AI revolution is '50x bigger' than the dotcom boom: SoftBank's Masayoshi Son to CNBC

SoftBank plans to invest 75 billion euros ($87 billion) in AI infrastructure in France, including 5 GW of AI data center capacity and 3.1 GW in Hauts-de-France by 2031. Masayoshi Son said AI is potentially 50x bigger than the dotcom revolution, framing the buildout as a major long-term growth opportunity. The initiative includes a partnership with Schneider Electric to develop a large-scale industrial production hub in Dunkirk.

Analysis

The signal here is not the rhetoric, but the sequencing: SoftBank is trying to lock up the bottlenecks in the AI buildout before the market fully prices a multi-year European capex wave. That matters because the first beneficiaries are not the model companies, but the picks-and-shovels stack—power management, cooling, switchgear, grid interconnect, and industrial construction—where demand is less elastic and lead times are already long. If even a portion of this program is real and repeatable, it creates a reference project that can crowd in follow-on capital from hyperscalers and sovereign-linked funds across Europe.

The second-order effect is a regional re-rating of the power and infrastructure ecosystem. Large AI data center campuses are fundamentally constrained by electricity access, permitting, and execution capacity, so the winners will be firms that can convert megawatts into delivered uptime, not just vendors with AI exposure in their labels. Schneider is an obvious beneficiary, but the broader basket likely includes European grid equipment, HVDC, transformers, and industrial automation names; the more interesting trade is that some software-heavy AI beneficiaries may underperform if capital intensity continues to rise faster than monetization.

The risk is that this story becomes a long-duration expectations trade with near-term disappointment risk. A 2031 buildout timeline means headlines can stay positive while order conversion, permitting, and financing cadence become the real gates; any slippage in power availability, local opposition, or macro funding conditions could compress the multiple on the entire AI infrastructure theme for months. The contrarian read is that investors may already be overbidding the “AI in Europe” narrative while underestimating how much of the value accrues to regulated utilities and industrials rather than to pure AI plays.

Key Decisions for Investors

• Long Schneider Electric (SU FP) on pullbacks over the next 1-3 months; target a 12-18 month rerating as AI power/cooling spend converts into backlog growth. Risk: the market treats this as one-off PR rather than repeatable order flow.
• Pair trade: long European electrical infrastructure basket (SU FP, ABBN SW, RHM GY only as industrial execution proxy) vs short a high-multiple AI software basket over 3-6 months. Thesis: capital intensity and power bottlenecks push near-term value capture toward hardware/infrastructure, not software. Risk/reward favors a modest basket spread with defined stop if software margins expand.
• Buy call spreads on utility/grid equipment names with European exposure for 6-12 months. The asymmetric catalyst is rising grid capex and data-center interconnect demand; risk is delayed permitting, which can defer but not eliminate demand.
• Avoid chasing pure-play AI infrastructure hype names without visible backlog or power access. Use any 10-15% post-news enthusiasm to fade crowded longs where the path to revenue is 24+ months out.