing classic HPC needs when hundreds of thousands of cores write to a single file, has to go through the IETF and the Linux community. "Like trying to solve peace in the Middle East," he said. AI workloads, by contrast, lean on many-to-many access, which the current kernel work already handles. Anvil, DSX, and data that moves while you're using it. Architecturally, there are two pieces in one software load. The Anvil metadata servers hold the file system metadata, essentially the brains. The DSX (data services) nodes present protocol access, move data, and can even act as data stores. Because the metadata is separate from the data, applications talk to the metadata server and then read in parallel. Because Hammerspace is the file system, data can move between storage systems, from slow to fast, or from one site to another, while users are working with it. Christofferson pointed to live data mobility, a capability the company introduced into pNFS. A policy engine (Hammerspace calls the policies "objectives," a desired state rather than a script) decides where each file's actual contents live, based on performance, cost, or protection level. "We're a new file system, but migrations go away. The migration for Hammerspace is changing a mount point." - Floyd Christofferson That migration claim is the one I'd poke at in a trial, and it's the one the customer stories lean on. Tier 0: the Storage you already paid for. If one idea carried the afternoon, it was Tier 0. Every GPU server ships with local NVMe. In NVIDIA's terminology that's G3. On a single node it's fine. Across a hundred or a thousand nodes, it's a scattering of isolated pockets that are unshared, largely unprotected, and mostly idle. As NVMe densities grow, that means a growing pile of stranded capacity. Hammerspace's approach is to fold that local NVMe into the same global namespace as every other tier and manage it with policy. Christofferson's math on speed was straightforward: local NVMe is about 2.5 times faster than a 400 Gb network, and in the cloud, where network bandwidth is far more modest, the comparison slide showed networked options running from about 10x to as much as 62x slower than local access. On top of that, you already own the media. His observation was that filling the empty NVMe slots in an off-the-shelf GPU server costs less than buying external networked flash and upgrading the network to feed it. Vanderbilt University: 10 petabytes nobody knew they had. Vanderbilt's ACCRE facility is the cleanest example. It supports around 2,750 researchers across nine schools and colleges and some 350 campus departments, runs a 750-node compute cluster with 80 GPU nodes, and was juggling roughly 40 PB across multiple systems. The university had a budget problem and a goal: cut storage costs and move to a more composable model built on commodity hardware. Hammerspace came in and unified the lot. The result was a roughly 10 PB cluster that includes Tier 0 storage built from GPU-server NVMe, 500 TB of Tier 1, and petabytes of archive, all under one namespace with automated orchestration. Christofferson said the team reclaimed close to 10 PB of local NVMe, so the Tier 1 SSDs they'd planned to buy never got bought. Storage costs dropped by about 48%. What I liked most was the less obvious win. Researchers on grant-funded projects can now be provisioned storage that fits their work without anyone touching the underlying infrastructure or disturbing other users. A global communications company on OCI. A similar pattern played out at a global communications technology company. It needed to push on-premises data into Oracle Cloud to reach GPUs there. The deployment is a 2 PB Hammerspace cluster with two Anvil metadata servers and six DSX nodes, with Tier 0 across a 144-node GPU cluster. Data is processed on Tier 0, checkpoints flow down to OCI Object Storage for long-term retention, and applications talk NFSv4.2 pNFS as well as NFSv3. His summary: fastest time-to-value, and lower cost because the customer didn't have to buy as much high-performance capacity inside OCI. More receipts: cloud labs, neoclouds, and a Glacier escape. The remaining examples show how far the same architecture stretches. I've kept some details light because a few of these customers aren't being named yet. A hyperscale AI research lab leaves Lustre behind. This one is the headliner for anyone who runs GPU clouds. The lab had ongoing operational headaches with Amazon FSx for Lustre and needed to move workloads between on-premises and multiple clouds. Its researchers also prefer NFS to S3. The plan is roughly 30 PB across multiple Hammerspace clusters in AWS, with an extension into the customer's own data center, creating a "data fabric" across public cloud and on-prem. Christofferson said Hammerspace beat FSx for Lustre in a head-to-head test. In the AWS pilot, checkpoint loads ran 36% faster and large-file operations ran 3x faster. Hammerspace's S3 connector presents an NFS mount point to users over object storage in the cloud, so the researchers never see the plumbing, and they can shift workloads across GPUs and even clouds without noticing. A three-site neocloud and a life sciences escape from Glacier. A neocloud offering inference-as-a-service needed a three-site global namespace for a client. Each site gets two metadata servers and two DSX nodes, inference clusters mount Hammerspace over NFS in each region, and data moves between sites through a shared object bucket. Metadata sync is asynchronous but lands within seconds, which means a colleague in Croatia, Singapore, or Los Angeles sees the same file system. Then there's a US life sciences research organization serving researchers across 26 labs. It had 5 PB of aging PowerScale and 2 PB in AWS Glacier, with archive retrievals taking 12 hours or more. With Hammerspace, the organization is shifting its data toward Glacier Instant Retrieval, staging it by policy instead of by hand, and retiring the old storage without taking anyone offline. The three clocks of AI economics. Christofferson's most memorable framing was what Hammerspace calls the three clocks of AI economics. Everyone in the industry talks about time to first token. He argued that's the wrong clock to start with, because getting a proof of concept to answer quickly is a very different problem than keeping a production pipeline fed. * Time to Very First Token. How long until the AI environment is ready at all? Traditional routes mean buying new storage (3 to 6+ months), deploying proprietary clients, and finishing bulk migration before anyone can use anything. Hammerspace's pitch is to assimilate existing data in place, with standard clients and zero migration. The company claims over a 10x reduction in time to AI-ready data compared with siloed AI storage. * Time to First Token. Once a request lands, how long does it wait for data? Policy-driven placement stages data ahead of demand, and parallel access speeds reads when the workload needs them. * Time per Token. While the pipeline runs, are the GPUs fed? This is where Tier 0 and orchestration earn their keep. That third clock is where the numbers get uncomfortable. Hammerspace's slides put average GPU usage today at around 18%, against 87% with intelligent orchestration, and claim that every 10% improvement is the equivalent of 10,000 GPUs' worth of value. Those are the company's figures and your mileage will vary. Christofferson's own version was simpler: if your GPU is 30% busy, 70% of your investment is burning a hole in your wallet. "The AI race won't be won by the company with the most GPUs. It will be won by the company with the fewest idle GPUs." - Hammerspace "It's not a new layer," and other fair questions. The IT Press Tour crowd is not shy, and a few of the exchanges were worth the price of admission. One journalist pushed on the obvious objection: aren't you just adding another layer to my infrastructure? Christofferson's answer was one of the better ones I've heard. Yes, it's a file system, and a new one. But customers already pass through multiple file systems today, plus a stack of point tools for migration, tiering, and gateways, plus a human running scripts overnight. Hammerspace consolidates all of that into one automated layer. "Sure, it's a layer, but it's not a new layer. It's replacing a manual, complex existing layer that was already there." - Floyd Christofferson The question of protection came up too. Hammerspace isn't a backup product, a migration tool, or a cloud gateway, he said, but it does all of those jobs as a side effect of managing data. Policies can require, for instance, that data carrying a certain metadata tag be protected to nine nines, automatically replicated, and audited. Global snapshots across NetApp, Isilon, and object storage can be stored wherever is cheapest. "We're not storage, but we do storage things," he said. "We're not data management, but we do data management things." Then there's the supply crunch. With SSDs and memory pricey and hard to get, customers are rediscovering hybrid HDD and SSD designs. Christofferson's line: tiering is back, "except it's not tiering like your grandpa's tiering. It's active data orchestration." On competition, he was relaxed about recent moves by NetApp into parallel NFS. If anything, he said, it validates pNFS as the parallel file system for AI, since the client is already in the kernel. NetApp has offered pNFS since 2011, but with NFSv4.1 rather than v4.2 with Flex Files, and he argued that a client capability alone isn't the point. Orchestration, multi-site data services, and native protocol bridging all have to be engineered into the platform. His claim: Hammerspace has a 10-year engineering head start and five or six years of shipping products. (That's Hammerspace's view, of course, and I'd expect the competition to disagree.) He also mentioned that Los Alamos' Gary Grider has been looking at pNFS with Hammerspace orchestration to reduce dependency on Lustre. Why this briefing stuck with me. Christofferson has been wrestling with the same problem for decades: a post-production facility in New York running SGI machines, then Pathfire, where he built a system that beamed syndicated TV content by satellite to a thousand stations, then SGI itself. Different decade, same headache: data trapped in silos, humans moving it by hand. His take is that AI finally makes the headache too expensive to ignore, because nobody is going to duplicate an entire IT estate to run one application. He is also candid that Hammerspace didn't predict AI. The company simply spent nearly a decade building a standards-based, orchestrated file system, and then the market arrived at its door. Hammerspace did work on training with Meta's Llama 2 and Llama 3 efforts, but the pitch now is aimed at the harder, less glamorous problem: inference and agentic workflows running on the messy enterprise data that already exists, including 40-year-old seismic archives. One engineer at a hyperscale customer reportedly called the technology "black voodoo magic." Christofferson's reply: there's science behind it. If you run infrastructure for AI, here's my advice. Keep an eye out for Hammerspace around SC26. They have a few announcements that will be released around the event. Christofferson hinted that the next layer is about connecting to AI pipelines and making deployment simpler. In the meantime, ask yourself a blunt question about your own environment: how much NVMe is sitting idle in your GPU servers, how many copies of your data exist, and how many months stand between your first pilot and production? If the answers make you wince, Hammerspace is worth a call.

70th IT Press Tour: Hammerspace says production AI dies in the data layer, and its fix already ships in your Linux kernel. Alex Zaharov-Reutt, Global AI and Technology Editor | Published 13 Oct 2026 70% of enterprise AI projects sail through the proof of concept and then quietly die before production. Hammerspace reckons it knows why, and it argues the cure is sitting on storage you've already paid for. I was at the 70th IT Press Tour in Redwood City, California, on 8 October 2026, where Hammerspace VP of product marketing, Floyd Christofferson, gave the assembled journalists a fast-moving tour of the company's technology, its customers and where it thinks enterprise AI is heading. Christofferson opened by placing his job "somewhere in the middle" of product and marketing. Sitting at the end of the table was Hammerspace VP of products and operations, Douglas Fallstrom, who I was told is one of the company's founding team. Christofferson described Fallstrom's daily life as "wrangling the engineers who are building this stuff into the product", which sounds about right for anyone running products at a storage company. The full video presentation is below, upon which the article that continues below is based upon, please watch, and read on! The 70% problem Christofferson had flown in from an HPC user forum in Germany, where he said research AI adoption is running hot (around 70% utilisation) on carefully curated data sets. Before that he'd been at the SNIA Storage Developer Conference in Santa Clara. Enterprise is a different beast. At SDC, Christofferson heard the CTO of the Berkshire Hathaway group of companies (an SGI alumna, which Christofferson, himself ex-SGI, clearly considered a strong reference) tell a room of technologists that roughly 70% of enterprise AI projects fail. They succeed at the POC stage. They fail on the way to production. Her reasons, as relayed by Christofferson, were simple enough. Companies can't justify duplicating infrastructure, they can't justify the risk of migrating or copying data, and they can't interrupt the business operations already running on that data. According to Christofferson, she lectured vendors about pitching her a separate new storage system for AI. "I've got existing business operations. I need to be able to run my existing business," she said, as he recounted it. That, he argued, is the gap Hammerspace was built to fill, years before anyone was talking about AI factories. Why the old model breaks The traditional enterprise IT model organises data by application. You buy storage for a workload, the data starts hot, ages down a hierarchy and eventually lands in an archive. Economics decide where it lives. AI scrambles that. "AI means that the data is not necessarily aligned to the primary application that was creating it," Christofferson said. An AI pipeline doing weld inspection on a manufacturing line, drug discovery imaging or seismic analysis on 40-year-old survey data needs to reach across every silo, site and cloud. "If it's in this silo over here, or that silo over there, how does the AI see it?" he asked. The industry's standard answer, Christofferson said, is a shiny new AI storage platform and an instruction to copy everything into it, which raises awkward questions about sovereignty and duplicated cost. The file system, pulled up and out Hammerspace's core argument is that the file system is trapped inside storage hardware. Buy a NAS and you're buying its file system. Fill it up and now you've got two file systems, then three, then ten, plus a stack of point products for tiering, migration, cloud gateways and data protection to paper over the gaps. Worse, for extreme performance many parallel file systems need a proprietary client installed on every compute node. When the AI factory sits in a neocloud or somewhere else entirely, that gets ugly fast. "Do I have to distribute custom clients to everybody?" Christofferson asked. Hammerspace's answer has been to push the performance features it needs into the standard Linux kernel, through the IETF and the Linux community, over the past eight or nine years. The company's CTO, Trond Myklebust, is the maintainer of the Linux kernel NFS client, and Christofferson said other kernel maintainers are on staff, with another joining the week of the briefing. The slide timeline ran from parallel NFS v4.2 with Flex Files (back in 2019) through LOCALIO and striped Flex Files to NFS Direct, with pNFS N-to-1 and client-side erasure coding still to come. Getting code through that process is slow. Christofferson compared it to "trying to solve peace in the Middle East. It takes a long time, and there's a lot of diplomatic effort." The payoff is the line he uses at every conference. "Our client is already installed in every data centre on the planet." If you run a standard Linux distribution, the plumbing is already there. One namespace, many protocols On top of that sits what Hammerspace calls an orchestrated global namespace. Every user and application sees the same file system, whether it comes in via NFS, SMB, S3 or a CSI driver for containers. The software has two parts in one load: Anvil metadata servers (the brains, where the file system metadata lives) and DSX data services nodes, which handle protocol access, move data and can store it too. Because Hammerspace owns the metadata, it can move the bytes between systems, sites and clouds while users keep working, using live data mobility features it contributed to pNFS. The company calls its policies "objectives", essentially a desired-state engine that can be triggered by job schedulers, user actions or metadata tags. Data on tape or deep archive still obeys physics, Christofferson conceded, but for any scheduled pipeline you'd be staging that data anyway. He also made a point that matters for anyone who's lived through a weekend data migration (anyone who's done one will wince). "The migration for a Hammerspace is changing a mount point. And then it's finished." Customers doing the heavy lifting The case studies were where the claims got concrete. Vanderbilt University's Advanced Computing Center for Research and Education (ACCRE) supports 2,750 researchers across nine schools and 350 departments, running a 750-node cluster with 80 GPU nodes and around 40PB of storage spread over multiple systems. They needed more high-performance storage and had no budget for it - a common experience for everyone on planet Earth wanting to buy SSDs in our memory and storage hungry AI age. Hammerspace built a 10PB cluster spanning Tier 0 (the NVMe inside the GPU servers), 500TB of Tier 1 SSD and petabytes of archive, all in one namespace. The slide claimed a roughly 48% reduction in storage costs. Christofferson said Vanderbilt reclaimed close to 10PB of local NVMe that had been stranded inside individual servers. ACCRE executive director of research computing operations, Hunter Hagewood, said at the time of the original announcement that Hammerspace gave the centre a "composable, open architecture that lets us unify GPU server-local, tiered and archival storage". A global communications technology company (one we "might actually be using right now", Christofferson teased) needed on-premises NAS data available to GPUs in Oracle Cloud Infrastructure, so it deployed a 2PB Hammerspace cluster in OCI with Tier 0 across a 144-node GPU cluster, with checkpoints tiered down to OCI Object Storage. A major hyperscale AI research lab is planning around 30PB across multiple Hammerspace clusters in AWS to replace FSx for Lustre, extending into its own data centre. Its researchers wanted NFS mount points, which they find easier to work with than S3. The AWS pilot showed checkpoint loads 36% faster and large file operations three times faster, according to Hammerspace's slides. Christofferson also shared a quote from one of that lab's engineers, who apparently called the result "black voodoo magic". (Arthur C Clarke would've nodded knowingly.) A European healthcare AI company, whose name is embargoed until SC26, is building an AI clinical assistant for oncology using imaging, records and clinical data. It runs a data lake in Azure while reaching GPUs at neoclouds Nebul and Nebius, growing from 128 GPUs to more than 1,000. And a US life sciences research organisation with 5PB on ageing Dell PowerScale and 2PB in AWS Glacier (with retrievals of 12 hours or more) is flipping that ratio, heading towards about 2PB on-premises and 5PB in Glacier Instant Retrieval, with automated orchestration replacing manual copies. Christofferson noted that old arrays coming off expensive support contracts often get a second life as tier-two or tier-three storage. "The storage still works," he said. Tier 0 and the network maths The most striking slide compared writing a 500GB checkpoint to local NVMe in an eight-drive GPU server against networked storage. Local NVMe at 112GB/s did it in about 4.5 seconds. A 400Gb/s network took around 11.1 seconds, making it 2.4 times slower, and that's a network few enterprises have deployed end to end. In the cloud, the gap widens dramatically. Hammerspace's figures showed AWS EBS gp3 at around 265 seconds (59 times slower), Azure Premium SSD v2 at about 278 seconds (62 times slower) and OCI Block Volume at roughly 207 seconds (46.4 times slower). "Clouds don't have fast networks, right?" Christofferson said, noting he'd updated the cloud bandwidth numbers on the plane the night before. NVIDIA's own nomenclature calls server-local NVMe "G3" and networked flash "G4". Hammerspace's pitch is that if you make G3 part of a shared namespace, you need less of the expensive G4 tier. Christofferson went further: it's cheaper to fill the empty NVMe slots in a half-populated GPU server than to buy networked flash and upgrade the network to feed it. (These are vendor numbers, of course. Independent benchmarks would be welcome, and I suspect Hammerspace would welcome them too, given its MLPerf Storage and IO500 submissions.) The three clocks of AI economics Christofferson's framing for the business case was "the three clocks of AI economics". Clock one is time to very first token: how long it takes to get an AI environment ready at all. With traditional approaches you scope, buy, install, migrate, configure and only then reach a pilot. Hammerspace's slides claimed procurement alone runs three to six months or more, with proprietary client rollouts and bulk migrations adding weeks or months. Hammerspace claims more than a 10-fold reduction in the time to AI-ready data. Clock two is time to first token, the familiar inference metric, where policy-driven placement can stage data ahead of demand. Clock three is time per token, keeping GPUs productive while pipelines run. The utilisation slide was the attention-grabber: 18% average GPU utilisation today versus a possible 87% with intelligent orchestration. "If your GPU is 30% utilised, you've got 70% of your investment just burning a hole in your wallet," Christofferson said. Tiering returns, with a twist I asked about the memory and SSD pricing crunch, which has made flash both expensive and hard to get this year, as I noted above with Vanderbilt. Christofferson pointed to a hybrid HDD and SSD deployment that only reached production because it didn't need to be all-flash. "We thought tiering was dead and had to always buy an all-flash array," he said. "Tiering is back." Hammerspace prefers the term active data orchestration. (I suggested tiering is "steering", and Christofferson jokingly offered to put me in touch with marketing so they could hire me.) NetApp, PEAK:AIO and the pNFS bandwagon The timing of this briefing is interesting. On 25 September, NetApp announced it would acquire PEAK:AIO, the Manchester-based company behind pNFS Lattice, an open-source scale-out metadata server for NFS v4.2 and Flex Files developed with Los Alamos National Laboratory and Carnegie Mellon University. Days later at NetApp INSIGHT, the company unveiled Novus, its AI data architecture that separates metadata from data. Asked by IT Press Tour founder Philippe Nicolas whether this was competition, Christofferson said no. "If anything, what we see is a validation that pNFS is really the parallel file system for AI." He argued that a client is only part of the story, and that orchestration, data services, multi-site and native protocol bridging are where the hard engineering lives. "We've got a 10-year advantage in engineering over this, and five or six years in actually shipping products," he said, while describing NetApp as "a very capable technical organisation". The broader competitive field includes VAST Data, WEKA, DDN with Lustre, Dell, Pure Storage and the hyperscalers' own managed file services. Christofferson also mentioned that Los Alamos National Laboratory's Gary Grider is looking at pNFS with Hammerspace orchestration to reduce dependence on Lustre. Momentum and money Hammerspace has plenty of wind in its sails. In April 2026, the company said its 2026 bookings to date were nearly 14 times those of all of 2025, with gross retention above 95% and a Net Promoter Score of 71 (company-reported figures, it's worth keeping in mind). That followed a US$100 million Series B in April 2025 led by Altimeter Capital with ARK Invest, valuing the company at more than US$500 million. In February 2026, South Korea's SK Square invested in Hammerspace through its overseas arm, TGC Square. Hammerspace founder and CEO, David Flynn, previously founded Fusion-io, which SanDisk bought for US$1.1 billion. In the April release, Flynn said "AI has moved beyond model building". What it means for Australia When I asked about Australian customers, Christofferson said one had just started in Australia and another in New Zealand, though he couldn't yet say more about their status. Hammerspace has an office in Singapore with people in Korea, China, the Philippines and India, and it's expanding. Christofferson had just returned from Korea's supercomputing show and urged us to pay attention to the AI work happening there. There's a local thread in Christofferson's own story, too. He was previously CMO at Melbourne's Arcitecta, after stints at SGI, a New York post-production house and Pathfire, where he helped build an IP-over-satellite system for around 1,000 US TV stations. "I've been playing this game for 30 years, and it's the same problem in different clothes in every organisation," he said. For Australian enterprises weighing AI against data sovereignty obligations, flash shortages and the eye-watering price of GPUs, the Hammerspace argument lands well. Metadata tags can keep PII and sovereign data out of training sets and away from offshore clouds, which Christofferson said is taken far more seriously in Europe than the US (and I'd argue in Australia too). What's next Hammerspace is promising new announcements at SC26 in November, with press pre-briefings the week of 9 November. Christofferson wouldn't say much, beyond describing it as "the next layer in connecting to AI pipelines" and making them simpler to deploy. I think the core pitch is a strong one. Most organisations have data scattered across NetApp, Dell, cloud buckets and forgotten NVMe, and the idea of plugging AI into all of it, in place, with clients that are already installed, is far more palatable than another forklift upgrade. The vendor benchmarks need independent testing, and NetApp's pNFS push means Hammerspace's head start will be tested. But if the 70% failure rate Christofferson quoted is anywhere near right, the company is pointing at the real bottleneck. As Christofferson put it, "it's really a data economics question."

Hammerspace to showcase how enterprises can dramatically improve AI infrastructure efficiency at RAISE Summit 2026. Hammerspace Jul 06, 2026, 08:00 ET Hammerspace to demonstrate how data orchestration improves AI economics by accelerating time-to-value, reducing time-to-first-token and keeping GPUs continuously productive. REDWOOD CITY, Calif., July 6, 2026 /PRNewswire-PRWeb/ - Hammerspace, the high-performance data platform for AI anywhere, today announced its senior executive team will participate in RAISE Summit 2026, taking place July 8-9 at the Carrousel du Louvre in Paris. At Booth 3B, Hammerspace will demonstrate how enterprises can win the AI infrastructure race before the first token by eliminating one of AI's largest hidden constraints: data readiness. As enterprise inference and RAG deployments accelerate, AI economics are not defined by GPU counts or storage performance alone. AI outcomes depend on how quickly environments become productive, how rapidly workloads gain access to the right data, and how efficiently infrastructure keeps GPUs generating useful output. The Hammerspace Data Platform addresses these challenges by unifying the data estate on existing storage and automating data orchestration to make data available where GPUs need it without large-scale data migration or new storage procurement. At RAISE Summit, attendees will learn how Hammerspace improves three metrics that determine AI success: * Time to Very First Token: Reducing the time required to make enterprise data AI-ready so projects can begin in days instead of months. * Time to First Token: Orchestrating and pre-positioning data so AI workloads can begin generating results faster. * Time per Token: Continuously supplying GPUs with the right data to improve utilization and lower the effective cost of AI operations. "AI leadership has been shaped by model innovation, infrastructure scale and access to data," said David Flynn, founder and CEO of Hammerspace. "The next challenge is efficiency: how quickly organizations can activate their own enterprise data and turn it into useful AI output without unnecessary delay, cost or operational complexity." Traditional AI deployment models often treat AI readiness as an infrastructure project, requiring organizations to acquire new storage, build new environments and migrate data before productive work can begin. Hammerspace's Data Platform instead treats AI readiness as a data availability challenge, enabling organizations to use data in place across existing storage systems and cloud environments while creating a unified global namespace for AI workloads. By continuously discovering, orchestrating and positioning data where AI workloads need it, Hammerspace reduces idle GPU cycles, minimizes unnecessary data movement and helps organizations improve productive GPU utilization while lowering cost per token. Hammerspace's Data Platform enables organizations to: * Activate enterprise data in place across existing storage systems and cloud environments * Eliminate storage procurement and migration delays that traditionally postpone AI initiatives * Unify distributed data through a global namespace without creating additional silos * Pre-position data before workloads require it, reducing time to first token * Continuously orchestrate active data throughout AI pipelines to maximize GPU productivity Organizations using traditional storage-first AI data approaches may spend between 14 and 30 weeks acquiring infrastructure and migrating data before productive AI work begins. Hammerspace can reduce that timeline to as little as a few days to one week, representing acceleration of up to 70x. "Hammerspace changes the operating model for enterprise AI by making data available where AI workloads need it, without forcing organizations into months of storage procurement, migration and manual data movement," Flynn added. "That means teams can start faster, use existing infrastructure more efficiently, and keep AI pipelines supplied with the data required to produce results." Executive Speakers at RAISE Summit David Flynn will present on the Main Stage on July 8, sharing how organizations can rethink AI infrastructure around the economics that matter most: speed to token generation, continuous GPU productivity and cost-efficient AI operations. Molly Presley, Hammerspace SVP of Global Marketing, will present on July 9 on the Grace Hopper Stage in a session titled "The Data Problem: What AI Actually Runs On," examining why data readiness, not infrastructure procurement, has become the defining challenge for enterprise AI. About Hammerspace Hammerspace is the high-performance data platform built to simplify and optimize AI infrastructure at scale. It makes all your data immediately accessible - anywhere across on-premises and cloud environments - without copying or migrating data into new silos. By integrating with existing storage, networking, and applications, Hammerspace creates a unified, high-speed data backbone for AI, enabling organizations to accelerate every stage of the AI pipeline while eliminating data silos. Media Contact Kim Pegnato, Hammerspace, 1 781-835-7118, [email protected], www.hammerspace.com SOURCE Hammerspace