Apple will "watch everything burn" when the AI bubble bursts, according to Ed Zitron, a newsletter writer and podcast host described by Politico as the AI boom's most "acerbic gadfly." Zitron argues that the economics of Large Language Models are fundamentally broken because their costs run contrary to standard software selling models. He notes that LLMs burn tokens at a per-million rate regardless of whether users get desired responses, and AI companies sell subscriptions with vague rate limits that allow users to burn far more in tokens than the subscription cost. SemiAnalysis found users can burn hundreds of dollars on a $20-a-month subscription and thousands on a $200-a-month subscription. Zitron's reporting shows OpenAI lost $20.9 billion on $13.07 billion in revenue in 2025. He claims that if Anthropic and OpenAI believed customers would pay real costs, they wouldn't give away 20 to 40 times the amount of tokens to subscribers. In March, both moved enterprise customers to token-based billing, and Uber spent its entire annual token budget in a quarter, with its COO saying it was getting "harder to justify" AI costs. Sam Altman called it a "huge issue" but declined to say how it might be fixed. Zitron says every AI startup, including Perplexity, Cursor, and GitHub Copilot, is unprofitable because users don't want to pay actual AI costs.