AI as Public Infrastructure: Three Insights from AI House Davos during the World Economic Forum

AI as Public Infrastructure: Three Insights from AI House Davos during the World Economic Forum

At the AI House Davos during the World Economic Forum this January (2026), one theme surfaced repeatedly: AI is no longer just a technology conversation, it is rapidly becoming an infrastructure question.

Having moderated a panel on “Large-Scale AI as Public Infrastructure” with leaders from academia, industry and frontier AI labs, three implications stood out that business leaders and policymakers should keep top of mind.

1. AI is moving from application layer to infrastructure layer

Leaders should no longer view AI merely as a collection of technology applications. Increasingly, it is taking on the characteristics of infrastructure: capital-intensive, systemic and foundational across sectors. This shift materially changes the strategic lens. Large-scale AI demands longer investment horizons, different governance models and much closer public-private coordination than traditional digital initiatives.

Implication: leaders should start treating AI capacity as a strategic infrastructure question, not just an innovation initiative.

2. Governance will be hybrid by necessity

The panel converged on a nuanced but realistic view: neither purely state-led nor purely market-driven models will be sufficient. Large-scale AI will likely evolve through hybrid governance arrangements involving:

• private sector innovation
• public sector oversight
• public-private partnerships
• open-source ecosystems

The example of AlphaFold illustrated the power of combining public data foundations with private-sector execution.

Implication: the future of AI governance will be architected ecosystems, not single-actor control.

3. Access to compute is emerging as a policy issue

Another theme gaining urgency is equitable access to compute. As AI capabilities concentrate around large-scale infrastructure, questions of access across regions, sectors and research communities are moving from technical debates into the policy arena. Several panelists emphasized that public-sector leaders in particular need more hands-on exposure to AI systems to make informed decisions about their governance and deployment.

Implication: compute capacity is becoming a strategic policy lever, not just a technical resource.

The window for shaping AI governance is still open, but narrowing. Perhaps the most important meta-signal was this: we are still early enough to shape the trajectory. Concrete decisions on governance models, funding structures and infrastructure access in the next one to two years will have outsized long-term impact. If AI is indeed evolving into a foundational infrastructure layer, then governance and access are ultimately design choices, not inevitabilities.

For boards, policymakers and executive teams, the task now is less about reacting to AI applications, and more about deliberately shaping the infrastructure layer that will underpin the next wave of digital transformation.