The planned integration of VAST AI OS into Hyperfusion’s containerized infrastructure aims to help governments and enterprises deploy AI inference faster while retaining control over their data. 

[For more news, click here]

Securing GPUs is only part of the challenge facing organizations looking to deploy AI. Power, cooling, and data center capacity can take considerably longer to arrange, delaying the move from computing hardware to production workloads. VAST Data and Abu Dhabi-based Hyperfusion are looking to address that gap through a partnership that combines modular AI data centers with a shared data platform, with the aim of bringing sovereign AI inference online in weeks rather than the years conventional data center capacity can take to establish. 

The planned integration will make VAST’s AI Operating System (AI OS) the preferred data platform for Hyperfusion’s Sentinel infrastructure and its larger HyperPod architecture. Sentinel packages a complete AI data center into a transportable 45-foot ISO High Cube container, with a reference configuration of 96 GPUs and approximately two to three petabytes of data capacity per scaling unit. Hyperfusion commissions the system in its factory before deployment, aiming to reduce the time required to bring it online at a customer site. For larger requirements, HyperPod extends the same modular approach to a 5 MW AI data center pod, with each pod scaling to approximately 20 petabytes of data capacity. 

Hyperfusion intends to take the systems to government, enterprise, telecommunications, energy, and other customers that need rapid or sovereign AI deployment. Customers can acquire the infrastructure as a customer-owned system or use it as a managed service. The Abu Dhabi-headquartered company also provides GPU compute through its managed cloud, which has been operating since 2024. 

Data Sovereignty and Distributed Infrastructure 

Beyond deployment speed, the partnership focuses on keeping data under customer control as AI infrastructure expands across locations. Hyperfusion says Sentinel can support confidential computing to protect models and data while in use and can operate fully air-gapped where required. With VAST AI OS as the preferred data platform, the companies aim to keep data in-country while allowing Sentinel and HyperPod deployments to share a single global namespace, providing a consistent data architecture across sites. 

“Every government and enterprise we speak to wants the same three things: capacity now, data that never leaves their control, and a platform that grows without starting over,” said Quentin Reyes, Founder & CEO. “Sentinel delivers the first two in a single container: a complete AI data center, commissioned in our factory and running on site in weeks, with confidential computing protecting models and data while in use. With VAST inside, every Sentinel extends the same data architecture as the last, so the tenth deployment scales as cleanly as the first.” 

The companies see particular relevance for AI inference, where computing needs to sit close to users, enterprise data, or available power. Hyperfusion also identifies multimodal models, computer vision, and physical AI as areas where data requirements are growing alongside computing needs. The modular approach is intended to let customers deploy individual units or expand into larger environments without having to establish an entirely separate data architecture for each site. 

VAST describes its AI OS as a unified software infrastructure stack combining data and compute services with agentic execution. Its underlying DASE architecture is designed to support performance, scale, simplicity, and resilience in distributed systems. Through the planned integration, the company aims to extend that data platform across Hyperfusion’s modular infrastructure. 

“The AI infrastructure conversation has been dominated by how many GPUs you can secure, when increasingly the gating factor is where you can actually put them,” said Jeff Denworth, Co-Founder at VAST Data. “Power isn’t arriving in neat, hundred-megawatt blocks on the timetable AI demands. Hyperfusion gives customers a way to build around that reality by deploying compute wherever capacity exists. Our opportunity together is to make the data layer just as portable, so customers can stand up serious inference infrastructure quickly, without every new deployment becoming its own data island.” 

Technical Validation Comes Next 

VAST and Hyperfusion will begin joint go-to-market activity, supporting customer introductions and commercial opportunities across their respective markets while completing technical validation of VAST within the Sentinel and HyperPod architectures. The companies also plan to explore a common control plane for managing distributed Sentinel deployments across multiple sites. 

The integration has not yet completed technical validation, and the proposed control plane remains an area for future exploration. The immediate next steps are therefore to validate the combined architecture and pursue customer opportunities as the companies work toward bringing the offering to market. 

Related Articles 

The Problem With AI Isn't the Model, It's Where the Data Lives 

VAST Data and AMD Expand AI Infrastructure Alliance to Tackle the Industry's Inference Bottleneck 

Before Physical AI Can Scale, Someone Has to Fix the Data Layer