If you walked the floor at Big Data London last year, you probably remember the sheer energy of every stage and booth buzzing with the same frantic energy: Deploy AI! Build agents! We are your harness! It felt like a collective sprint into the unknown.
Stepping onto the floor now a year later, the vibe was noticeably different.
The frantic sprint has paused, and now we are taking a deliberate, collective breath. Enterprise leaders aren't showing up asking How fast can we launch an AI agent? anymore. The real questions being asked over lunch, at the booths, and during sessions were far more grounded: How do we keep these agents from acting rogue? Where is our data actually going? And are we about to accidentally break a national privacy law?
After spending two packed days talking to customers, partners, and peers, two main themes stood out, and they explain exactly where enterprise AI is headed next.
Enterprise leaders are more focused on first adding guardrails than deploying
Attendees want to talk about bringing AI to their data, not the other way around
1. The Hype Hit a Wall: Why Everyone Is Adding Guardrails
Last year was all about agentic AI, giving autonomous agents the keys to the kingdom so they could run queries, trigger workflows, and make decisions without human hand-holding.
The promise is huge, but after reading the news over the past twelve months, reality has set in:
Agents made unauthorized calls because they lacked context.
Security blind spots opened up the moment AI tapped into hybrid clouds.
Shadow AI quietly popped up across departments, leaving risk teams sweating.
The big takeaway from this year’s panels wasn't that people are giving up on agents, it's that everyone is focused on building guardrails and governance first.
No one wants to stifle innovation, but everyone wants to avoid regulatory missteps. The general consensus was clear: an AI agent is only as good as the data foundation it sits on. Without context and tight control, governance, and line-of-sight lineage, autonomy becomes a massive liability.
2. The Sovereignty Pivot: Bring the AI to the Data, Not the Other Way Around
The second theme that came up in almost every conversation was AI sovereignty. Think of AI sovereignty as a company building its own tech future rather than renting it from someone else. It sits squarely inside the broader movement toward Digital Sovereignty.
At its core, digital sovereignty requires two foundational pillars: Data Sovereignty (guaranteeing not just physical data residency, but complete authority and access control over your data) and Sovereign Cloud (having true choice and operational control over your underlying infrastructure). Only when you hold the keys to both your data and your infrastructure can you realistically achieve true Sovereign AI, ensuring your models, logic, and long-term tech strategy are built on your own terms rather than rented from foreign tech giants or subject to external regulatory shifts.
With regulations tightening globally, leaders, especially in highly regulated industries like finance, healthcare, and the public sector, are facing a hard truth. You simply cannot haul sensitive data out of your secure perimeter and dump it into a public cloud LLM just to get smart insights.
The realization hitting the industry right now is simple: while bringing AI to where your data lives is the foundation of Private AI, true Sovereign AI goes a crucial step further. Sovereignty means total control and complete ownership across the entire AI lifecycle, giving you absolute authority over your data, your choice of models, and the underlying infrastructure and operations. Whether hosted on-premises, within a sovereign cloud, or across national boundaries, your stack remains under your operational control, meaning neither your data nor your governance ever has to be compromised.
Where Cloudera Fits into the Picture
This shift toward control and sovereignty is right in our wheelhouse, which is why we chose to present our session at the event focused on Sovereign AI and gaining visibility for your AI initiatives. We’ve been focusing our strategy around a core idea: Your data stays where it is, and we bring the AI to where your data lives.
Here is how that’s playing out in practice:
Giving You Choice in AI Models (Including Mistral): True sovereignty means having options, which is why we give you full freedom to choose the models that fit your specific requirements, whether that’s leveraging premier open-weights providers like Mistral AI or integrating other leading foundation models. By bringing your choice of LLM directly into Cloudera, your teams can run private inference, fine-tune on proprietary IP, and keep everything locked down within your own security boundaries, even in air-gapped setups.
Partnering for Sovereign Infrastructure: To give European enterprises true data residency and regulatory confidence, we are aligning with dedicated sovereign cloud environments, starting with support for the AWS European Sovereign Cloud. This allows organizations to run high-performance data and AI workloads locally without cross-border compliance headaches, with plans to expand across additional sovereign infrastructure providers as regional demand grows.
Making "Data Anywhere" Practical: To power all of this safely, Cloudera delivers an open data lakehouse foundation built on standards like Apache Iceberg and Apache Polaris, which are supported by sovereign infrastructure providers. This architecture gives your AI agents a clean, unified view of your enterprise data wherever it sits—across multi-cloud, on-prem, or sovereign environments—without requiring massive, expensive data migrations.
The Bottom Line
This year's Big Data London showed that industry leaders have redefined how to run a good race. They've moved from prioritizing speed above all else to tempering optimism with caution. The prize won't go to the team that rushes the fastest free radical agent into production. It will go to the teams that build trusted, secure, and sovereign foundations first so they can scale AI with confidence instead of constant anxiety.
