Inside the AI Operating System Tackling Data Starvation as the UAE Scales GPU Capacity

At Ai Everything Abu Dhabi 2026, Dany Khoury, Regional Channel Director at VAST Data Middle East, met with The Recursive to discuss why the next phase of AI adoption will depend on how organizations manage their data, the growing demand for sovereign AI infrastructure, and the shift from experimentation to production.

The Middle East is investing heavily in the computing infrastructure needed to power artificial intelligence. But as governments and technology providers race to build GPU capacity, another challenge is emerging: making the vast amounts of data held by enterprises accessible, secure, and useful for AI.

While governments and emerging AI cloud providers are currently driving much of the region's demand, the next wave could come from enterprises moving beyond experimental AI projects toward production-ready applications.

For VAST Data, which has evolved from a data storage provider into what it calls an AI operating system company, this transition represents a significant opportunity. Its regional growth includes a strategic partnership with Abu Dhabi's Core42, an expanding presence in Saudi Arabia, and growing demand for infrastructure capable of supporting sovereign and agentic AI.

During Ai Everything Abu Dhabi, held on October 6–7, 2026, The Recursive co-founder Etien Yovchev spoke with Dany Khoury, Regional Channel Director at VAST Data Middle East, about the company's journey in the region, the infrastructure challenges facing AI adoption, and why the real competition among enterprises is only beginning.

VAST Data is originally a US-based company. How did your journey in the UAE begin, and how has the business developed in the region?

I've been in the UAE for almost 18 years, working with different technology vendors, including HP, Dell, EMC, and VCE. My experience has always revolved around data.

Over the last five or six years, I started seeing significant growth in the data market. Around 2021, emerging technologies and artificial intelligence were becoming increasingly important, and the UAE was launching initiatives to embrace AI.

I saw an opportunity for a new kind of data offering, one engineered specifically for data-intensive AI workloads and high-performance computing.

That's when I learned about VAST Data and decided to join the company. What attracted me was the opportunity and the new architecture it was bringing to the market.

We initially entered with our Universal Storage approach, positioning the platform as a way to consolidate unstructured data. This is particularly relevant because unstructured data represents a significant part of the information used in AI workloads.

Haider Aziz, who established VAST Data's presence in the region, played an important role in that journey.

We started working with our partner and alliance ecosystem to introduce this new approach to storage and build the foundations for the AI data market.

The UAE was among the first markets to embrace the technology. One of our biggest breakthroughs was with Core42, which became one of our most significant customers and partners in the region..

If we fast-forward four or five years, the market has developed significantly. We're now seeing demand across different types of AI workloads, particularly with the emergence of neocloud providers and GPU-driven infrastructure.

With so much investment going into GPUs and computing capacity, where does VAST Data fit into the AI infrastructure stack?

When you're looking at GPU-driven workloads, you have to think seriously about the data platform sitting behind them.

You need a properly engineered data platform that avoids data starvation at the GPU level. You don't want GPUs sitting idle because you need to monetize every single compute cycle.

That's where platforms like VAST Data come in.

Globally, our AI operating system is supporting more than three million GPUs. Customers are increasingly recognizing that legacy storage solutions cannot sustain the requirements of these new workloads.

But we didn't stop at storage. We evolved our platform into an AI operating system.

When you're building an AI operating system, you need to be able to bring in legacy data, make sense of it, contextualize it, and make it available to AI applications.

That requires software components and data services that support the entire lifecycle, from ingesting data to generating insights and powering AI workloads.

Today, we believe we're uniquely positioned to combine the AI data lifecycle with the infrastructure needed to support it.

Traditionally, organizations would acquire storage separately, then add software tools and data services to make that information useful for AI. Our approach brings these capabilities together.

This becomes even more critical as we move toward agentic AI.

The number of AI agents being developed will grow significantly, creating new demands on infrastructure. Without the right underlying data platform, organizations will struggle to scale these workloads.

And of course, everything needs to be properly governed, audited, and secured, giving customers control over their data and infrastructure.

Sovereign AI and sovereign cloud infrastructure are particularly important topics in the UAE. How does VAST Data contribute to this agenda?

Sovereignty is absolutely critical, not only in the UAE but across the region.

The UAE has taken a leading role in embracing sovereign infrastructure, and I believe other countries are following its example.

When we talk about sovereignty, we're talking about the entire technology stack, starting with sovereign compute, then networking, and finally data.

All of these elements need to operate within an appropriate governance framework.

VAST Data sits at the heart of the data sovereignty discussion because we provide the underlying data platform, together with the security, governance, and control capabilities required to operate these environments.

As AI adoption grows, organizations need the ability to use their data while maintaining control over where it resides, how it's accessed, and how it's processed.

Looking at the market in 2026, where is most of the demand coming from? Is it still primarily governments, or are enterprises starting to catch up?

Today, demand is still largely led and driven by governments.

The UAE has been particularly active in building the foundations for agentic AI, creating an example for the rest of the world to follow.

We're also seeing growing maturity among neoclouds and platform service providers. They play an important role, partly because of the global shortage of compute capacity.

As enterprises start adopting AI, many rely on these providers for the infrastructure they need.

However, over time, I expect the market to shift increasingly toward enterprises as their AI adoption matures.

There's an enormous amount of data sitting within enterprises today. Unfortunately, much of it remains in silos.

Different teams own different datasets, and those datasets aren't necessarily connected or accessible across the organization.

Enterprises are beginning to realize the value they can unlock from this information. They're exploring how to consolidate their data, generate better insights, and improve business processes to become more competitive.

We believe data is becoming the competitive edge.

Unless you have the right data infrastructure and can apply intelligence to that data, you're going to struggle.

I would say we're entering a race over the next couple of years. Enterprises will be competing to see who can move faster, become genuinely AI-driven, and gain a competitive advantage.

The interesting thing is that almost every company today describes itself as AI-driven. But whether they're actually taking action is another question.

Absolutely. There's talk, and there's walk.

Enterprises are still on a learning curve.

For the last couple of years, much of the conversation has been around experimental AI. And I think experimentation will continue.

But we're approaching an inflection point where enterprises will start moving toward production AI.

This is particularly relevant for VAST Data because we work with large enterprise customers and major AI cloud infrastructure providers, including companies like G42, CoreWeave, Nscale, Nebius, and Mistral, to name a handful.

The next phase, however, will increasingly be about enterprises recognizing the value of the data they already possess.

How can they consolidate that data? How can they start making sense of it? How can they translate it into something that improves their operations?

That's where we see ourselves playing a critical role: helping organizations accelerate their time to value, or what I would call their time to insights.

You're also expanding in Saudi Arabia. How is that market developing for VAST Data, and what opportunities do you see there?

We've already established an office in Saudi Arabia, and we believe the market represents a significant opportunity for us.

We're currently in discussions with several customers, and we already have existing customers in the country through our OEM relationship with HPE.

We're looking at opportunities to expand further while demonstrating how the VAST Data platform can support fully sovereign AI data infrastructure.

One thing that's particularly important when discussing sovereignty is security.

We recently announced VAST DataEnclave, which represents the next step in delivering secure, sovereign AI infrastructure.

The idea is to allow customers to keep sensitive data within trusted environments while securely bringing advanced AI models to that data. Using confidential computing, hardware isolation and cryptographic attestation, customers retain control of their data and keys while model providers retain control of their proprietary model weights.

Traditionally, customers often had to move their data to the cloud to run these models. That creates security and sovereignty concerns, especially for organizations handling sensitive information.

What we've developed in collaboration with NVIDIA is a way to bring models and data together within a secure computing environment.

Customers can run models on-premises using confidential computing and cryptographic protections.

The model providers maintain control over their model weights, while customers retain control over the keys to their data.

This creates an environment where organizations can use multiple AI models without worrying about sensitive data leaving their premises or crossing national borders.

It's an important capability for governments and enterprises that want to benefit from advanced AI while maintaining sovereignty over their information.

Looking three or four years ahead, how do you see agentic computing evolving, and what role will VAST Data play in that future?

We're definitely building for that future.

Over the coming months, we'll be introducing additional software capabilities that allow customers to run agentic workloads on our platform within fully governed and controlled environments.

That's going to be critical because we expect significant growth in agentic AI adoption.

We're already seeing customers and partners developing agentic AI workflows and leveraging VAST Data's platform to support them.

As more organizations move in this direction, the underlying infrastructure will become increasingly important.

It's not just about running an AI model anymore. It's about creating an environment where AI workloads can access data, operate securely, and scale as organizations expand their use of intelligent systems.

That's the future we're preparing for.