---
title: "The Real Cost of Artificial Dreams"
description: "5 source-backed AI claims from The Real Cost of Artificial Dreams, with key statistics, context, and the original AI Adopters Club source."
url: "https://kbanc.com/claims-library/the-real-cost-of-artificial-dreams"
source: "https://aiadopters.club/p/the-real-cost-of-artificial-dreams"
date: "2025-07-02"
topics: ["strategy", "business"]
generated: "2026-08-31"
---

# The Real Cost of Artificial Dreams

By Kamil Banc | July 2, 2025

## Claims

1. **AI's Massive Energy Appetite** (source summary): The AI boom requires adding half to 1.2 times the UK's annual energy consumption within five years.
2. **Exploited Kenyan Data Workers** (source summary): Kenyan workers were paid two to three dollars per hour to review toxic content for AI training.
3. **Data Center Power Surge** (source summary): Deloitte projects data center electricity demand could double to approximately 1,000 TWh by 2030.
4. **OpenAI's Mission Shift** (source summary): OpenAI began as a nonprofit in 2015 before adopting a capped-profit structure concentrating control.
5. **Efficient Tiny AI Alternative** (source summary): Tiny AI models like MiniBERT run 9.4x faster while being 72 percent smaller than original models.

## Evidence

### Quote
> "The AI revolution everyone talks about isn't inevitable. It's a series of choices made by people with specific interests." - Kamil Banc

### Key Statistics
- **22% of U.S. households' electricity**: MIT analysis estimates AI could consume electricity equal to 22 percent of U.S. household usage, mostly fossil-fuel powered.
- **$2-3 per hour**: Wages paid to Kenyan workers reviewing toxic and explicit content to train AI models, causing psychological trauma.
- **~1,000 TWh by 2030**: Deloitte projects data center electricity demand could double by 2030, equivalent to Japan's entire energy consumption.
- **$1 trillion**: Goldman Sachs estimate of AI spending while questioning economic returns, with some analysts calling it a bubble.

## Context
Kamil Banc, an AI practitioner who works with the technology daily, grounds his critique in documented sources including MIT, Deloitte, Yale, Goldman Sachs, and the IMF rather than speculation. The article traces AI's origins, environmental footprint, labor practices, and market structure to argue that current industry direction reflects deliberate choices rather than technological inevitability. For business leaders, the practical takeaway is to scrutinize AI adoption claims against real infrastructure, labor, and financial costs. Practitioners can also consider efficient alternatives like tiny, on-device AI models that deliver strong performance with a fraction of the energy and infrastructure demands. The piece functions as a due-diligence framework for evaluating AI investments beyond marketing rhetoric.

## Source
- Original: [The Real Cost of Artificial Dreams](https://aiadopters.club/p/the-real-cost-of-artificial-dreams)
- Cite: kbanc.com/claims-library/the-real-cost-of-artificial-dreams

## Primary Evidence
- [half to 1.2 times the UK's entire annual energy consumption](https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/) (technologyreview.com; supports claim 1)
- [systematic exploitation disguised as global opportunity](https://www.noemamag.com/the-exploited-labor-behind-artificial-intelligence) (noemamag.com; supports claim 2)
- [CBS News found workers describing themselves as "overworked, underpaid, and exploited"](https://www.cbsnews.com/news/labelers-training-ai-say-theyre-overworked-underpaid-and-exploited-60-minutes-transcript/) (cbsnews.com; supports claim 2)
- [Deloitte projects data center electricity demand could double to ~1,000 TWh by 2030](https://www2.deloitte.com/us/en/insights/industry/technology/technology-media-and-telecom-predictions/2025/genai-power-consumption-creates-need-for-more-sustainable-data-centers.html) (www2.deloitte.com; supports claim 3)
- [OpenAI started as a nonprofit in 2015](https://clsbluesky.law.columbia.edu/2024/03/05/the-untold-nonprofit-story-of-openai/) (clsbluesky.law.columbia.edu; supports claim 4)
- [Infosys research shows tiny AI models like MiniBERT run 9.4x faster](https://www.infosys.com/about/knowledge-institute/documents/tiny-ai-sustainable-digital-future.pdf) (infosys.com; supports claim 5)