YouTube Summarize
Open YouTube

How AI Became More Expensive Than The Workers It Replaced

Economy Mediaautoenpublicupdated

Read in about 3 minutes instead of watching 11 minutes.

AI adoption and job displacement

  1. The video argues that after ChatGPT launched in 2022, companies rapidly adopted AI because it appeared cheap, efficient, and capable of replacing work across coding, media creation, operations, and customer service.
  2. AI adoption was linked to layoffs and workforce reductions across tech companies and service industries, including examples such as tech firms and fast food drive-thrus.
  3. The central question presented is how a technology promoted as cheaper than workers became expensive enough to challenge that assumption.

Token maxing inflated demand

  1. From 2023 onward, companies pushed employees to use AI tools, and token usage became a performance signal or status marker.
  2. Employees allegedly began using excessive tokens to look more productive, a behavior described as token maxing.
  3. The transcript says token-heavy usage expanded from text prompts into video, image, code, and minor engineering tasks, driving costs higher.
  4. Statements from executives, including Nvidia's CEO, are presented as encouraging heavy token spending, which may have inflated demand beyond real business need.

Infrastructure constraints and rising prices

  1. The video claims data center construction delays and component shortages limited AI infrastructure supply while demand continued to rise.
  2. It questions whether reported AI demand reflects real enterprise adoption or inflated internal usage metrics.
  3. The transcript says token maxing, data center delays, and global AI adoption contributed to significant token price increases.
  4. It cites average LLM token costs rising from $1.01 per million tokens in December 2025 to $2.12 per million tokens by May 2026.

Enterprise budgets under pressure

  1. At large companies, small per-token price increases can translate into tens or hundreds of millions of dollars in monthly or yearly spending.
  2. The transcript gives examples such as AT&T using nearly 8 billion tokens per day and Meta allegedly consuming nearly 60 trillion tokens in a month.
  3. Anthropic and OpenAI are described as benefiting from enterprise AI adoption, with Anthropic revenue projected to rise sharply in 2026.
  4. Companies are reportedly exhausting AI budgets quickly, making it less clear that AI remains cheaper than human labor.

Companies reconsider AI costs

  1. Microsoft reportedly instructed engineers to stop using Anthropic coding tools, with high Claude Code costs presented as the main reason.
  2. The video says other companies are also pulling back as finance departments struggle to justify fast-growing AI expenses.
  3. It cites forecasts that AI could become a large share of enterprise technology spending, with AI agent and model spending projected to reach $680 billion by 2027.
  4. The transcript argues that if OpenAI and Anthropic face shareholder pressure after public listings, token prices may rise further to support profitability.

AI versus human labor

  1. The video concludes that AI agents are already comparable to human labor costs in some tasks and more expensive in others, especially call centers.
  2. It says AI remains cheaper for some coding tasks, but offers smaller savings in data entry and may be less cost-effective than humans in customer support work.
  3. The final point is that companies may soon reassess whether paying rising AI costs makes sense compared with rehiring or retaining human workers.