AI Is Too Expensive

This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).

A detailed financial critique arguing that AI is fundamentally uneconomical at every level. Hyperscalers have invested over $800 billion in AI infrastructure with no clear path to recouping costs — Microsoft, Google, Amazon, and Meta would need to generate $100+ billion each in AI revenue within a year just to avoid catastrophic losses. OpenAI and Anthropic burn billions while lacking any credible path to profitability, with gross margins declining and compute costs spiraling. Enterprise customers like Zillow, Stripe, and ServiceNow are burning through annual AI token budgets in months with no measurable ROI. The piece also examines how Anthropic obscures cost telemetry from enterprise customers, making budgeting nearly impossible, and how corporate pressure to 'use AI for everything' is creating technical debt and organizational chaos without measurable productivity gains.

#ai

#llm

#openai

#anthropic

May 19•45m read time•From wheresyoured.at

Table of contents

AI Is Too Expensive To Ever Pay Off Hyperscalers’ Capex Investments AI Is Too Expensive for Any AI Lab To Overcome Anthropic and OpenAI Need To Make Or Raise More Than $1.25 Trillion In The Next Four Years How Does Anthropic or OpenAI Turn Profitable?AI Is Too Expensive For Its Customers Zillow Is AI Chernobyl AI Token Burn Is Increasing — and ROI Is Impossible To Measure The Revenge of the Business Idiot