OpenAI CEO Sam Altman warned during a podcast about a potential AI compute oversupply within two to three years. Several factors could trigger this reversal: increasingly efficient AI models, declining unit costs for intelligence, abundant cheap energy, and breakthroughs enabling local model execution. This outlook contrasts with current reality where supply remains constrained, yet companies are investing billions in data centers expected to come online in two to four years. Meta has already begun selling excess compute capacity as of mid-2026. Should this oversupply materialize, data center utilization rates would decline, pressuring margins for cloud providers and undermining investment narratives for crypto infrastructure projects built on assumptions of persistent compute scarcity.
Source: Read the original article
