Trip Insights: Europe and the United States - May & June 2026

The trip argues that the next investable phase of AI is shifting from crowded chip exposure toward power infrastructure, mission-critical software and other bottleneck enablers with clearer earnings visibility.

• The analyst met more than 50 companies at four investor conferences across Europe and the US over three weeks in late May and early June 2026, finding that AI spending remains robust but that the key questions are sustainability, bottlenecks, value capture and free-cash-flow conversion.
• The binding constraint is moving below GPUs toward electricity, grid connections, advanced packaging, memory and data-centre power systems; NVIDIA-native 800V HVDC racks are expected to begin first delivery in 2H27, while only a small minority of facilities appear ready for the architecture at launch.
• Memory remains attractive but more dependent on pricing, tight supply and mix than on unit growth: HBM is still the tightest segment, NAND pricing and related equities have rallied sharply year to date, and meaningful greenfield capacity appears to be a CY27 story.
• ServiceNow and Autodesk illustrate the preferred software model, in which AI is embedded in workflows and supported by proprietary data, distribution, governance and measurable customer ROI rather than sold as a standalone feature.
• New or prioritised ideas include Applied Materials, Prysmian, Robinhood and Marex: Prysmian has an approximately €17bn Transmission backlog and FY2025 adjusted EBITDA of approximately €2.4bn, while Robinhood has over $350bn of assets, approximately 4.5m Gold subscribers and around 75% of support tickets automated.
• Marex reported FY25 revenue growth of 27% to approximately $2.0bn, adjusted profit before tax growth of 30% to approximately $418m and adjusted ROE of around 30%; Q1 2026 revenue rose 48% and adjusted profit before tax 59%.

Takeaway: Investors should favour power equipment, cables, grid infrastructure, semicap, mission-critical platforms and selected capital-markets infrastructure, while resisting crowded AI-semiconductor multiple expansion and applying strict valuation, balance-sheet and free-cash-flow discipline.