General Motors pitched an unconventional solution to the energy crisis sparked by AI's data-center demands: vehicle-to-grid (V2G) technology. The thesis: an EV fleet collectively stores and discharges power to stabilize grid strain during peak AI inference loads.
The proposal acknowledges a real problem. Training and running large language models consume electricity at scales comparable to small nations. As AI inference scales, grid operators face bottleneck challenges.
GM's pitch is clever but speculative. V2G requires coordinated infrastructure rollout, regulatory changes, and consumer adoption of bidirectional charging. The timeline is 5-10 years minimum.
More immediately, data centers are shifting to renewable partnerships and efficiency optimizations (quantization, pruning). V2G is a future lever, not an immediate solution.
