This is AI's version of WW2 food rationing. When rationing occurs, it means that the thing being rationed (food/compute capacity) is priced far below what the market would bear. In WW2, there was severe food shortages and rationing in the UK. As a result, not only did food prices rise, but so did land prices due to government attempts at boosting food production. You can map this WW2 analogy onto AI: - Food: AI inference, tokens, or access to the model. - Farms/food producers: AI labs like Moonshot, OpenAI, Anthropic etc. - Land: data centers and fab capacity. - Tractors: GPUs, memory and networking/optics etc. - Food rationing: subscription pauses, waitlists, usage caps, and rate limits. Ultimately, since food (model access) is being rationed, we need more land (DC capacity), meaning we need more tractors (GPUs, memory etc) to actually farm the land (fit the DCs). This is why it's been funny, yet shambolic, seeing clickbaity articles from reputable outlets claiming Kimi K3 to be another "DeepSeek moment".