Data Gravity Meets Acceleration: Foundation for Sovereign AI

The AI industry is entering a new era, one defined not by experimentation but by execution.

Enterprises that once ran pilots are now asking a harder question: ‘How do we operationalise AI securely, sustainably and at scale?’.

At the centre of that question is data gravity, the principle that as data accumulates, it attracts more computation, analytics and intelligence around it. 

It’s not just a technical concept; it’s an economic and architectural shift reshaping how the next generation of AI infrastructure is being built.

Sovereignty and the shift to data-centric AI

As enterprises modernise, the relationship between data and AI is inverting. Instead of moving data toward centralised models, the emerging paradigm is to bring AI to where data already lives –securely, locally and sovereignly.

“In today’s enterprise, we’re faced with a new notion that your data carries that same gravity,” says Simon Lightstone, Director, Product Management at EDB. 

“The more concentrated the data, the greater the pull – and within that pull lie both the greatest innovations and new vulnerabilities.”

That principle defines the foundation of sovereign AI: intelligence built around data, not away from it.

Nave Algarici, Group Product Manager, Generative AI at Nvidia expands on that shift: “The main pull is around the gravity of the data, right? If, previously, the paradigm was to bring your data to where your AI sits, now it shifts to bringing your AI to where your data resides.”

He continues: “By combining the solutions from Nvidia AI software and Nvidia NIM to EnterpriseDB [running on Supermicro servers], we’re enabling that turnkey solution where your data is already sitting in EnterpriseDB. You turn a key – now it’s AI powered, AI accessible and connected to your AI application.”

This approach preserves data sovereignty while unlocking speed and accessibility. It’s a design principle that reflects where the entire AI infrastructure market is now headed.

And it’s precisely at this intersection of control and capability where enterprises are now recalibrating their strategies. 

Kevin Dallas, CEO of EDB, captures the shift: “We know that 95% of organisations say they plan to become their own data and AI platform within three years, yet only 13% have successfully done so today. The leaders are the ones treating sovereignty not as a constraint, but as the foundation for competitive advantage.”

These perspectives reinforce the industry-wide movement toward architectures that make sovereignty a strategic enabler rather than a barrier.

Acceleration with efficiency: Redefining performance and ROI

Operationalising AI at scale requires more than just compute power – it requires economic and architectural efficiency. That’s where hardware optimisation and database performance meet.

Simon highlights the business impact: “You’re essentially cutting that tax of having to pay for your performance every single month.... You’re talking about 6x acceleration just from working with Supermicro. You’re actually getting something that works well together and is really optimised as one package.”

The “tax” he describes – escalating cloud costs for persistent performance – has become a growing challenge for enterprises deploying generative and agentic AI workloads.

Somik Behera, General Manager, Cloud, Datacenter & AI Software Products at Supermicro explains how the right architecture changes that equation: “We take EDB’s market-leading AI database and AI software, marry it together with Supermicro’s building-block architecture and Nvidia’s GPUs, so that you can not only have AI available, sovereign, in your data centre, but you can have it AI fast.”

The result is a cloudlike experience at a fraction of the cost. It’s performance with control, acceleration without the compromise of lock-in. 

Removing bottlenecks, simplifying adoption

Acceleration only delivers value when it’s easy to adopt. Many organisations struggle to take the latest AI innovations and actually apply them in production environments. 

“Often simplicity is the barrier and the reason why customers really struggle to take the latest technology and then actually apply this to their business. When you can deploy high availability in three clicks…that’s really key,” says Simon. 

That simplicity is what enables real operationalisation – the ability to move from prototype to production without massive reengineering. From faster embedding to fresher data, the emphasis should be on keeping intelligence flowing continuously and securely. 

Enabling the next wave of agentic AI

As AI systems become more autonomous, their need for data freshness and availability intensifies. Agentic AI – multiple reasoning agents interacting and iterating – depends on systems that are both fast and sovereign.

“AI interactions are going to scale deployments quite significantly,” Nave says. “These iterative agents and reasoning agents ... need that data highly available and keeping that data fresh, keeping that data up to date.” 

He points to Nvidia NIM as a critical enabler. The containerised service enables users to deploy AI models into production and into enterprise applications, delivering a flexibility that allows enterprise customers to “really build amazing applications.”

In this model, sovereignty isn’t a constraint. It’s a competitive advantage. It ensures that the more intelligence a company builds, the more control it retains over its data, models and value.

The architecture of the future: Sovereign, accelerated, simple

The points raised by EDB, Nvidia and Supermicro represent a broader transformation in the AI industry: a move toward architectures that are fast, open and sovereign by design. 

EDB provides the data substrate, with its two decades of Postgres expertise to manage and secure data at scale. By collaborating with Nvidia and Supermicro, enterprises can build sovereign AI anywhere – across public clouds, private data centres or edge environments. 

Harnessing that gravity while maintaining control is what defines this new era of AI infrastructure.