Two stories landed this week that, read together, form a single damning thesis: the AI boom is not free, and someone always pays. PJM Interconnection, the largest US power grid, is buckling under data center demand. Meanwhile, Cloudflare, Coinbase, Upwork, and others are citing AI as justification for mass layoffs. The machine is hungry. It consumes watts and workers with equal indifference.
AI's Environmental Footprint Is No Longer Abstract
A 2026 paper by Lambert and Luccioni in arXiv CS.CY delivers the receipts: a full life-cycle review of AI's environmental costs, from chip fabrication to inference at scale. The findings aren't metaphorical. The carbon embedded in a single large model's training run rivals the lifetime emissions of multiple cars. PJM's grid operators are not dealing with projections anymore. They're dealing with transformers catching fire. The irony is structural: the AI that promises to optimize energy systems is itself the reason those systems are failing. TurboFund's Signal Report tracked 6 AI/LLM investor signals this week alone, with Sam Altman publicly evangelizing GPT-5.5's leap in capability. Every model upgrade compounds the load.
Layoffs as Externalized Cost
Truecaller's 44% ad revenue decline and subsequent 70-person cut is a different story but the same chapter. When platforms automate their way to efficiency, the displaced workers don't appear on any carbon balance sheet. The LLMorphism paper by Valerio Capraro argues that humans are increasingly modeling their own cognition on LLMs, flattening self-conception to match the tools. The workforce isn't just shrinking around AI. It's reorganizing its self-image around it. The grid strain and the job cuts are two symptoms of one metabolic disorder: an economy ingesting AI faster than it can process the waste.