The Man Who Calls BS On AI: They’re LYING About AI, 2027 Is When It All Breaks! | Ed Zitron
- Core Thesis: Ed Zitron characterizes the generative AI industry as a "con" and a "roccoco bubble" (Rock-con bubble), arguing that the technology is currently a half-arsed, unprofitable machine that has been marketed as magical and transformative to mislead the public, journalists, and investors.
- Financial Reality: Major AI players are running at "horrifying losses," with OpenAI losing $20.9 billion in a single year, while most revenue cited by public companies (Microsoft, Google, Amazon) stems from existing businesses rather than AI.
- Subsidized Demand: Current adoption is described as the "largest non-consensual push of technology in history," where users are forced to use AI tools by interface design (e.g., Google Docs, Word, Amazon), and the vast majority of revenue comes from two unprofitable companies (OpenAI and Anthropic) subsidized by their investors.
- Capital Expenditure vs. Revenue: Companies have spent over $1 trillion in capital expenditures on data centers and GPUs, yet the actual revenue generated from AI services is estimated to be only a small fraction of that investment (e.g., Microsoft's $24.1 billion AI revenue against $115 billion in capex).
- Cost Structure: Enterprise customers face unsustainable costs; analysis suggests a $200/month ChatGPT subscription can burn $14,000 worth of tokens, while companies like Uber burned their annual token budget in three months before realizing the costs were not justified by productivity gains.
- Myth: Economic Growth: The AI industry is not creating genuine economic growth; the "growth" is largely an artifact of speculative investment in GPU hardware and the circular flow of money between cloud providers (Microsoft, Google, Amazon) and AI labs (OpenAI, Anthropic).
- Myth: US vs. China Race: The narrative of a US vs. China AI race is dismissed as "insane," with no clear strategic objective beyond maintaining market share, noting that Chinese companies have already acquired Nvidia GPUs and possess comparable models.
- Myth: Job Replacement: There is no economic data supporting the claim that AI will replace all or most human jobs; studies show no correlation between AI spending and revenue per employee, and disruption is currently limited to "contract labor" rather than core white-collar roles.
- Technological Limitations: The technology is described as a "directionless egregore of capitalism" that requires a "Rube Goldberg machine" of human intervention (prompt engineering, harnessing) to function, rather than being the autonomous "magic" promised.
- Quality Degradation: The widespread use of AI coding tools has led to a decline in software quality, with increased downtime for major platforms like GitHub and Google Docs due to the volume of unvetted, buggy code being pushed into production.
- Hallucination Rates: While benchmarks show hallucination rates on simple tasks have dropped (e.g., from 21.8% to 0.7%), these tests are criticized as rigged and do not reflect reliability in complex, high-stakes environments like financial modeling or medical transcription.
- Environmental Impact: The industry's infrastructure is described as "environmentally destructive," utilizing massive gas turbines in sensitive areas (e.g., Louisiana, Vineland, NJ) and condensing power consumption at rates that dwarf entire cities (e.g., Stargate Abilene data center vs. City of Bristol).
- Market Manipulation: Public companies use undefined metrics like "annualized run rate" (which can mean month × 12 or week × 13) to obscure the lack of tangible AI revenue, engaging in "financial engineering" to prop up stock prices.
- Risk Narrative Shift: The industry narrative has pivoted from "existential danger" (extinction) to "age of abundance" (unlimited growth) as a marketing tactic to induce fear or greed, with CEOs like Sam Altman and Dario Amodei accused of using these narratives to secure investment.
- Venture Capital Bubble: The AI sector has absorbed over 50% of venture capital funding in recent years, with valuations based on "paper gains" and speculative futures rather than actual profitability, creating a fragile ecosystem where most startups will fail.
- Predicted Collapse (2027): Zitron forecasts a collapse of the AI bubble around 2027, triggered by OpenAI's inability to secure further funding or successfully IPO, leading to a cascading "tech depression" that could severely impact retirement funds and the broader stock market.
- Systemic Failure: The business model is deemed unsustainable because the cost of training and inference (energy, compute) is rising, while the revenue from AI services is not scaling proportionally, meaning the "progress" of the technology is entirely dependent on continuous capital injection.
- Regulatory Failure: The lack of regulation is attributed to a "cult-like worship" of the wealthy and a failure of the SEC and media to demand transparency, allowing companies to operate without accountability for the economic and environmental costs they impose.
- Consumer Cost: Unlike the "dot-com bubble" where demand eventually materialized, the AI bubble is fueled by subsidies; once users are forced to pay the "honest cost" (per token), adoption is predicted to plummet as the value proposition disappears.
- Corporate Desperation: Major tech firms (Microsoft, Amazon, Meta) are investing heavily in AI not because it is currently profitable, but because their core legacy businesses are plateauing and they need a "new growth narrative" to satisfy stock markets.
- Psychological Manipulation: The industry relies on "fear-mongering" (e.g., "if you don't use AI, you will be left behind") and "frenzy" tactics to force adoption, creating a professional culture where employees are penalized for not using AI, regardless of its utility.
- Hardware Bottleneck: Despite massive spending, there are no significant hardware breakthroughs (Moore's Law for GPUs has stalled) to drive down costs or increase efficiency, meaning the only way to improve models is by spending exponentially more money.
- Conclusion on Value: The ultimate value of AI will likely be limited to low-stakes tasks (tech support, basic coding snippets), while the high-value, creative, and contextual work remains uniquely human; the "commoditization" of content generation makes human curation and "taste" the new scarce asset.
- Future Outlook: Zitron advises investors and the public to be skeptical of AI promises, live in cash, and avoid over-exposure to tech stocks, warning that the "magnificent seven" valuations are built on a foundation of speculative debt and false growth narratives.