Enterprises Keep AI Models Local as Data Privacy and Costs Reshape Strategy

A fundamental tension sits at the heart of enterprise artificial intelligence adoption: whether to move sensitive data outward to AI models or bring models inward to where data resides. Regulatory pressures and geopolitical concerns are tilting the balance decisively toward the latter approach, reshaping how data infrastructure companies position themselves in the market.

Cloudera Inc. has centered its AI strategy on the principle of keeping models close to data. Rather than requiring customers to transfer their proprietary information to external AI services, the company focuses on deploying models within the environments where data already sits. According to Wim Stoop, product director at Cloudera, this represents a fundamental reversal of conventional thinking.

Rather than saying, 'Just give us all your data and we'll in return provide you access to a model that's then contextualized based on that data,' we turn things around. We say, 'We'll bring AI to where the data is and where the data lives.'

Wim Stoop, product director at Cloudera

Stoop discussed these strategic shifts during an appearance at Cloudera EVOLVE26, where he spoke with theCUBE's Rebecca Knight and Paul Gillin about how enterprises can integrate AI while preserving control and managing expenses.

Sovereignty moves to the center of enterprise AI strategy

Cloudera's partnership with Mistral AI SAS exemplifies this shift toward localized AI deployment. The collaboration integrates Mistral's models directly into Cloudera's data platform, with particular appeal for European organizations. Stoop emphasized that Mistral's geographic location and European heritage carry significance alongside its technical capabilities.

Certainly for us in Europe, having a French organization, a local company, that provides these models and provides the complete stack is very attractive for a lot of our customers. Mistral was a very easy choice for us to say, 'Those are the ones that we're going with.'

Wim Stoop

Data sovereignty now occupies a central position in enterprise AI planning, sitting atop a maturity curve that encompasses sovereign cloud infrastructure and localized operations. Regulatory frameworks including the European Union's Digital Operational Resilience Act and the U.S. CLOUD Act are compelling organizations to demonstrate genuine control over their systems rather than merely asserting it. Yet regulatory compliance alone cannot propel AI beyond experimental phases.

The economics of model operation at scale present a more immediate obstacle for many organizations. Token generation demands substantial computational resources, and reliance on public cloud providers creates escalating costs as usage increases. Stoop identified this tension as the central challenge confronting enterprises attempting to scale AI deployment.

https://www.youtube.com/embed/uot59HI53-M?feature=oembed

If you use a public provider for that, the more you use your model, the more you get hit on cost. How do I get that high utilization of a good model inside my organization without being exposed to horrific costs? That, I think, is the biggest challenge that organizations are struggling with right now.

Wim Stoop