Consumer electronics, energy, and water prices are rising, and the ongoing AI gold rush, built on a mountain of debt, is contributing to these costs. Hundreds of billions in AI investment have been racked up for 2026, with borrowed money spent on data center capacity, even though consumers would likely prefer a cheap Mac over an AI subscription. Memory vendors are shifting capacity to produce high-value, high-bandwidth memory for server farms at the expense of consumer electronics suppliers. Sequoia’s David Cahn estimates the AI revenue gap between infrastructure expenditure and the revenue to justify it has already fallen $600 billion a year short. Some analysts think capex growth in the sector could halt by mid-2027, and memory vendors will feel reduced forward orders by the end of 2026. Apple has already figured out how to run tasks like web searches, writing assistance, and document summarization on device, which barely stretch the capabilities of large language models (LLMs) currently sitting in data centers.