DeepMind Exec Predicts Trillion-Dollar AI Capex Driven by Recursive Self-Improvement
A top Google DeepMind executive is predicting that AI-related capital expenditures will reach the trillion-dollar mark by 2026, driven in part by a concept called recursive self-improvement (RSI). RSI occurs when an AI system becomes good enough to improve itself without human intervention. This development has significant implications for industries such as crypto and digital infrastructure.
According to Goldman Sachs' estimates, the cumulative capex forecast for AI-related spending from 2026 through 2031 is approximately $7.6 trillion. The current collective investment guidance from major hyperscalers like Alphabet, Microsoft, Amazon, and Meta points to a 77% increase in spending, from around $410 billion in 2025 to roughly $725 billion in 2026.
Alphabet has revised its own capex forecast for 2026 to between $195 billion and $205 billion. The executive at Google DeepMind emphasizes the importance of RSI as a key focus area for development in 2026, alongside multi-agent systems and traditional scaling. This concept is seen as crucial not only for achieving Artificial General Intelligence (AGI) but also for pushing toward even more advanced Artificial Superintelligence (ASI).