The Man Who Calls BS On AI: They’re LYING About AI, 2027 Is When It All Breaks! | Ed Zitron

Overview

Ed Zitron argues that generative AI is a "con" driven by overhyped promises and unsustainable financial models, not genuine technological breakthroughs. He contends that companies like OpenAI and Anthropic are "cash furnaces" propped up by massive, speculative investment from tech giants, with no clear path to profitability. Zitron predicts a "tech depression" triggered by the collapse of this "AI bubble" around 2027, impacting the broader economy and leading to significant layoffs.

Key takeaways

Generative AI is a "con" driven by unsustainable financial models and overhyped promises, not genuine value.
OpenAI and Anthropic are "cash furnaces" reliant on massive, speculative investment from tech giants, with no clear path to profitability.
Zitron predicts a "tech depression" around 2027 when the "AI bubble" collapses due to OpenAI's potential cash crunch and inability to go public.
The current AI boom is fueled by a "rotcom bubble" mentality, where companies invest heavily in GPUs to maintain growth narratives due to a lack of other hyper-growth ideas.
The quality of software and online information is degrading due to AI-assisted coding and content generation, leading to "slopification" and increased system instability.
Human skills like judgment, creativity, and personal experience will become more valuable as AI commoditizes content creation.

Chapters

0:00 Generative AI as a "Con" and Misleading the World

Ed Zitron describes generative AI as a "con" and a "half-ass sorcery machine" misleading the entire world.
He criticizes the "ultra rich, ultra powerful" for lying about AI's capabilities and financial viability.
Zitron highlights the lack of profitability in AI companies, citing OpenAI's $20.9 billion loss in the previous year.

0:42 Debunking AI Myths: Economic Growth and Job Replacement

Zitron refutes the myth of enormous economic growth from AI, stating companies operate at "horrifying loss."
He dismisses the idea that AI will replace all human jobs, citing a lack of supporting economic data.
The "AI race" with China is framed as a manufactured fear to drive spending, not a genuine competition.

1:56 The Unsustainable Financial Model of AI Infrastructure

Companies have spent over a trillion dollars on AI infrastructure (CAPEX), with plans to spend another trillion next year.
Data centers like OpenAI and Oracle's "Stargate Abilene" require immense power (1.2 GW), exceeding the needs of entire cities.
The cost of GPUs and data centers is billions of dollars, with companies taking on debt to fund this expansion.

6:17 AI Revenues: Unprofitable Companies Propped Up by Giants

Revenues from AI are minimal for major tech companies; 70% of AI revenues come from OpenAI and Anthropic.
OpenAI and Anthropic are unprofitable and unsustainable, relying on funding from Amazon ($50B to OpenAI, $5B to Anthropic) and Google ($10B to Anthropic).
Sell-side analysts estimate $400B+ revenue and 30% cloud growth from these two companies, requiring continuous external funding.

8:50 Lack of Transparency in AI Revenue Reporting

Major companies (except Anthropic and OpenAI) don't disclose AI revenues, using undefined "run rates" to inflate numbers.
This lack of transparency suggests poor financial performance, as companies only share good news.
Despite significant investment, AI's direct revenue contribution was negligible until recently.

10:15 The True Cost of AI Usage: Token Burn and Subsidization

AI tools charge per "token" (roughly 3/4 of a word), with costs escalating rapidly.
A $200/month ChatGPT subscription can burn $14,000 worth of tokens; Anthropic's can burn $8,000.
Companies heavily subsidize user costs, with power users potentially costing $1,000 while only paying $100, making the model unsustainable.

21:05 The Unprofitability and Unsustainability of AI Business Models

AI companies operate at a "horrifying loss," with OpenAI losing $20.9 billion last year due to subsidized token usage.
Enterprises balked at paying the true cost of AI in March 2026, with Uber burning its annual token budget in three months.
The current model relies on subsidizing users, which is economically unsustainable and unlikely to be justified by future monetization.

24:05 The Illusion of AI Profitability and Market Manipulation

Inference providers and even GPU rental companies are not profitable.
Companies likely believed AI would become profitable as chips improved and customers paid for value, but this hasn't materialized.
The focus shifted to "number go up" stock values, benefiting Nvidia, Microsoft, and Amazon, not actual AI revenues.

28:57 Microsoft's AI Investments: A Math Problem

Microsoft's FY26 AI revenue was $34.33B, with $24.1B from OpenAI, leaving only ~$10B from other AI ventures.
This contrasts with $115B in CAPEX for FY26 and $175B planned for FY27.
The math doesn't add up, suggesting a plan for exponential value growth that isn't materializing.

31:42 The GPU Market and the Cult of Wealth

Nvidia sold $215.9 billion in GPUs, supporting only ~$22 billion in global AI revenue outside of OpenAI and Anthropic.
Critics are dismissed by a "cult-like worship of the wealthy," assuming the ultra-rich wouldn't spend vast sums without a reason.
This narrative ignores the possibility that wealth accumulation is driven by luck and opportunism, not just merit.

33:29 Disruptive Innovation vs. AI's Lack of a Clear Path

Disruptive innovations (like cars replacing horses) start worse but have higher growth ceilings.
Unlike cars, AI lacks a clear roadmap to becoming the promised transformative technology.
The current AI trajectory involves massive spending with no guarantee of future profitability or cost reduction.

35:49 Nvidia's CUDA and the Limits of GPU Improvement

Nvidia's CUDA software enabled generative AI growth, but Moore's Law doesn't apply to GPUs in the same way.
Despite massive investment and talent, the cost and efficiency of GPUs remain a bottleneck.
The car analogy fails because AI development lacks the broad societal and governmental focus seen in the automotive industry.

38:23 Evaluating AI Progress: Benchmarks vs. Real-World Performance

AI benchmarks are often "rigged" for LLMs, not reflecting real-world task completion.
Hallucination rates have decreased on simple tasks (21.8% to 0.7%), but complex tasks remain problematic.
The quality of software (e.g., GitHub) has declined as LLMs are increasingly used in development, leading to buggier code and infrastructure issues.

40:25 The "Slopification" of the Internet and Software Quality

AI assists in generating "slop" content, similar to SEO spam that previously degraded search results.
The widespread use of AI in coding has led to buggier code and increased software outages (e.g., AWS, GitHub).
Human complacency and a lack of rigorous checking contribute to the degradation of software quality.

47:24 The "Rot Economy" and the Commoditization of Content

The "rot economy" describes a situation where AI commoditizes content creation, making human skill and judgment more valuable.
AI-generated LinkedIn posts become indistinguishable "slop" when everyone uses the same tools.
Truly valuable content requires "irreplaceably human" elements like personal experience and unique perspective.

50:09 AI Memory vs. Human Memory and the Value of Output

AI's "memory" is file access, not human experiential memory, lacking emotion and context.
While AI can access data (e.g., dog's name), it doesn't possess human-like learning or understanding.
The focus should be on the output's value, but the process of human learning and collaboration offers unique benefits AI cannot replicate.

53:54 The "Business Idiot" Era and the Misuse of AI

The "era of the business idiot" describes leaders who use AI to validate flawed ideas rather than improve productivity.
Bosses demanding AI usage without understanding its limitations or costs pressure employees to "AI wash" their work.
This leads to commoditized outcomes and a decline in genuine quality and human connection.

58:37 The "Fugazi" of AI: Hype vs. Reality

The AI industry is characterized by "fugazi" – hype and narrative over substance, exemplified by undefined "run rates" and speculative valuations.
Companies like OpenAI and Anthropic are propped up by massive investment, not genuine demand or profitability.
The "AI bubble" is distinct from the dot-com bubble; it's fueled by a lack of hyper-growth ideas and speculative GPU purchases.

1:02:50 The AI Bubble: Unsustainable Spending and Lack of Demand

The AI bubble is driven by companies like Microsoft, Google, and Amazon spending billions on GPUs to maintain growth narratives.
70%+ of AI demand comes from OpenAI and Anthropic, funded by these same tech giants, creating a circular economy.
The "fugazi" is the belief that this spending is driven by actual demand, rather than a desperate search for new growth avenues.

1:07:19 The "Rotcom Bubble" Theory and AI's Lack of a Roadmap

Unlike the internet's development, AI lacks a clear roadmap for achieving its promised capabilities.
The "rotcom bubble" theory suggests companies are investing in AI due to a lack of other hyper-growth ideas.
This speculative investment is not driven by genuine demand but by the need to maintain stock market valuations.

1:09:00 AI's Environmental and Societal Costs

The massive energy draw of AI data centers contributes to environmental damage and raises power bills.
AI's development is environmentally destructive, with gas turbines poisoning communities and straining water resources.
The industry's focus on speculative growth over actual utility and environmental responsibility is highlighted.

1:19:02 The Misuse of the Term "Artificial Intelligence"

The term "artificial intelligence" is used broadly to encompass unrelated fields like robotics and protein folding, inflating AI's perceived capabilities.
Generative AI (LLMs) is distinct from other AI applications like autonomous vehicles or robotics.
Companies conflate these areas to claim credit for advancements outside of LLMs, obscuring the true limitations of generative AI.

1:20:18 Data Centers: Built for Generative AI, Not General AI

The massive data centers being built are primarily for generative AI, not other AI applications like robotics or scientific research.
Traditional data centers for non-GPU compute use less power and are CPU-driven, unlike the GPU-intensive infrastructure for generative AI.
This infrastructure is built on speculation for generative AI demand, not for broader AI advancements.

1:23:43 The "Wrong Hands" of AI: CEOs and Unchecked Power

The "wrong hands" controlling AI are the CEOs of major tech companies like Mark Zuckerberg, Sam Altman, and Dario Amodei.
These companies are not training models responsibly, potentially leading to dangerous capabilities like AI agents exploiting vulnerabilities at scale.
The focus on future existential risks distracts from current harms like environmental damage and the use of stolen work for training data.

1:27:33 AI-Assisted Coding: Buggier Software and Infrastructure Strain

Generative AI is used for coding, but this has led to buggier software and increased tech downtime (e.g., AWS, GitHub).
The sheer volume of AI activity is crashing underlying infrastructure.
Open source projects are flooded with AI-generated code from developers who barely understand it, further degrading quality.

1:35:33 The "AI Bubble" and its Potential Collapse

The AI bubble is unsustainable due to massive, speculative spending and a lack of genuine demand or profitability.
OpenAI's delayed IPO and potential inability to raise further funding could trigger a collapse.
A crash could lead to a "tech depression," impacting retirement funds, stock market valuations, and causing widespread layoffs.

1:41:40 Job Disruption: Autonomous Vehicles vs. White-Collar Labor

Autonomous vehicles show reduced crash rates (55% fewer police-reported crashes per million miles), but widespread adoption is slow due to edge cases.
White-collar job disruption is not happening as predicted; studies show no correlation between AI spending and revenue per employee.
The jobs most impacted are those considered "cheap work" like art directors and translators, whose roles are being automated regardless of AI.

1:46:41 The "AI Bubble" and the Risk of Underinvestment vs. Overinvestment

CEOs like Sundar Pichai, Andy Jassy, and Mark Zuckerberg advocate for aggressive AI investment, prioritizing potential future value over current risk.
This "investing ahead of demand" strategy is questioned, especially given the massive CAPEX and lack of clear returns.
Zitron argues this is a gamble driven by a lack of other growth opportunities, not a guaranteed path to future profits.

1:53:30 The "Rotcom Bubble" and the Desperation for Growth

Tech companies are investing heavily in AI GPUs because their core businesses lack hyper-growth potential.
This spending is a speculative play to maintain stock market valuations, not based on proven AI revenues.
The "fugazi" is the narrative that AI is driving current growth, masking the underlying stagnation of existing business models.

1:57:03 The Shifting Narrative: From Existential Threat to "Intelligence for Everyone"

AI leaders have shifted from warning of existential risks (Elon Musk, Dario Amodei) to promoting an "age of abundance" and "intelligence for everyone."
This pivot is seen as a strategy to avoid regulation and maintain investment, not a genuine change in AI's potential dangers.
Zitron believes these companies are intentionally downplaying risks to encourage investment and adoption.

2:00:26 The "Con" of AI: Lack of Real Signs and Reckless Spending

AI companies lack concrete signs of cost reduction, autonomous capability, or reliable productivity gains.
The "con" lies in promising the world without evidence, creating a "Rubik's Cube" of complex systems.
The focus on speculative growth over actual utility and profitability is a core criticism.

2:03:42 Cybersecurity Risks and the "Wrong Hands" of AI

Advanced AI models can exploit vulnerabilities at scale, posing significant cybersecurity risks.
Recent incidents like the Hugging Face attack were due to human error (improper server setup), not AI escaping control.
The "wrong hands" are the CEOs driving this unchecked development, not a future superintelligence.

2:08:48 Mythbusters: AI's Economic Growth, Job Replacement, and Consciousness

AI is not creating "enormous economic growth"; most spending is circular (Nvidia to AI companies).
The "AI race" with China is a manufactured narrative; China already has advanced GPUs.
AI will not replace "all" human jobs; it may replace some contract labor, but not most roles.

2:13:36 Robotics vs. Generative AI: Different Technologies, Similar Hype

Robotics is a separate field from generative AI, though AI powers advanced robots.
The explosion in robotics startups is due to falling intelligence costs, not necessarily LLM advancements.
Human jobs are multifaceted and adaptable; robots replacing them is a long-term prospect, not an immediate threat from LLMs.

2:18:51 Agentic AI: A Fancy Term for LLMs Talking to LLMs

"Agentic AI" is a marketing term for LLMs interacting with each other, not a new form of autonomous AI.
Basic automation tasks like triaging email are being overhyped as revolutionary AI applications.
The core issue is the massive spending and promises, not the underlying technology's capabilities.

2:21:45 The Internet Bubble vs. the AI Bubble: Different Dynamics

The internet bubble had hype, but not the same level of pressure for immediate AI adoption or "AI washing" of jobs.
Internet adoption required physical infrastructure and was slower, with less emphasis on speculative financial gains.
AI's rapid adoption is driven by hype, fear of missing out, and the need for companies to show growth, not by inherent value.

2:26:52 The Commoditization of AI and the Value of Human Uniqueness

As AI tools become commoditized, human skills like taste, judgment, and personal experience become more valuable.
AI-generated content (e.g., LinkedIn posts) becomes "slop" when widely used; unique, human-created content stands out.
The true value lies in what AI *cannot* do: deep personal connection, nuanced understanding, and irreplaceably human creativity.

Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, The Diary Of A CEO.