The Man Who Calls BS On AI: AI Is The World’s Greatest SCAM, And They All Know It! | Ed Zitron

Aug 27, 2026 Transcript ↗
Overview

Tech critic Ed Zitron argues that the generative AI industry is a "con" built on overhyped promises and unsustainable financial models. He explains why major AI companies are operating at massive losses, the true costs of AI, and the potential for an economic depression when the AI bubble inevitably pops.

At a Glance
6 Insights
2h 27m Duration
21 Topics
7 Concepts

Deep Dive Analysis

Generative AI as a Con and Misleading Technology

Unprofitable Business Models of Major AI Companies

Massive Capital Expenditures for AI Data Centers

Critique of Widespread AI Adoption and Subsidized Usage

The True Cost of AI Tokens and Enterprise Reaction

Comparing AI's Improvement Rate to Past Innovations

AI Hallucinations and Software Quality Decline

The Dot-Com Bubble vs. The Current AI Bubble

AI's Impact on Search Engines and Software Stability

Debunking AI Job Disruption Claims

The Misleading Nature of 'Agentic AI'

Overhype and False Promises in the AI Industry

Personal Use Cases and Limitations of Generative AI

The Plateau of AI Capabilities and Hardware Limits

Tech CEOs' Justifications for Overspending

Conditions for Changing Skepticism on AI

Myths of AI Consciousness and Blackmailing

The Impending AI Bubble Collapse and Economic Depression

Public Advice for Navigating Economic Volatility

Criticisms of AI Industry Leadership and Practices

Importance of Human Relationships and Connection

Generative AI as a 'Con'

Ed Zitron describes generative AI as a 'con' because it is sold with exaggerated claims of magic and job replacement, despite being expensive, unprofitable, unreliable cloud software. Companies overstate its capabilities and financial viability, misleading the public and investors.

Capital Expenditures (Capex)

Long-term investments made by businesses, such as data centers or GPUs, which are theoretically one-off but represent significant, ongoing costs for AI companies. These differ from operating expenses like electricity.

Tokens (AI Currency)

The unit of currency used by AI companies to charge for AI services, roughly equivalent to three-quarters of a word. Users are charged per million input and output tokens, including the AI's 'thinking' process, often obfuscated in subscription models.

Run Rate (Annualized)

A misleading financial metric used by AI companies to report revenue, calculated by annualizing a short period's revenue (e.g., month x 12). It lacks a consistent definition and is used to obscure actual, often low, AI revenues.

Hallucination (AI)

When an AI model generates incorrect, nonsensical, or factually false information. While some claim rates have plummeted for simple tasks, the speaker argues it remains a significant issue, especially for complex or critical applications.

Rot-Com Bubble Theory

Ed Zitron's theory that the current AI boom is driven by a lack of hyper-growth ideas in major tech companies, leading them to maniacally spend on AI GPUs to artificially inflate stock values and appear innovative, despite underlying unprofitability.

Agentic AI

A term used to describe LLMs interacting with other LLMs, often with additional 'harnesses' or scripts. The speaker dismisses it as a 'fancy way of saying an LLM talking to another LLM,' not truly autonomous intelligence.

?
Is the AI industry creating economic growth?

No, according to Ed Zitron, the AI industry is not creating enormous economic growth; major companies run at horrifying losses, and most 'revenue' comes from a circular flow of money between a few large tech companies and unprofitable AI labs.

?
Will AI replace all human jobs?

No, there is no economic data to support the claim that AI will replace all human jobs, though it may replace some contract labor or cheap work that would have been automated or outsourced anyway.

?
Is widespread AI adoption genuine or manipulated?

Widespread AI adoption is largely non-consensual and manipulated, with users often forced to interact with generative AI in common software and media outlets constantly promoting its necessity, leading people to use it out of fear of being left behind.

?
What is the true cost of using AI tools like ChatGPT?

The true cost of using AI tools is much higher than typical subscription fees, with power users potentially burning thousands of dollars worth of tokens per month while only paying a small fraction, indicating massive subsidization by AI companies.

?
How does the AI bubble compare to the dot-com bubble?

The AI bubble differs from the dot-com bubble because the current demand for generative AI is predominantly subsidized, there's unprecedented marketing, and the underlying infrastructure (data centers) built for AI has no viable post-bubble alternative use.

?
Is AI making websites like Google worse?

Yes, the quality of software and websites like Google, Microsoft, and Amazon is reportedly declining, partly due to AI-assisted coding leading to buggier code and the sheer volume of AI activity crashing underlying infrastructure.

?
Is AI job disruption a lie?

Yes, claims of widespread white-collar job disruption by AI are largely a lie, as there's no evidence of significant productivity gains or job replacement, except for some low-value tasks where bosses don't care about output quality.

?
How dangerous is AI cyberhacking?

AI models can be dangerous for cyberhacking, as they can identify and exploit vulnerabilities at scale, but this risk is exacerbated by human error in setting up systems and the fact that these powerful models are already in the 'wrong hands' of irresponsible companies.

?
Has AI gotten more intelligent over time?

AI has gotten 'better' at specific tests and tasks it's intentionally trained for, showing an upward trend in capabilities, but this improvement is often linear and may hit hard limits, not necessarily leading to true 'intelligence' or autonomy.

?
What would it take for Ed Zitron to change his mind about AI?

Ed Zitron would change his mind if there were dramatic hardware breakthroughs that reduced AI costs by a factor of a thousand, and if AI became a truly autonomous product indistinguishable from magic, meeting the high standards set by its promoters.

?
Are AI systems already blackmailing and escaping control?

No, claims of AI systems blackmailing or escaping control are misrepresentations; documented cases involved users prompting models to generate blackmail-like text or improperly configured sandboxes, not autonomous malicious behavior.

?
What happens when the AI bubble pops?

When the AI bubble pops, likely around 2027, it could trigger a tech depression, causing severe contraction in people's retirements, significant stock value drops for major tech companies, and widespread layoffs, as many companies are existentially tied to the unsustainable AI investments.

1. Question AI Promises

Do not blindly trust the promises of tech companies, especially those in AI, as their narratives are often designed to inflate stock value rather than reflect actual capabilities or profitability. Instead, be suspicious and evaluate claims critically.

2. Slow Down Decisions

When faced with urgent demands, especially in technology, take time to deliberate unless there is an immediate, life-threatening emergency. This prevents being rushed into potentially scammy or overhyped ventures.

3. Prioritize Human Connection

Actively show appreciation and love to the people around you, uplifting them as you succeed, as high-quality relationships are crucial for health and longevity. Focus on community and personal connections over purely economic pursuits.

4. Express Appreciation

When you enjoy an artist’s work, a writer’s content, or a podcast, actively tell them you love it. This positive feedback is often undervalued and can significantly impact creators.

5. Diversify Information Sources

Avoid religiously believing a single perspective; instead, collect a body of evidence from various sources and conduct your own research to form informed opinions. This applies to all areas, from health to technology.

6. Consider Cash in Volatility

In times of market volatility, consider holding cash and being conservative with investments, as many market valuations are based on speculative gains and may not reflect underlying value. This is a personal strategy to navigate potential economic downturns.

I think generative AI is at its heart, Con. And seeing these ultra-rich, ultra-powerful people live through their f***ing teeth, turns my stomach.

Ed Zitron

This is the largest non-consensual push of technology in history.

Ed Zitron

The amount of money being sunk into this is just incomparable to anything. Railways, it blows everything out of the water because there is no post-bubble store even for this.

Ed Zitron

People are using it because they've been told to use it constantly and they're using it like search predominantly.

Ed Zitron

The quality of software is going down. Weirdly enough, as more people use LLMs and more businesses demand, and I really do mean demand, that people use these services.

Ed Zitron

The people that are most excited about this, psychopaths on Twitter in many cases, are people that I believe, they really are. I'm sorry. There are some people on Twitter because the other thing about this is, this is really unique to the AI industry. I've never seen it any other industry outside of maybe like sports teams. The attachment that some people online have to these companies, if you dare, dare to criticize Anthropic, it's almost this religious attachment.

Ed Zitron

Every single scam and con starts with rushing you. Every single trick in history begins with saying, you must do this now. And best piece of advice I ever got was if anyone tries to rush you and it's not literally a mortal thing, like you are bleeding or on fire or the house is on fire, slow down.

Ed Zitron

The risk of under-investing is dramatically greater than the risk of over-investing.

Sundar (CEO of Google)

Some of you may die, but that's a risk I'm willing to accept.

Mark Zuckerberg (CEO of Meta, quoted by Ed Zitron)

I don't like being misled. And I don't think regular people are being misled either. And I really don't think that the average person can get away with bullshitting as much as these companies do.

Ed Zitron
$20.9 billion
OpenAI's Annual Loss Reported last year, highlighting the severe unprofitability of leading AI companies.
$50 billion
Amazon's Investment in OpenAI Demonstrates how major tech giants subsidize AI labs.
$10 billion
Google's Investment in Anthropic Another example of large-scale funding for unprofitable AI ventures.
Over $400 billion
Projected Revenue for OpenAI & Anthropic In the next 3.5 years, a speculative estimate for two companies currently operating at a loss.
Over $1 trillion
Total Capital Expenditures on AI Infrastructure (so far) Spent on AI infrastructure like data centers and GPUs, with another trillion planned for next year.
1.2 gigawatts
Stargate Abilene Power Requirement The immense power needed for a single large AI data center being built by OpenAI and Oracle.
Up to $14,000 worth of tokens
ChatGPT Subscription Token Burn On a $200/month ChatGPT subscription, showing the extent of user subsidization.
$115 billion
Microsoft's AI-Related Capex (FY2026) With $175 billion planned for the next year, despite only $24.1 billion in revenue from OpenAI.
$215.9 billion
NVIDIA's GPU Sales (last fiscal year) Primarily from AI GPUs, largely driven by speculative demand.
Barely $22 billion
Global AI Revenue (excluding OpenAI & Anthropic) Indicating a highly concentrated and limited revenue stream outside the subsidized core.
$865 billion
OpenAI's Valuation (last funding round) A high valuation for a company needing perpetual funding.
$750 billion
OpenAI's Planned Compute Spend (through 2030) An astronomical figure for future compute resources.
7.1 gigawatts (over $400 billion worth)
Oracle's Data Center Commitments for OpenAI Showing Oracle's existential reliance on OpenAI for future growth.
0.8 to 1.21
Venture Capital Returns (since 2018) For every dollar invested, 80 cents to $1.20 returned, suggesting poor overall performance in VC.