All the major AI models are burning money (OpenAI, Anthropic, xAI, Google, Meta etc) Examples: OpenAI burning $9B in 2025, $15B-$27B expected in 2026. Anthropic burning $5B in 2025, $5-10B expected in 2026 Compute is the technical bottleneck but energy is included in the price bottleneck. Currently models are priced unprofitably, seeking to grow market share, train new models and scale infra. It’s important to note that infra also must be maintained, and is not a one time cost. The high performance GPUs in AI infra have roughly a ~3 year lifespan before needing to be replaced, so CAPEX costs will come on cycles This is why an argument has been made that AI companies will eventually be owned by the largest tech companies (Meta, Google, Apple, Microsoft, Amazon) because these tech companies have the means to subsidize AI costs/growth with their other businesses, while OpenAI and Anthropic must continue to raise and sell more and more of their company to sustain their burn Long term this also means either costs have to come down, or pricing for AI subscriptions/requests have to come up. Likely both will occur Right now we’re truly in a golden age where retail users and developers can access AI requests below cost. In the future, the applications of AI will narrow to applications where the efficiency or revenue gained by active use of AI outweighs the costs. Many applications of AI will fail this test, certain AI models will fail to survive the competition, pricing will increase, the gap of who can access and use AI will widen Now is the time to lock in and take part in the revolution, and capture value while it’s artificially cheap