Wand AI adds StorONE to cut sovereign AI storage costs

Wand AI has added StorONE's Real-Time Tiering technology to its sovereign AI offering, making the storage software available through the Wand AI ecosystem.

The move adds a storage-efficiency layer to Wand AI's pitch to governments and institutions building sovereign AI systems, where infrastructure costs depend heavily on how fully compute and storage assets are used.

StorONE's software is designed to reduce the amount of flash storage needed for AI workloads by writing data to flash first, then shifting inactive blocks to higher-capacity media within the same volume. According to the companies, this keeps all data immediately accessible and avoids separate archive systems, restore steps, or policy-based movement between tiers.

The addition is optional for customers using Wand AI's backend. Organisations can continue running on existing storage arrangements where costs are already acceptable, while using StorONE where flash spending, physical footprint, or power use are the main constraints.

Storage economics

The tie-up reflects a broader issue in AI infrastructure. While much attention has focused on improving the use of scarce GPUs, storage remains a large and often underused part of the bill. AI systems are commonly provisioned for peak demand, leaving large amounts of expensive flash holding inactive data.

Wand AI argues that sovereign AI programmes need high utilisation across both compute and storage to remain affordable at scale. Its backend is intended to increase the use of shared national compute resources by allowing multiple data owners to run on common capacity rather than dedicated allocations.

StorONE is meant to address the other side of that equation. The company said its Real-Time Tiering can deliver average flash savings of about 90%, with around 10% performance and resource overhead. It contrasted that with legacy data-reduction methods, which it said typically produce 30% to 40% capacity savings while consuming far more CPU and memory resources.

The system also supports block, file, and object storage on standard hardware, allowing capacity to be pooled across workloads rather than tied up in separate arrays. That is particularly relevant for public-sector or national infrastructure projects that may need flexibility in sourcing servers and drives.

Deployment example

The companies pointed to one enterprise deployment as an illustration of the operational savings the approach can deliver. In that case, the storage footprint fell from nine cabinets to two, cutting data centre footprint by 80% and power consumption by 78%.

Such reductions matter in AI projects because rack space, electricity supply, and cooling can all limit how much compute infrastructure an organisation can install. Lower storage demands can therefore affect not only storage budgets but also the practical scale of an AI deployment.

Gal Naor, Founder and Chief Executive Officer of StorONE, said the issue is often overlooked when organisations focus only on GPU efficiency.

"Nations are sizing AI programs around GPU utilization and then quietly losing the same money one layer down, on flash that is mostly holding cold data. Real-Time Tiering moves only the inactive blocks to high-capacity media inside the same volume, so everything stays immediately accessible. One of our enterprise customers went from nine cabinets to two, resulting in an 80% smaller footprint and 78% drop in power usage. At national scale, that is capacity, megawatts, and capital returned to the AI program," said Naor.

Sovereign focus

Sovereign AI has become a growing area of interest for governments and regulated sectors seeking to retain control over data, models, and infrastructure. Vendors have responded with combinations of privacy, governance, and confidential computing tools aimed at customers seeking domestic or tightly governed AI deployments.

In that context, Wand AI positions its platform as infrastructure for what it calls AI labour across ministries, agencies, and other institutions. Adding StorONE broadens that offering by addressing storage use alongside compute use, rather than focusing only on model deployment or governance layers.

Cristian Felix, Chief AI Architect of Wand AI, linked the deal directly to the cost structure of sovereign AI systems.

"Sovereign AI is an economics question before it is a technology question, and the answer is utilization on both sides of the stack. We are pleased to welcome StorONE into Wand's sovereign technology ecosystem, adding a storage efficiency capability that ministries, institutions, and agencies can draw on where flash economics are the constraint, serving the same workloads on far less flash, and spending the difference on the AI labor itself," said Felix.

The integration sits alongside other elements in Wand AI's sovereign offering, including inference privacy, confidential computing, and model robustness tools. According to the companies, those components can be adopted independently, with StorONE intended for customers that need to reduce storage costs without giving up immediate access to data.

For buyers weighing the economics of large AI deployments, the development underlines how storage architecture is becoming part of the conversation alongside chips, networking, and power supply.