AI Is Too Expensive

2026-05-19 · Source: Ed Zitron's Where's Your Ed At · Field: Finance & Economics — Capital Markets & Investment Management, Corporate Finance & Treasury, Economic Analysis & Policy · Depth: Intermediate, extended

Summary

An analysis published on May 19, 2026, by Ed Zitron argues that AI is not economically viable for most participants, except hardware companies like NVIDIA. Hyperscalers have invested over \$800 billion in the last three years, with plans for an additional \$700 billion in 2026 and \$1 trillion in 2027, requiring trillions in AI-specific revenue to break even. Microsoft alone spent approximately \$100 billion on its OpenAI partnership, representing about 30% of its \$293.8 billion capex since FY2023. Despite reported AI revenue "run rates" (e.g., Microsoft's \$37 billion, Amazon's \$15 billion), actual revenues are minimal compared to these massive investments. OpenAI and Anthropic, major compute consumers, are projected to burn hundreds of billions, with Anthropic spending \$3 to make \$1 of revenue on a compute basis. Enterprise customers like Zillow are exceeding token budgets, with Zillow spending over \$1 million in Q1 2026, potentially reaching \$7M-\$10M for the year, nearly 50% of its \$23 million 2025 net income, while experiencing increased human review and "AI slop." The author concludes that AI's high operational costs, lack of transparent ROI metrics, and reliance on unprofitable AI labs make current investments unsustainable.

Key takeaway

For technology executives and investors evaluating AI strategies, recognize that current AI investments by hyperscalers and enterprises are largely unsustainable. You should demand clear, auditable ROI metrics and transparent cost breakdowns from AI vendors, rather than relying on projected "run rates" or unverified profitability claims. Prioritize measurable outcomes over mere AI adoption targets to avoid significant financial losses and "AI slop" in your operations.

Key insights

AI's current economic model is unsustainable due to massive hyperscaler investments, high operational costs, and unproven profitability of AI labs.

Principles

• AI costs increase linearly with revenue.
• Inference profitability is unproven.
• Compute capacity is a "knife-catching" game.

In practice

• Scrutinize AI token budgets for ROI.
• Demand granular cost data from AI vendors.
• Avoid tying metrics to AI usage targets.

Topics

• AI Economics
• Hyperscaler Investments
• OpenAI
• Anthropic
• Data Center Capex
• Enterprise AI Adoption
• LLM Costs

Best for: Executive, Investor, Consultant

Related on AIssential

• AI Doesn't Have ROI — The AI industry's economic model is fundamentally flawed, built on obscured costs and unproven ROI, leading to unsustainable enterprise spending.
• What is causing OpenAI to lose so much money compared to Google and Anthropic? — OpenAI's high costs stem from its broad strategy, infrastructure choices, and large free user base.
• Premium: AI's Circular Psychosis — The AI economy's perceived growth is an illusion, sustained by circular funding and compute demand from two unprofitable companies.
• The bubble is slowly popping, investment isn't able to keep up — The AI market faces a critical debate regarding its investment sustainability and the long-term viability of current pricing models.
• How Do We Fix This? — Focus AI investments on use cases where returns outpace costs to achieve economic viability and avoid negative unit economics.
• $700 Billion in Capex. $50 Billion in Revenue. AI’s Math Is Broken. — Despite exponential growth, major AI labs struggle with profitability due to costs scaling faster than revenue and rapid commoditization.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Ed Zitron's Where's Your Ed At.