AI Infrastructure

90% of AI Initiatives Will Fail -- And Infrastructure Is Why

90% of AI Initiatives Will Fail -- And Infrastructure Is Why

BlackRock's AI ETF increased overweight on infrastructure (chips, data centers, networking) during AI selloff — rotating away from model vendors toward picks-and-shovels plays with more predictable cash flows

Saudi Arabia, India, EU, and US are all treating AI infrastructure as national/sovereign infrastructure — fragmenting AI before it matures, unlike cloud which consolidated globally around US hyperscalers

Nvidia restructured its OpenAI investment from $100B long-term to $30B milestone-based — a 70% reduction that signals execution risk assessment from the company with the best visibility into AI infrastructure economics

Nvidia restructured its OpenAI investment from $100B to $30B milestone-based — a 70% reduction reflecting execution risk concerns from the company with the deepest visibility into AI infrastructure economics

Hyperscalers spending $660-690B on AI infrastructure in 2026 while AI-driven cloud growth decelerates and investors punish capex announcements — the spend and the skepticism are simultaneous

EU, Saudi Arabia, India, and the US Stargate project are all building sovereign AI infrastructure — fragmenting AI by policy before the technology matures, unlike cloud which consolidated around US hyperscalers

Microsoft's free Office agents + Meta's trillion-dollar DC + TSMC 37% capex increase = simultaneous democratization and concentration at different stack layers