The Man Who Calls BS On AI: AI Is The World’s Greatest SCAM, And They All Know It! | Ed Zitron
The Diary of a CEOAI Summary
→ WHAT IT COVERS Ed Zitron, a 16-year tech industry veteran, argues that generative AI represents the largest financially unsustainable technology push in history. He presents data showing OpenAI lost $20.9 billion in 2024, that 70% of AI revenues flow between just two unprofitable companies, and that trillion-dollar infrastructure spending has no credible path to profitability or the transformative productivity gains being promised. → KEY INSIGHTS - **The Token Subsidy Problem:** AI companies are selling services at a fraction of their actual cost. A $200/month ChatGPT subscription allows users to burn up to $14,000 worth of compute tokens, and a $20/month plan allows up to $400. This means companies are subsidizing users at ratios potentially exceeding 40-to-1. When Microsoft attempted to shift enterprise clients onto cost-reflective pricing in early 2026, Uber burned through its entire annual token budget in just three months. - **Circular Revenue Structure:** Approximately 70% of all AI revenues across major players flow exclusively through OpenAI and Anthropic — both of which are funded by the same companies claiming AI revenue growth. Amazon sent $50 billion to OpenAI and $5 billion to Anthropic; Google sent $10 billion to Anthropic. Microsoft generated roughly $34.3 billion in AI-related revenue in fiscal year 2026, but $24.1 billion came directly from OpenAI — while spending $115 billion in capital expenditures that year. - **Benchmark Manipulation as Distraction:** When AI companies cannot demonstrate real-world productivity gains, they redirect attention to internally defined benchmarks. Hallucination leaderboards measure only simple summarization tasks, where error rates dropped from 21.8% to 0.7% over four years. However, complex tasks, multi-step reasoning, and code refactoring remain unreliable. Readers should demand task-specific, real-world performance data rather than accepting benchmark scores as proof of general capability improvement. - **Non-Consensual Adoption Inflates Usage Stats:** The rapid user adoption figures cited as proof of AI value — 100 million ChatGPT users in 60 days — are distorted by forced integration. Google embeds Gemini into Search and Docs by default; Microsoft pushes Copilot into Word; Amazon inserts AI into shopping. Three years of media coverage warning professionals they will be fired without AI adoption further coerces usage. Organic, paid-at-cost adoption would produce substantially lower numbers. - **Software Quality Is Declining Under AI Coding Pressure:** Industry data shows tech outages and software bugs have increased as AI-assisted coding scales. GitHub experiences frequent downtime. Amazon Web Services went down multiple times in one year linked to AI coding tools. The mechanism is compounding: developers who rely on AI-generated code review subsequent outputs less rigorously, allowing errors to accumulate. Businesses mandating AI tool usage without verification protocols are systematically introducing technical debt and instability. - **Job Displacement Claims Lack Economic Evidence:** OpenAI's own internal study found zero statistical correlation between organizational AI token spending and revenue per employee. Actual documented displacement is concentrated in contract creative roles — translators, transcribers, and art directors — where employers were already seeking the cheapest available option. White-collar disruption in law, finance, and management remains anecdotal, driven largely by senior professionals who do not perform the detailed analytical work they claim AI is replacing. - **Infrastructure Overbuild Has No Post-Bubble Recovery Path:** Unlike the fiber optic dark cable overbuild of the dot-com era — which eventually found demand as Internet usage grew organically — AI GPU data centers serve a single, narrow function: running generative AI inference and training. Sightline Climate identified 190 gigawatts of data centers in planning stages, requiring $1.6 to $3 trillion in annual demand to justify. Current total global AI revenue outside OpenAI and Anthropic sits below $22 billion annually, making utilization projections mathematically implausible. → NOTABLE MOMENT Zitron describes how Google's search quality deliberately degraded after an internal 2019 code yellow meeting, where ad-focused leadership overruled engineers who warned that increasing search queries required giving users worse answers. The executive responsible was subsequently placed in charge of Gemini development — suggesting AI integration into Google Search was driven by ad revenue logic rather than user value. 💼 SPONSORS [{"name": "NetSuite by Oracle", "url": "https://netsuite.ai/bartlett"}, {"name": "Fiverr Pro", "url": "https://pro.fiverr.com"}, {"name": "Salee eSIM", "url": "https://apps.apple.com/salee"}] 🏷️ Generative AI Economics, AI Bubble, Tech Industry Accountability, AI Job Displacement, Software Quality Decline, AI Infrastructure Spending, OpenAI Financial Losses
