Empowering Innovation
with Solidigm

The race to build AI infrastructure is defined, in most coverage, by megawatts announced, leases signed, and capital deployed. What gets less attention is whether any of it actually gets built on time. Atif Ansar, co-founder of Foresight Works, has spent 15 years studying why large projects fail, including through academic work at Oxford University. In my recent conversation with him and Solidigm’s Jeniece Wnorowski, Atif shared an insight from that work that shapes his company’s direction: the limiting factor in the AI data center boom is not technological, but human.

Foresight Works offers a project delivery platform combining AI, proprietary data, and scheduling methodologies developed through Atif’s academic research. The platform is built around the idea that projects are human systems, and that basic psychological biases that can harm outcomes can be overcome with processes that ensure both micro daily commitments and larger goals are met.

The Real Bubble Risk

In discussion about a potential AI infrastructure bubble, the concern usually voiced is about overbuilding. Will the supply of new data center capacity outpace demand? Atif rejects that framing. He points to accelerating demand for AI-enabled services and the contractual structure of hyperscaler leases as evidence that demand is not the problem. The problem, he argues, is delivery.

“These data centers are being built under contract from very high rated companies like hyperscalers,” he explained. “And as a result, they have 15-year leases. So the developers are not taking huge amount of risk.” But that contractual protection cuts both ways. A single month of delay on a 100-megawatt facility can cost $15 million or more in lost early revenue, and service level agreement penalties from hyperscalers compound the damage further. Atif estimates that more than a three-month delay can dissipate the net present value of even a very large build. “The penalties for missing your targets become potentially insolvency causing for companies that are not well managed,” he said.

The Watermelon Problem

One of the more memorable concepts Atif brought to this conversation was what he called the watermelon problem, a phrase for projects that are “green from the outside until they suddenly turn red.” Atif noted that the gap between perceived and actual progress is typically a product of optimism bias and political pressure, not malice. Teams hope they will catch up. But often, they do not.

The interdependencies inside a data center build make this particularly dangerous. A purchase order that should have been issued today, but was not, means a missing piece of equipment six months from now. The delay is invisible until the moment it becomes a crisis. “What lets people down is the accumulation of these very small variances,” he noted. “It’s actually those micro details that make the difference.”

Foresight is designed to function as a control tower, allowing organizations to manage the macro picture while having a handle on micro variances early enough to act on them. In one case, the platform uncovered a five-month delay buried beneath distorted reporting.

Where Schedules Break Down

Atif has identified recurring patterns in how project schedules fail. One is an overemphasis on civil and structural work, which is visually dramatic and easy to represent, at the expense of mechanical, electrical, and plumbing work, which is complex and harder to schedule. “Nearly 60 to 80% of a data center is MEP,” he noted. Treating it in broad strokes rather than mapping it with precision is a reliable path to delay. A second pattern involves compressing commissioning into the final phase of a build, when factory and site acceptance tests should begin far earlier, alongside procurement and equipment delivery.

Foresight’s AI layer helps identify structural gaps in submitted schedules, particularly missing milestones or poorly sequenced dependencies. “We help them look at what milestones they are likely missing or what gates they’re likely missing and help them insert those back in the appropriate places,” Atif said.

A Skills Supply Crisis Beneath the Capital Crisis

The execution problem extends well beyond data centers. Digital infrastructure, the energy transition, and defense spending are all generating enormous project delivery demands simultaneously on top of existing demand from evergreen sectors like pharmaceuticals and civil infrastructure, and the construction workforce is not keeping pace. Atif estimates that data centers represent roughly 10% of the global construction market by value, yet the people working in the industry amount to just 0.05% of the global construction labor force.

“So it’s still a cottage industry in terms of the footprint,” he said. “They need a lot of technology, automation, and AI to simply keep pace. They also need education…and I think that we need to upskill people and train them in the art of becoming better project managers.”

The TechArena Take

While day-to-day execution doesn’t typically garner headlines, Atif's work makes a compelling case that it deserves far more attention from executives committing capital to AI infrastructure. The trillion-dollar build underway globally requires disciplined upfront planning, clear governance, and repeatable processes. The developers who get this right are those who invest in process maturity early, building the kind of credibility that holds up with communities, investors, and hyperscalers alike. Platforms like Foresight Works represent an important step toward making that discipline accessible at scale, at a moment when the cost of getting it wrong compounds with every month of delay.

To learn more, listen to the full podcast or visit foresight.works.

Sustainability, when it even enters the conversation for data centers today, means different things to different people. It is a broad topic, including energy use, water use, e-waste, and the effects of rapid growth. Gabriel Lazar, head of sustainability at Submer, a  Barcelona-based full-stack data center company, sees a need for both greater specificity and more comprehensive systems thinking as part of organizational strategies.

In a recent discussion with Gabriel and Solidigm’s Ace Stryker, Gabriel laid out a pragmatic framework for thinking about data center impact: one that spans energy, community relationships, and the uncomfortable question of whether the industry is using what it builds.

The Cost of Growth Without Durability

The data center industry is struggling with sustainability, and Gabriel is direct about it. About half of data center operators don’t even have sustainability plans in place. Many companies that do have them are still moving in the wrong direction. Part of the difficulty is the recent push for higher performance and scale hasn’t considered what will make what is being built now durable for years to come.

What the industry needs, Gabriel argued, is a genuine reduction in resource intensity, pursued with enough specificity to actually be achieved. He expressed optimism that data centers and digital infrastructure can continue to have a smaller environmental footprint than heavy industries like steel making or car manufacturing. . Getting there requires separating growth from impact, finding ways to scale the industry while bending the resource curve downward. “We need to start seeing that graph divide, with impact going down and the growth going up.”

Thinking in Systems, Not Silos

One reason sustainability conversations fall short, Gabriel argued, is that they tend to stay within the boundaries of the sustainability field itself. The forces shaping data center infrastructure today, including geopolitical risk, supply chain scarcity, energy market structures, and community relations, do not respect those boundaries. Addressing them requires a multi-disciplinary approach that borrows from fields with more practice at managing complexity.

“If we were to pick a field that is best at doing this right now, it’s probably risk, and more specifically, if you look at insurance companies, they’re amazing at capturing this because they have so much data and they’re able to crunch it,” he said. Having people from different backgrounds collaborating and communicating is the best path to addressing the extremely complex, interconnected challenges such as power shortages that the industry is facing today.

Heat Reuse: Valuable, but Not Universal

Gabriel leads a heat reuse workstream for the Open Compute Project, so his cautious take on the topic may come as a surprise. The logic of heat reuse is straightforward: nearly all the energy flowing into a data center converts to heat, and using that heat for other purposes rather than dissipating it is sensible in principle. In practice, though, the economics depend heavily on geography and local infrastructure.

District heating networks, common in Germany and other northern European countries, make heat reuse viable at scale. In Spain, where Submer is headquartered, or in the United States, those networks largely do not exist. Trying to mandate heat reuse uniformly across markets will produce uneven and often poor results.

“It might just be 20% of data centers worldwide that are applicable,” Gabriel said. “I think that makes more sense than trying to hit it all and not getting anything.”

He flagged desalination and water treatment as underexamined applications for heat reuse, noting their round-the-clock demand profile. The constraint, again, is proximity: only data centers sited near coastlines or water treatment facilities would benefit. The pattern across all of these examples is the same: broad ambitions need to be matched with specific, locally grounded analysis before they translate into real outcomes.

The Community Dimension

Beyond grid dynamics and thermodynamics, Gabriel emphasized that sustainable infrastructure is also a social proposition, and one that requires the same place-specific thinking. Heat reuse, for all its complexity, offers something rare in the industry: a tangible benefit that local communities can see and understand. Unlike grid flexibility programs or ancillary services markets, a data center supplying heat to a nearby manufacturer or supporting a local business is a concrete value proposition to the people who live nearby.

That matters for the permitting process. “It goes a long way for that regulatory process, the buy-in, the social license to operate,” Gabriel said. Operators who engage communities early, rather than as an afterthought or not at all, tend to move faster through approval and run into fewer delays.  This practice also helps with future projects as a reputation builder. The TechArena Take

Gabriel’s argument is ultimately a call for better engineering of the sustainability problem itself. Sustainability as a vague aspiration produces vague results, if any. Sustainability as a set of specific, locally calibrated initiatives focused on relevant topics  can produce real outcomes. Infrastructure should not be considered in isolation, especially not when it plays such a crucial role for societies. In the race to build data centers, systems and longer-term thinking should be key. To learn more about Submer’s work, watch our full podcast or visit submer.com.

As general manager of MiTAC Computing, Raymond Huang is building the case that validated, pre-integrated infrastructure is the best path to AI at scale. His company, long known as a manufacturer of high-performance compute systems, has evolved into a provider of end-to-end AI infrastructure solutions. In a recent TechArena Data Insights episode, Solidigm’s Jeniece Wnorowski and I talked to Raymond about the strategy, the partnerships, and the technical decisions driving that evolution.

The Evolution to Delivering Turnkey AI Capacity

MiTAC’s transformation has been driven by three converging pressures: the technical demands of modern AI workloads, economic pressure on margins and differentiation, and a customer base that increasingly wants faster time to deployment. In response, the company is developing turnkey AI infrastructure solutions built around pre-integrated rack-level systems in which servers, GPUs, networking, power distribution, and cooling are validated together before they ever reach a customer site.

“The AI factory is a new category we're working on,” Raymond said. “MiTAC is aligned around the modular AI cluster as a building block. We don’t just build servers; we deliver AI capacity.”

That positioning matters because the traditional model, in which customers sourced components separately and spent months integrating them, is increasingly untenable at the pace AI demands. Raymond noted that MiTAC’s approach compresses that timeline significantly, moving customers from months of integration to days or weeks of installation.

Solving the Density, Power, and Cooling Equation

AI clusters are constrained by how tightly GPUs can be packed without hitting thermal or power ceilings. MiTAC addresses this through pre-designed high-density rack configurations validated for specific GPU setups, ranging from 32 air-cooled GPUs per rack to 96 GPUs per rack with liquid cooling. This eliminates the trial-and-error phase that often accompanies high-density deployments.

Power delivery is equally central to the approach. Modern GPU racks can draw easily from 30 kilowatts up to 130 kilowatts, and instability in that power supply creates downstream problems that are expensive to diagnose and fix. Because MiTAC controls both the design and manufacturing of its systems, it can pre-match power profiles to specific GPU and CPU configurations from the ground up. That end-to-end ownership allows the company to engineer across voltage options including 208V, 415V, 480V, and potentially the upcoming 800V DC standard.

On cooling, the company has developed direct liquid cooling systems capable of removing up to 95 percent of generated heat via a liquid loop. Raymond pointed to MiTAC’s G4826Z5, a 4U system built around dual AMD EPYC 9005 CPUs and AMD Instinct MI355X GPUs, as a demonstration of what liquid cooling makes possible.

“Without the DLC, this kind of density wouldn’t even be thermally practical,” Raymond said. “The system isn’t performance throttling even during long AI training runs. So the system sustains peak GPU utilization instead of cycling down due to heat. This is critical to the large language model training or HPC simulation multi-AI workload jobs.”

A Partner Ecosystem Built Around the Full Stack

In our Data Insights series, we often discuss how as AI infrastructure becomes more data intensive, application performance relies on compute and storage working together optimally. Raymond described a set of partnerships that MiTAC has developed to address the layers of the AI stack to enable high performance. For example, with Solidigm and DDN, MiTAC has built a storage solution pairing its servers with Solidigm NVMe drives and DDN software to eliminate the I/O bottlenecks that can leave expensive GPU clusters sitting idle waiting for data.

For orchestration, MiTAC has partnered with Rafay, whose managed Kubernetes platform and GPU orchestration tools simplify cluster management across multi-node environments. The combination allows customers to go from power-on to a usable cluster significantly faster, with centralized policy management that scales consistently from a single rack to 500.

Another collaboration reflects MiTAC’s commitment to reducing the energy footprint of AI infrastructure: its work with Akash Systems using diamond-based cooling. Because diamond conducts heat at roughly five times the rate of copper, the resulting systems consistently run up to 10 degrees Celsius cooler than standard configurations, delivering measurably better performance per watt without increasing energy consumption.

Manufacturing Localization and the Road Ahead

MiTAC is also expanding its North American manufacturing footprint to meet demand for localized supply chains. Raymond described the company’s capacity as several thousand racks per month, with a focus on building, fulfilling, and supporting AI infrastructure locally in each region it serves.

Looking a few years out, Raymond sees data center design converging around liquid cooling as the universal standard, rack densities climbing toward 100 kilowatts to potentially one megawatt per rack, and hybrid on-site power generation becoming common. The organizing principle Raymond returned to is modular, repeatable architecture.

“Think of it like Lego blocks for AI capacity,” he said. “This is extremely critical for deploy to speed needs.”

The TechArena Take

MiTAC’s evolution to an AI infrastructure provider reflects a broader maturation in how the market thinks about deploying AI at scale. Validated, pre-integrated solutions that compress deployment timelines and reduce operational risk are becoming increasingly crucial. Raymond’s message to technology decision makers is straightforward: the competitive question is how quickly and reliably you can put your chosen infrastructure to work. MiTAC understands that customers who face that pressure will value solutions that arrive ready to run.

For more information, listen to the full podcast episode, or visit MiTACcomputing.com.

There’s a long-held idea in enterprise technology that people who make safe, conventional procurement decisions stay employed. Today, that logic is playing out as procurement teams and IT leaders choose dominant global cloud platforms not because they’ve done a thorough evaluation, but because it feels like the default.

Solidigm’s Ace Stryker and I recently sat down with Albane Bruyas, chief operating officer of Scaleway, who made a compelling case that the “safe” choice may come with unexpected risks, and that having a backup plan is critical. With 25 years of cloud experience behind it, Scaleway is working to demonstrate that a homegrown European complement to the global hyperscalers is essential for organizations that care about control, cost transparency, and long-term resilience.

More Than Sovereign: The Case for Full-Stack Control

Scaleway describes itself as both a sovereign European cloud provider and a neocloud built for AI infrastructure. For Albane, however, the word “sovereign” only gets you so far.

“What is very interesting for our clients is that we sell full autonomy,” she said. “We have control on all the value chain. This is what makes the pure difference for our clients.”

That control extends across software, hardware, network transit, and pricing. Scaleway operates within a major French telecommunications conglomerate, which gives the company ownership over connectivity that most cloud providers must outsource. The result, Albane argues, is a level of transparency and pricing stability that organizations cannot get from vendors who depend on third-party components throughout their stack.

Sustainability is also woven into this argument. Because Scaleway built its data centers, primarily in France, it has been able to optimize for energy efficiency. That includes tracking and reducing carbon footprint at the hardware level, a capability made more precise by the company’s participation in the Open Compute Project.

A Procurement Argument Rooted in Risk

Before joining Scaleway, Bruyas worked in industrial procurement, and she brings that lens directly to conversations with enterprise buyers. In industry, she notes, you always ensure you have at least two suppliers for physical goods. Her message is straightforward: operating differently with your digital suppliers doesn’t eliminate risk. It invites it.

“In the digital world, people are just forgetting that their principal strategic supplier is a unique supplier they cannot move out from,” she said. “There is no solution if there is a crash. There is no solution if different prices come out. No solution if there is an external government that asks for data. You have no choice.”

Her advice to organizations anchored to a single hyperscaler is practical, not confrontational: “You will never be fired because you choose one of the scalers. So it’s just like, test an alternative. You need to have one, and you will be happily surprised,” she said.

In a similar spirit, Scaleway’s active participation in the Open Compute Project is not simply a technical preference; it is a supply chain hedge. By building on open hardware standards, the company can source components from multiple vendors, reducing dependence on any single manufacturer and creating competitive pricing leverage.

“If you have the most open hardware you can, then you have the capacity to buy from different suppliers,” Albane explained. “If you have the capacity to buy from different suppliers, you can have a better price, and you can have more capacity because you can go to different places.”

Building for an Agentic AI World

As AI workloads shift from simple inference queries toward longer, more complex agentic workflows, the infrastructure requirements change substantially. More context, more memory, tighter latency budgets, and greater demand for diverse compute options are all part of that transition.

Scaleway’s strategy is, once again, to embrace sourcing from different providers. The company offers multiple GPU types, is actively collaborating with next-generation chip designers, and maintains capacity for CPU-based inference where appropriate. Albane noted that Scaleway has a history of being early to emerging architectures, having been among the first providers to offer Arm-based servers.

“We want to continue to be really at this level of technology where we can put something in place that nobody has,” she said.

The TechArena Take

Scaleway occupies an unusual position in the cloud market: mature enough to offer a full public cloud catalog, yet structured in a way that gives it operational visibility and pricing control that most providers lack. Albane makes a credible case that full-stack ownership is not just a differentiator in the marketing sense but a concrete operational advantage, particularly for organizations that need cost predictability, data residency assurance, and a genuine second-source option. Enterprise buyers that prioritize control and transparency can benefit by making a strategic choice to diversify their cloud strategy before vendor dependency becomes a liability.

Hitesh Kumar explores AI infrastructure design, GPU clusters, networking, storage, cooling, and scalability. Learn how modern AI systems are evolving beyond GPUs to full-stack architecture.

Enterprise storage has long had a potential pain point with growing unstructured data, but most organizations have tolerated the ballooning storage needs of their documents, video files, images, sensor outputs, project archives, and more because the cost of inaction seemed manageable.

That calculus has changed. The surge in AI workloads has triggered an acute shortage of high-performance memory chips, the same components that underpin enterprise storage hardware. As AI infrastructure competes for that supply, vendors have passed the cost along quickly, leaving organizations to urgently search for new solutions.

Recently I sat down with Solidigm’s Jeniece Wnorowski and Krishna Subramanian, co-founder and chief operating officer of Komprise, about how enterprises can better handle unstructured data, which makes up roughly 90% of all data generated worldwide. We explored how a problem that has been simmering for years has suddenly jumped to the top of to-handle lists for enterprise IT leaders.

The Budget Shock No One Planned For

The numbers are striking. “Just in the first few months of this year, most storage companies raised their prices by anywhere from 30% to about 75%,” Krishna noted. For IT leaders already managing double-digit data growth, that is a significant jolt to infrastructure budgets that were built around very different assumptions. Organizations are being asked to absorb substantially more data while spending far more per unit of capacity, with little room to maneuver.

The problem is compounded by poor visibility. Most enterprise IT teams know they are sitting on large volumes of cold data, files that are no longer actively used but continue to consume expensive primary storage. What they typically lack is the precision to act on it. “You can’t manage what you don’t know,” Krishna said. “Most IT leaders, I think the problem is they don’t know really what the issues are with their unstructured data.”

Unstructured data is sprawling by nature, generated across different users, applications, and systems with no consistent format to make analysis straightforward. Knowing the general shape of a problem is different from knowing which specific data can be safely moved, where it lives across on-premises and cloud environments, and what the real savings opportunity looks like.

Reclaiming Capacity Without Buying More

Komprise designed its Flash Stretch Assessment to close exactly this gap. Available to qualified customers with at least 500 TB of data for no charge, the service analyzes storage environments across both on-premises and cloud infrastructure to produce a concrete picture of data activity, growth patterns, and cost distribution by user, application, and data type.

The goal is to identify cold data that can be moved to lower-cost storage tiers, freeing up primary capacity without requiring new hardware purchases at today’s elevated prices. “We’re freeing up all that space, but you can now put new data onto it,” Krishna said. “You’re kind of reclaiming existing capacity without having to buy at these exorbitant prices.”

When organizations act on the assessment, the financial impact can be significant. Unstructured data accounts for roughly 30% of the average IT storage budget, and that figure encompasses not just primary storage but also backup and disaster recovery copies. Komprise reports that customers who tier their cold data can reduce storage and backup costs for their unstructured data by around 80%.

Transparent Access, Not a Trade-Off

A predictable concern with any tiering strategy is disruption. Moving data to lower-cost storage only delivers value if users and applications can still reach it without friction. Komprise addresses this through its patented Transparent Move Technology, which relocates data to the cloud or another tier while leaving a dynamic link in its original location.

“It looks like the X-ray image is still local,” Krishna explained. “Your applications can still open it, but when they go to open it, we stream that data from the cloud instead of it sitting locally. It’s transparent to users and applications so they don’t see any change.”

For data that follows cyclical patterns, such as project archives that go dormant for months before becoming relevant again, the platform supports bulk recall that restores entire datasets on demand, giving IT teams flexibility to handle the rhythms of how enterprise data actually gets used.

The TechArena Take

The current storage crisis is not a temporary disruption that patient organizations can wait out. Memory chip shortages tied to AI demand will likely ease, but the need to optimize your existing resources instead of planning to scale to meet increasing demand alone is likely the new normal.

The current pressure is forcing a discipline that should have existed all along: Many enterprises have been storing unstructured data without meaningful visibility into what they have, how fast it is growing, or what it is actually costing them. That was always inefficient. The difference now is that the inefficiency has become expensive enough to demand attention. Organizations ready to engage with decisions about what data to keep, where to keep it, and how to organize will be best positioned in the race to improve AI performance, slow infrastructure spending, and increase operational agility.

To learn more, listen to the full podcast or visit komprise.com.

Doug Finke joins an episode of Data Insights to discuss quantum computing, commercialization, cybersecurity risks, and the road to enterprise adoption.

Just two years ago, most companies were simply asking what AI could do in an enterprise setting. In 2026, they are asking a harder question: how to scale without breaking their reliability or their budget. That shift from curiosity to capacity is where Isayah Young-Burke, go-to-market strategist at IONOS, spends most of his time.

In a recent TechArena Data Insights episode, I sat down with Isayah and Solidigm’s Jeniece Wnorowski to explore why security and access risks are the underexamined obstacle in enterprise AI, how data sovereignty is reshaping infrastructure decisions on both sides of the Atlantic, and why storage is now one of the most strategic layers in an AI-ready stack.

An Expansive Vantage Point Across the Stack

IONOS, part of the publicly traded IONOS Group with more than 6.6 million customer contracts globally, occupies a distinctive position in the cloud market. The company serves customers ranging from an individual registering their first domain to an enterprise running a multi-client managed service provider business. That breadth, Isayah explained, provides a kind of ground-level intelligence that shapes how the company serves customers and thinks about AI adoption.

“It's that customer service and that experience we carry behind our brand. It has to be good at every level,” he said. “AI adoption…doesn’t just start with AI. It starts with that digital footprint that grows into infrastructure. AI becomes that natural next step, just like after you get a website, you start thinking about cloud storage and cloud infrastructure. So we get to see that whole journey.”

The Real Gaps: Security, Data and Skills

When asked where he sees the biggest gaps as organizations operationalize AI, Isayah was direct: most enterprises are focused on the wrong thing. While model selection often dominates the discussion, choosing the “right” model is not what predicts success.

“Most AI challenges at scale — it’s not really a capability problem. It’s a system problem, not the model. And increasingly, they are a trust and access problem,” he said.

He drew on a panel discussion at IT Expo where a fellow speaker raised concerns about the level of access AI agents are granted within enterprise environments. An agent embedded in a company’s internal systems can do more than answer questions. It can write, delete and trigger workflows across an entire environment. “That’s a very different risk profile than a website chatbot,” Isayah noted.

Beyond security, he identified data readiness and workforce skill gaps as persistent obstacles. IONOS has responded by building tools like IONOS Momentum and the AI Model Hub, designed to make AI infrastructure accessible to small-to-medium businesses and public sector organizations that need practical solutions, not just raw compute.

Data Sovereignty and the Regulatory Divide

Operating across the US and Europe gives IONOS a useful vantage point on how regulatory environments shape AI infrastructure decisions. In Europe, regulations like GDPR and initiatives like Gaia-X have made data residency a front-line concern from day one. In the US, speed and innovation tend to dominate, but that is shifting.

Isayah pointed to a dimension of US cloud law that often goes unexamined: the Cloud Act gives the US government legal authority to access data held by American cloud providers, even when that data is stored in Europe. IONOS operates under a different legal framework in Europe, because it is a subsidiary of a German company. This distinction matters significantly to companies that do business overseas.

“Knowing where your data lives and who has access to it under what conditions really matters,” he said. “Providers who can give answers to those questions have a real advantage.”

Storage as Strategic Infrastructure

Nowhere is the infrastructure shift more visible than in storage. Isayah described storage as having “quietly become one of the most strategic layers in AI,” noting that as AI-enabled workloads scale, enterprises must manage massive volumes of unstructured data, including text, images, logs and embeddings, that traditional storage architectures were never designed to handle.

With this new challenge, he noted, there’s been a shift toward object storage. The medallion architecture approach, organizing data into bronze, silver and gold enrichment tiers, has become a common framework for managing this complexity. These practices have become the backbone for data lakes, the central repositories of where raw data lives before being processed. S3-compatible object storage has emerged as the de facto standard for these data lakes, valued for its scalability, cost efficiency and — through IONOS — API accessibility.

Preparing for Agentic AI

Looking ahead, Isayah sees agentic AI as the next major infrastructure challenge. “AI agents aren’t just generating outputs,” he said. “They’re interacting with back-end systems. They’re triggering workflows from different applications and different software, making decisions across platforms in real time.”

That shift demands decentralized architecture, low-latency edge and cloud environments, strong API interoperability and, above all, rigorous security controls. He referenced Anthropic’s recent decisions around its Mythos model, where it chose not to release the model publicly after it was tested offensively in a sandbox experiment, found system vulnerabilities and escaped the test environment as instructed, as a reminder of what is at stake.

“Without fail-safes, it’s unwise to release this into the public,” Isayah said. “The foundation for these automated systems has to be solid.”

The TechArena Take

For technology decision-makers, the practical takeaway from this conversation is straightforward: the infrastructure decisions being made now around storage architecture, data governance and agent access controls will determine the ability of organizations to scale AI later. IONOS’s position gives Isayah a grounded view of where those decisions are going well and where they are not. Organizations still treating storage as a commodity and AI security as an afterthought, may find that catching up later is considerably more expensive than getting it right now.

To learn more, listen to our full conversation in the published podcast, and read about IONOS’s cloud offerings, including the AI Model Hub, at ionos.com.

Foresight’s Atif Ansar explains delivery risk, project delays, governance, and how AI can improve large-scale data center buildouts.

Our next TechArena Data Insights comes live from Xcelerated Compute, where Jeniece Wnorowski and I had a chance to sit down with David Moehring, general partner at Cambium Capital and former CEO of IonQ, to unpack how quantum computing fits into the broader computing landscape, and why the real story is less about breakthrough moments and more about building entire ecosystems.

From Physics Labs to VC

David has had a long and illustrious career working in quantum across academia, government and industry. His career began in academia, where he pursued a PhD in atomic and optical physics and quantum computing before moving into applied research at Sandia National Labs. From there, he transitioned into government, funding advanced quantum computing initiatives, and became the founding CEO of IonQ before his move to Cambium Capital.

That multidisciplinary background continues to shape how he evaluates companies today. “I found it very useful to understand the motivations of all of these different parts of the ecosystem,” he said, adding that it greatly informed how he looks at investing into quantum.  

Systems-Level Thinking  

Cambium Capital is an early-stage venture capital fund that invests in advanced computer hardware, and we talked about their recent moves in investing in the total cost of ownership (TCO) of AI data centres.

Rather than chasing a single breakthrough, the firm invests across the compute ecosystem, focusing on improvements to be made in different areas. “We have companies that we’re looking at as part of our portfolio that work in power delivery, movement of data, embedded memory, and packaging across the ecosystem,” he explained, “because if you move just one of them forward, everything else is still backed up.”

This systems-level thinking extends to how Cambium evaluates startups. Technical depth is essential, but so is market readiness. “You just can’t have your one good technology and toss it over the fence and think other people are going to integrate it,” he notes. “You need to be very deeply technical in your field, but you also need to really understand where it comes in.”

Complement, Not Replace

When it comes to quantum computing, David is clear-eyed about both its promise and its limitations. The technology is not poised to replace classical systems; it will augment them. “It’s not that if you make quantum computers then classical computers or GPUs will go away because they really solve very different problems,” he said.

Instead, quantum systems will tackle specific classes of problems that are either infeasible or inefficient for classical architectures. This mirrors the evolution of GPUs and specialized accelerators, each designed for distinct workloads.

“There are jobs that just cannot be done by classical computing,” David explained, framing quantum as another layer in an increasingly heterogeneous compute landscape.

On the relationship between AI infrastructure and quantum computing, he envisioned both evolving in parallel, solving very different problems in the long run. That being said, he noted that classical computers will still be required to control quantum computers.  Beyond that, foundational innovations in materials science and lasers are likely to benefit both domains.

Building the Quantum Ecosystem

Cambium’s investment strategy isn’t limited to headline-grabbing processors. The firm also has strong conviction in the new quantum investment vehicle, 55 North, where David is the Board Chairman.  

“There’s still a lot of hardware development that is needed, not just for the kind of the quantum processor itself, but the ecosystem,” David said.

That includes everything from laser systems for atomic qubits to cryogenic infrastructure for superconducting systems, components that rarely make headlines but are essential to scaling the technology. This mirrors Cambium’s broader philosophy: meaningful progress happens when the entire stack evolves together.

Misconceptions and Early Signs  

Despite growing interest, quantum computing remains widely misunderstood. Even quantum physics itself can be very counterintuitive, and many physicists struggle with some of the rules. That gap between theory and practical understanding often fuels unrealistic expectations. His best advice? Talk to the experts.

Looking ahead, David points to a specific category where quantum could first demonstrate real-world value, the field of biopharmaceuticals. “We strongly believe that’s where you’re going to see first real kind of advantage from quantum,” he predicted.

At the same time, he remained cautious about overblown claims, adding that some of the news was driven more by hype than understanding.

The Tech Arena Take  

David built on his years of experience working on quantum in different capacities to deliver a pragmatic take on where the field is headed. Our key takeaway was that along with the big-ticket processors, investments need to focus on building the underlying infrastructure needed to keep quantum compute running, something Cambium Capital is keenly focused on. Pushing one piece of the puzzle without solving other challenges would only lead to bottlenecks being moved down the line.  

Watch the interview here.

AI adoption depends on accessibility. A.J. Camber explains how computer vision, data quality, and simplified workflows enable enterprises to scale AI without relying on scarce data science talent.

Data Insights podcast with Infleqtion CTO Pranav Gokhale on combining quantum, GPUs, and CPUs to deliver fault-tolerant systems and real-world impact.

‍

Maya Kalyan’s career has always lived at the intersection of disciplines. With a background in biomedical engineering and more than a decade in life sciences, she now serves as a staff algorithms and AI engineer in the molecular diagnostics space at Thermo Fisher Scientific. In a recent TechArena Data Insights episode, Solidigm’s Jeniece Wnorowski and I heard Maya offer a practitioner’s view on where AI is genuinely delivering value in healthcare, and where significant work remains.

Four Problems, One Goal

The starting point for any meaningful discussion of innovation in diagnostics, Maya explained, is understanding what problems the industry is actually trying to solve. She identified four primary areas of focus: accuracy and reliability, turnaround time, cost reduction, and automation.

On both cost and turnaround time, she pointed to how the increasing demand for molecular testing is being met by innovations like multiplexing, which is the ability to detect multiple pathogens within a single sample. “We have respiratory virus tests that can detect COVID and flu and RSV viruses all within the same test,” she said. “That reduces the reagent use and lowers your consumable cost while also increasing throughput.” The broader goal, she noted, is building diagnostic systems that are simultaneously faster, more reliable, more affordable, and capable of handling the volume demands of high-throughput clinical and research environments.

Where AI Works, and Where It Doesn’t

Maya offered a measured perspective on AI’s current capabilities, drawing a clear line between where the technology performs well and where it has room for growth.

AI tends to be most effective, she said, when working with large, well-structured datasets toward a defined predictive outcome, such as pattern recognition in biological data, quality monitoring in experimental workflows, and domain-specific assistants that help researchers navigate documentation or troubleshoot instruments. The benefit, she noted, is improving the user experience and reducing manual touchpoints.

The limitations, however, are equally important for technology decision makers to understand. “When it comes to large language models, specifically the risk of hallucinations and its non-deterministic nature — where it can make up things or not say the same thing each time — can be a barrier to adoption in scientific or healthcare settings.” Her prescription is a hybrid approach: one that keeps human expertise in the loop by design, even as agentic AI systems grow more capable of autonomous workflows.

Regulatory Reality

Building AI-enabled diagnostic products is not simply a technical challenge. Maya outlined a layered set of constraints that shape every deployment decision, starting with data governance. Healthcare datasets often contain sensitive patient or genomic information. Considerations for privacy affect how data can be accessed, shared, and used in ways that go well beyond standard HIPAA compliance.

There are also practical deployment decisions with regulatory implications: whether AI systems run in the cloud or directly on an instrument, and how factors like connectivity and latency influence what’s feasible. And once a model is deployed, the work isn’t over. “Teams need some kind of post-market surveillance plan,” she said, “which requires a strong model observability service where they can monitor the performance of the model and identify any drifts.” In practice, applying AI in this space means balancing innovation against a set of strenuous operational and regulatory realities.

The Data Foundation Problem

Before AI can meaningfully contribute to product development or diagnostics, Maya emphasized that organizations need to get their data house in order. That begins with rigorous data curation, ensuring experimental data is well-annotated and collected consistently so models can learn real patterns rather than artifacts of poor methodology.

Accessibility is the other piece. In many research organizations, data is scattered across instruments, labs, and databases with no unified infrastructure to bring it together. Maya pointed to large open biomedical datasets such as the Cancer Genome Atlas curated by the National Institutes of Health as important resources the research community already relies on. Looking ahead, she sees federated data approaches, which enable collaboration without requiring the sharing of raw patient data, as critical to accelerating AI’s role in diagnostics at scale.

The TechArena Take

While grounded in biomedical engineering, Maya’s perspective reflects broadly applicable lessons: the most durable AI deployments are built on disciplined data practices, realistic expectations, and a clear-eyed understanding of evolving regulatory requirements. In a field where the stakes are measured in patient outcomes, the pressure to get it right is acute. If AI lives up to its potential in life sciences, the payoff won’t just be operational efficiency. It will be earlier diagnoses, more personalized treatments, and meaningfully better quality of life for patients facing some of the most challenging medical conditions.

A conversation with qBraid’s Kanav Setia on quantum computing’s evolution, developer challenges, and the role of software in shaping its future.

‍

Submer's Gabriel Lazar joins TechArena and Solidigm at OCP EMEA on data center sustainability, heat reuse, and lower-impact AI infrastructure.

‍

There’s a conversation happening at the edges of the AI infrastructure world that hasn’t quite broken through to the mainstream yet. It’s not about which GPU cluster wins the benchmark race or which hyperscaler is adding the most capacity. It centers on something far more fundamental: the cost of moving data.

In a recent Data Insights episode, I sat down with Solidigm’s Jeniece Wnorowski and Nilesh Shah, VP of Business Development at ZeroPoint Technologies, to work through where this friction in modern AI systems lives.  

Ten Times More Expensive to Move Than to Compute

Nilesh began with an often-overlooked aspect of data storage: the amount of power moving data takes. Moving a single bit of data from storage, through high-bandwidth memory or low-power double data rate (LPDDR), and into the on-chip static random access memory (SRAM) where computation actually happens costs roughly ten times more in power than performing the computation itself. That ratio explains why inference chip innovators like Groq, Cerebras and SambaNova are focusing on data movement and memory hierarchies over compute.

Zero Point Technologies was founded on the premise that the need for data and memory is going to increase rapidly, and one of the ways to tackle that challenge is through lossless memory compression. By reducing the volume of data physically moving across the system, you increase the effective bandwidth and capacity of the compute engine.

Agentic AI Compounds the Problem

On the question of whether AI workflows were being constructed correctly for the management of data, and how this could change as enterprises start scaling inference into different parts of their business, Nilesh pointed out that the key problem to be solved is agentic AI entering the workflow.

A pattern seen at recent tech conferences was that chip designers were integrating multiple specialized AI agents into a single electronic design automation (EDA) workflow, each handling a distinct task, like error detection or chip verification. This would mean having domain-specific inference solutions for even EDA operations, fundamentally changing the way enterprises will need to think about data.  

Prefill vs Decode Stage

As data becomes a challenge, memory bandwidth could become a bottleneck. Nilesh pointed out that agentic workflows and inference takes place in two stages, prefill and decode. The prefill stage processes the input prompt and is genuinely compute intensive. Modern GPU clusters handle this part reasonably well. The decode stage, where the output is generated, is extremely memory intensive and is what’s really limiting tokens per second.

When it comes to responsiveness at enterprise scale, say 100,000 employees simultaneously interacting at that scale across multiple streams of data, the decode phase becomes a real bottleneck. At NVIDIA GTC 2026, a lot of the keynotes revolved around developing heterogenous architectures that can manage the decode phase more efficiently.

We talked about when quantum computing would enter the picture. “What is the ChatGPT moment for quantum computing? That’s the favourite question I like to ask,” said Nilesh. He predicted that it could make sense to attach quantum processing units to data centers to efficiently offload some of the compute work that quantum tends to do well. There are currently examples of banks deploying early quantum computers, and another use case could be encryption and creating more secure encryption protocols.  

Future Innovation in Memory  

When I asked Nilesh what he sees on the horizon for memory and storage technology, he outlined three distinct directions where investment and innovation are converging.  

The first is alternative memory technologies. Dynamic random-access memory (DRAM) is a decades-old architecture that hasn’t changed fundamentally, and its limitations are starting to bite at exactly the moment AI workloads are scaling fastest. The second is new interfaces between memory and compute that will transform how memory communicates with the compute engine.

The third is the most significant shift in perspective: the unit of infrastructure design is moving from the chip, to the server, to the rack, and now to the data center as a single coherent system. Organizations are thinking about AI infrastructure in terms of megawatts allocated to a data center, with memory, storage, and compute all traded off within that power budget.

The biggest misconception, he felt, was the assumption that scaling AI output will keep being built on a proportional increase in power. “I expect a breakthrough that someone will come up with an entirely new style of physics that will break that linear assumption that to go from 100 LLMs to a million or going from a million users to 100 million, we’ll just multiple the megawatts of power,” he said.  

The TechArena Take

My conversation with Nilesh clarified a change in direction I’ve noticed at many recent tech conferences. The 10x cost differential between moving data and computing on it is the reason the entire inference chip landscape looks the way it does. It’s a significant engineering constraint that companies like ZeroPoint are building directly against. The prefill-decode distinction matters because enterprises planning inference deployments at scale need to architect around the decode phase as a distinct bottleneck.

We’re excited to see what new innovations take place in the memory space, and if, as Nilesh believes, someone will eventually find a way to scale AI without the linear progression of more compute meaning more power.  

Watch the podcast here.

Quantum computing conversations tend to get pulled toward the exotic: superposition, quantum entanglement, and what a world with exponentially faster compute will look like. But at the Xcelerated Computing Conference in New York, Solidigm’s Jeniece Wnorowski and I spoke with Burns Healey, quantum infrastructure lead for Dell, who offered a grounded perspective. For quantum technology to matter, it first needs a place to land, and that place is working alongside the classical data center.

It’s a framing that shapes everything about how Dell is approaching the market and one with practical implications for technology decision makers weighing when and how quantum fits into their roadmap.

A Hybrid Future

Our conversation started with reconsidering the terms we use when we talk about quantum technology. “Quantum computers are almost a bit of a misnomer,” Burns said. “When you say quantum computer, I prefer to use the term quantum accelerator, because really that’s what they are. They’re an add-on to HPC or data center infrastructure that give you specialized options for computing specific workloads.”

This perspective that quantum technology is best considered as an extension of high-performance computing environments can be helpful to enterprise leaders who may feel pressure to engage with quantum. Organizations attempting to adopt quantum before they’ve pushed classical computing to its limits are, in his view, getting ahead of themselves.

“Going to a quantum computer before you’ve attempted to use classical HPC or large data center environments is a bit like trying to run before you’ve walked,” he said. “Only once you hit those limits in your data center, in your HPC environment, will you start to think about what quantum can do that you can’t currently do.”

Lowering the Barrier with Logical Qubits

Much of the early quantum conversation focused on physical qubit (quantum bit) counts and error rates, but recent conversations in quantum computing have shifted toward logical qubits and error correction as the field considers what usable quantum will look like.

Burns drew a direct analogy to classical computing. Just as error-correcting code allows applications to run more reliably, logical qubits aim to provide a stable, abstracted layer above the physical qubit substrate.

“The way we use them from a vendor and hardware supplier viewpoint is that we are going to aim to abstract away a lot of that physical layer complexity from the end user,” he said. “It’s a lowering the barrier to entry question in my mind, and the best way we can help onboard new people to the technology.”

Closing the Latency Gap With NVIDIA

When you think of quantum computers as quantum accelerators, the importance of the infrastructure that enables quantum and classical computing to work seamlessly becomes paramount. Rather than building quantum processing units (QPUs), Dell is helping produce the ecosystem and infrastructure appliances that will make quantum devices usable within real data center environments. A major challenge in that area is latency between quantum and classical systems. Burns pointed to Dell’s collaboration with NVIDIA as a current example of this work.

NVIDIA has developed a framework called NVQLink, designed to minimize the round-trip latency between QPUs and classical compute. Using NVQLink on Dell PowerEdge servers, the two companies recently demonstrated sub-4-microsecond latency, a result Burns described as meaningful progress toward the kind of tight integration that real quantum workloads will require.

“We’re really looking at what the technology needs in terms of specifications and hitting those targets to make this infrastructure usable for real quantum computing,” he said.

Dell is also engaged with quantum partners including QuEra and IQM, as well as a joint research initiative with Ernst & Young, all documented on Dell’s hybrid quantum-classical computing page.

The Path to Scale

When asked what needs to happen technically and operationally for quantum to move from research settings to deployable infrastructure, Burns identified two parallel tracks.

On the software side, progress is already underway. Frameworks like IBM’s open-source Qiskit are helping developers work with quantum gates and algorithms today. The next meaningful shift will come when developers can work at a Python-level abstraction, or eventually through application-specific tools that require no quantum expertise at all.

On the hardware side, cabling is one of the more pressing unsolved problems. Superconducting qubit systems require analog signals routed to each individual qubit. At 50 or 100 qubits, that is manageable. At thousands or millions of qubits, the cabling architecture becomes an issue. Ideas to address this include embedding classical components inside dilution refrigerators and more sophisticated multiplexing approaches, both of which introduce their own challenges.

The TechArena Take

Dell’s positioning in the quantum space is as perceptive as you would expect from one of the world’s classical computing giants. Rather than competing with QPU vendors, the company is focused on the infrastructure layer that will make quantum systems usable in real enterprise environments.

Burns’s framing of quantum as an accelerator, not a computer, is a useful corrective for organizations trying to calibrate their engagement with the technology. For most enterprises, the near-term question is not whether to adopt quantum, but how to ensure that classical infrastructure is ready when quantum workloads become viable. The organizations with the strongest HPC foundations will be best positioned to take advantage of it.

Listen to our conversation in the full podcast episode, and for more information about Dell’s hybrid quantum-classical computing work is available on Dell’s quantum computing site.