Projected spending from hyperscalers and Goldman Sachs estimates points to a sharp buildout in AI infrastructure, with implications for power markets, semiconductors and crypto-linked data center competition.
AI-related capital expenditures could approach $1 trillion by 2026 as recursive self-improvement, or RSI (AI improving its own capabilities), moves closer to the center of the industry’s long-term thesis. The projection comes as combined 2026 capex guidance from Alphabet, Microsoft, Amazon and Meta reaches roughly $725 billion, up 77% from about $410 billion in 2025, while Goldman Sachs estimates annual AI capex at around $765 billion for 2026 and about $7.6 trillion cumulatively from 2026 through 2031. DeepMind CEO Demis Hassabis has described RSI as a key focus for 2026, alongside multi-agent systems and traditional scaling, as researchers look beyond AGI (AI matching human-level general reasoning) toward ASI (AI exceeding human capability across domains). The spending surge matters beyond Big Tech balance sheets because AI data centers are already competing with Bitcoin miners for power in markets such as Texas and the Nordics, potentially tightening demand for energy, chips and networking infrastructure.