Insights from Adara Ventures' 2026 Industrial Advisory Committee

The standard reading of the global AI race is that the US is winning, China is competing, and Europe has missed the moment. That verdict is too coarse to be useful. AI is not a single market, and the geography of leadership looks very different depending on which layer of the stack you examine. In the US, roughly 70% of venture capital now flows into AI-related companies, close to 90% of recent Y Combinator cohorts are AI startups, and 88% of organizations have deployed AI across at least one business function.

While the US continues to dominate foundational AI models and compute infrastructure, Europe has emerged as one of the few regions competing at similar scale in AI x Energy. If Europe is to compete meaningfully in AI, it will likely do so through industrial and energy systems rather than frontier models. The region's advantage is rooted in decades of expertise across utilities, grid infrastructure, automation, and energy operations.

As AI deployment moves closer to physical infrastructure and operational systems, these capabilities become increasingly strategic, positioning energy as one of Europe's strongest AI verticals.

Cracking the Energy Problem

In April 2026, Adara Ventures hosted its fourth annual Industrial Advisory Committee in Bilbao, bringing together more than 30 leaders across utilities, TSOs, energy infrastructure operators, industrial corporates, corporate venture groups, and AI-native companies. The discussion centered on one question: where is AI actually creating value in energy, and where is it a patchwork of pilots?

The discussions revealed a clear disconnect between the pace of AI capability development and the ability of industrial organizations to operationalize it safely. While AI performance is improving rapidly, deployment inside critical infrastructure environments remains constrained; accuracy alone is insufficient in industrial environments. Systems must also be auditable, physics-aware, and capable of operating reliably under changing real-world conditions.

Adoption friction also rises sharply as AI moves closer to operational decision-making. Workflow automation has gained traction across the sector, but systems that interact directly with engineering processes and physical infrastructure continue to face a substantially higher bar.

Keynote: The Paradox, the Opportunity, the Imperative

The session challenged the assumption that AI adoption is moving slowly because the technology is not ready. The opposite may be true: AI capabilities are advancing faster than organizations can absorb them. Enterprise AI spending is already vast, but only a fraction is reaching production with measurable ROI. At the same time, model access is becoming ubiquitous, non-engineers are shipping production applications, and the cost of generating intelligence is collapsing.

For energy, the implication is direct: the bottleneck is no longer technical. It is whether organizations can build the guardrails, feedback loops, and domain-specific systems needed to deploy AI safely inside physical infrastructure.

"The risk is no longer experimenting with AI, but failing to understand how quickly it is already entering the organization."

– Arun Subramaniyan, CEO of Articul8

Five Shifts We're Tracking

The next phase of AI adoption in energy will likely be defined less by model capability and more by operational deployment. Five themes consistently surfaced across the discussions:

1. The cost of experimentation has collapsed. Tasks that previously required months of development and dedicated engineering resources can increasingly be prototyped in days, and in some cases hours.
2. Industrial software is moving closer to the operator. As AI systems reduce technical barriers, operational software is increasingly being shaped by people with domain expertise rather than pure software backgrounds.
3. Physics-aware AI is emerging as a critical infrastructure layer. Generic AI models are increasingly insufficient for critical infrastructure environments; value is shifting toward architectures that embed physics, simulation, operational constraints, and engineering logic directly into AI-driven decision systems.
4. Synthetic data is becoming operationally viable. Advances in simulation, synthetic data generation, and hybrid modeling are beginning to reduce dependence on perfectly structured historical datasets.
5. Infrastructure bottlenecks are becoming AI problems. From compute capacity and cooling to transmission and grid management, energy systems are rapidly emerging as one of the primary constraints to AI scaling itself.

Thank you to the utilities, TSOs, energy infrastructure operators, industrial corporates, corporate venture groups, and AI-native companies who joined this year's Industrial Advisory Committee and contributed to the discussions shaping the future of AI across industrial systems and critical infrastructure.