michelle weaver

A 40 GW power shortfall and critical shortage of specialized electricians will keep data center supply in check for years, with political pushback (NIMBYism, state moratoriums) adding further constraints.

Structural shortages of electricians (labor) and a 40GW power shortfall through 2028 will keep data center supply growth checked, maintaining high compute prices and preventing oversupply.

Power infrastructure cannot keep pace with AI demand, with a 40GW shortfall through 2028 and innovative solutions (bitcoin site conversion, fuel cells) still leaving 10-20% gap; labor shortage of specialized electricians is the number one bottleneck.

63% of companies already use both closed and open models; switching costs and employee skill stickiness favor closed models, while specialized/cheap use cases favor open-weight, but geopolitical concerns over overseas models create hesitation.

Enterprise AI adoption is early but sticky: 25% of S&P 500 companies now quantify AI benefits (up from 14%), median user spends <$11/month on tokens, and power users are rationalizing spend rather than maximizing consumption.

Enterprise token spend is declining from 'tokenmaxxing' peaks (median user <$11/month), while companies adopt both open and closed models (63% use both) with stickiness favoring incumbent closed models for existing workflows.

Data center growth faces three buckets of local opposition: power bill impacts, environmental concerns (water/air), and quality-of-life issues; brownfield redevelopment and off-grid solutions are emerging mitigations, but New York's moratorium signals rising political risk.

US companies hesitate to fully commit to overseas open-weight models due to regulatory risk (potential bans) and business continuity concerns, while state-level moratoriums (e.g., New York) add political uncertainty to data center builds.