Elon Musk has repeatedly highlighted energy (electricity/power supply) as the next major crisis or bottleneck for AI scaling, especially in recent statements from late 2025 into 2026.
He has warned that AI compute (chips and hardware) production is advancing so rapidly that we’re approaching (or already hitting) a point where there will be more chips available than electricity to power them.
Musk has stated variations of “We’re very soon going to be producing more chips than we can turn on,” emphasizing that electricity generation and grid capacity – not chips, transformers, or algorithms – are now the primary constraint for AI growth.
He predicted challenges with power generation hitting by mid-to-late 2025 or into 2026, with AI’s explosive demand outpacing infrastructure. By early 2026, reports and his posts suggest this is materializing, with surplus compute but unusable without enough power.
Electricity is becoming “the most valuable commodity on Earth,” shifting focus from silicon shortages to energy wars. He notes that global grids (especially in the US) aren’t scaling fast enough, potentially leading to idle hardware, higher costs, blackouts, or slowed AI progress. China, with massive energy additions, could gain an edge in the AI race due to this.
His proposed solutions and context:
Massive scaling of clean/abundant energy (solar, nuclear, etc.).
Long-term vision: Orbital/space-based data centers powered by constant solar energy (no batteries needed, natural cooling in vacuum). He claimed at Davos 2026 that space could become the lowest-cost place for AI within 2–3 years, with Starship enabling huge solar-powered AI satellite constellations (potentially hundreds of GW per year). This ties into xAI/SpaceX integration ideas.
Near-term: xAI’s own facilities (like Colossus in Memphis) have faced real-world energy issues, relying on temporary gas turbines (drawing environmental criticism and regulatory scrutiny) while pushing for grid upgrades or dedicated power.
This aligns with industry trends: AI data centers are projected to consume enormous electricity (potentially rivaling entire countries or large percentages of US household use by late this decade), driving investments in nuclear restarts, new plants, and alternative sites. Musk sees this as part of a path to abundance via AI/robotics/solar loops, but warns of short-term crises if energy doesn’t catch up –possibly disrupting everything from training massive models like future Grok versions to broader economic/tech progress.
Musk views energy problems as the current/next big crisis in AI, far beyond previous chip bottlenecks.
/Х/