MurmurCast
OpinionDiscussion

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

The Diary Of A CEO with Steven Bartlett2h 27m

Ed Zitron argues that generative AI is fundamentally a con—an expensive, unprofitable technology oversold by ultra-wealthy executives to mask the tech industry's lack of genuine innovation. He predicts a major economic collapse around 2027 when funding dries up, triggering a broader tech depression with severe consequences for retail investors and the broader economy.

Summary

Ed Zitron presents a comprehensive critique of the generative AI industry, arguing it represents a massive financial con rather than genuine technological progress. He contends that the entire AI boom is driven by circular financing: Microsoft, Google, and Amazon feed money to unprofitable AI companies like OpenAI and Anthropic, which then spend lavishly on GPUs from Nvidia, creating the illusion of enormous economic value. In reality, he claims, these AI companies lose staggering amounts of money—OpenAI lost $20.9 billion last year—and cannot survive without continuous external funding.

Zitron distinguishes between different types of AI, emphasizing that the criticism applies specifically to generative AI and large language models, not AI broadly or robotics. He argues that while hallucinations have decreased on benchmark tests, these benchmarks are specifically designed for the models and don't reflect real-world reliability. He points out that when enterprises were forced to pay actual token costs (rather than subsidized subscriptions) in March 2026, they immediately cut back dramatically, with companies like Uber burning through entire annual token budgets in three months. This reveals that organizations don't genuinely value the technology at its actual cost.

On the economic structure, Zitron explains that the AI industry has created a "rock-con bubble"—tech giants have run out of genuine growth ideas in their core businesses, so they're desperately spending $1 trillion-plus annually on AI CapEx to appear innovative and sustain stock valuations. Microsoft alone spent $115 billion on capital expenditures in one year while only making $34 billion total revenue, with just $24 billion from OpenAI. The math doesn't work: these companies are building massive data centers that can only be filled by the unprofitable AI labs they're funding.

Zitron predicts that OpenAI will run out of cash around 2027 because it needs to raise at least $100 billion annually just to survive, largely to continue training runs that cost tens of billions each. When OpenAI tried to go public at a $1 trillion valuation, advisors said it was too aggressive. If OpenAI can't go public, SoftBank (which holds about $100 billion in paper value of OpenAI stock) faces severe problems, and the entire investment thesis collapses. This triggers a cascade: Amazon, Google, and Microsoft will have to restate guidance downward, venture capital—which has been returning less than $1 per dollar invested since 2018—will face a reckoning, and the stock market will experience a significant contraction.

The broader economic consequences would be severe: the S&P 500 is heavily dependent on these mega-cap tech stocks, with Nvidia alone representing 7-8% of the index. A 50-70% decline in Nvidia's revenue would ripple through the entire market. Retail investors' retirements could contract 20-40%. This leads to job losses, hiring freezes, and economic contraction. Zitron argues this isn't like the dot-com bubble because these are massive companies with enormous leverage making gigantic promises; the fallout will be more severe.

Zitron also critiques the social and environmental harms: data centers are being built recklessly in communities using gas turbines that damage air quality, require immense water resources, and increase local power consumption dramatically. He argues this represents inequality—while regular people struggle to get business loans or mortgages, AI companies get billions in funding and GPU contracts. He's particularly critical of how executives deliberately spread fear about AI being dangerous and powerful while actually knowing it's unreliable, using this fear-mongering to justify massive spending and secure funding.

When challenged on whether he could be wrong, Zitron says it would require hardware breakthroughs reducing costs by 1,000x (not happening despite all efforts), or the models becoming truly autonomous and reliable at scale (which he doesn't believe is possible given the mathematical constraints of large language models). He respects some critics like Dario Amodei but thinks they're ultimately part of the con, using safety concerns to justify massive spending while later walking back doomsday narratives.

The host pushes back with the disruptive innovation theory—that early technologies often look worse than what they replace and seem uneconomical until they don't. Zitron acknowledges this but distinguishes that we're seeing no path forward like there was with cars, the internet, or AWS. He notes that unlike those technologies, there's no clear roadmap to making generative AI profitable, efficient, or genuinely useful at scale. The technology has hit diminishing returns on improvement despite massive spending.

Zitron emphasizes the media and analyst complicity, noting that tech companies don't disclose actual AI revenues (using vague metrics like "annualized run rate") and that skepticism has become rare. He calls for people to be suspicious of promises, to recognize that markets are being manipulated through selective information, and to focus on what companies are actually achieving today rather than speculative futures.

About this episode

Tech critic Ed Zitron exposes the AI bubble, why OpenAI and Anthropic are burning billions, the fake AI boom, and why the crash could wipe out the ENTIRE economy! Ed Zitron is a British AI critic and one of the most cited voices warning that the AI industry is one giant bubble. He hosts the 'Better Offline' podcast, reaching over a million monthly downloads, and writes the newsletter 'Where's Your Ed At'. He is the founder and CEO of the PR firm EZPR, and is currently writing his upcoming book, 'Why Everything Stopped Working'.  He explains:  ■ Why he believes generative AI is a “con”  ■ The real reason OpenAI and Anthropic can't turn a profit  ■ Why data centers could leave a $500 billion debt bomb  ■ Why superintelligence is a myth sold by tech billionaires  ■ Why AI won't take your job, no matter what CEOs promise Chapters 00:00:00 Intro 00:02:15 AI Is A Con 00:05:55 How Much Power Data Centres Really Need 00:07:42 Is Widespread AI Adoption Manipulation Or Do People Actually Like Using It? 00:11:40 The Actual Cost Of AI And How Tokens Actually Work 00:15:49 Is The Spending Of AI Companies Justifiable? 00:19:47 Will The Rate Of Improvement Of AI Go Up, Like Previous Innovations? 00:24:03 How Bad Are AI Mistakes? 00:26:34 Comparing Human Error To AI Hallucinations 00:31:11 If The Output Is The Same, Does It Matter If Humans Or AI Created It? 00:33:58 Can We Trust AI Like We Trust Humans? 00:38:17 Would People Use AI If They Paid The Honest Cost? 00:42:15 How Does The AI Bubble Compare To The Dot-Com Bubble? 00:47:22 Does AI Demand Match The Cost And Risk Of Data Centres? 00:52:26 Is AI Making Websites Like Google Worse? 00:58:26 Ads 01:00:30 Is AI Job Disruption A Lie? 01:10:02 Could Your Narrative Be Helping AI Companies? 01:14:02 How Dangerous Is AI Cyberhacking 01:17:10 Is The AI Industry Creating Economic Growth? 01:18:53 How Would The US Beat China In The AI Race? 01:19:33 Is Robotics A Threat To Jobs? 01:23:03 What Do You Think About Agentic AI? 01:24:43 Is The Adoption Of AI The Same As The Rise Of The Internet? 01:27:46 The Overhype Of AI 01:30:03 What Do You Use Generative AI For? 01:33:21 Has AI Gotten More Intelligent? 01:34:09 Will AI Start To Do More Jobs As It Gets More Capable? 01:36:07 What Does The Future Look Like As AI Grows? 01:38:08 You Don't Think People's Workflows Have Been Transformed By AI? 01:40:21 Will All AI Be Powered By Data Centres? 01:43:20 Ads 01:44:34 Is Overspending On AI Due To Demand Or Something Else? 01:54:44 Tech CEOs Rebuttal 01:56:54 What Would It Take For You To Change Your Mind About AI? 02:00:12 Are AI Systems Already Blackmailing? 02:07:51 Are We In An AI Bubble And What Happens When It Pops? 02:12:48 The Tech Depression Is Coming 02:18:30 What Should The Public Do? 02:21:07 Why Do You Have A Bone To Pick With AI CEOs? 02:24:29 What Should We Be Doing To Improve Our Relationships And Social Connection? Follow Ed Zitron: Linktree: https://link.thediaryofaceo.com/C6fKrVK Better Offline: https://link.thediaryofaceo.com/A9awRDM  X: https://link.thediaryofaceo.com/GTr0z7R  Where's Your Ed At Newsletter: https://link.thediaryofaceo.com/CZ3JLap  You can get $10 off your first year of Where's Your Ed At Premium, here: https://link.thediaryofaceo.com/91LBdmi   The Diary Of A CEO: ◼ Join DOAC circle here - https://doaccircle.com/ ◼ Buy The Diary Of A CEO book here - https://link.thediaryofaceo.com/BWjLTZK ◼ Shop The Diary Of A CEO collection: https://thediary.com/collections/shop ◼ Get email updates - https://link.thediaryofaceo.com/5IB1H6E ◼ Follow Steven - https://link.thediaryofaceo.com/AGU9QP4 Sponsors: Fiverr - https://fiverr.com/diary and get 10% off your first order when you use code DIARY Saily - Download from the app store and use code DOAC at checkout for 15% off. For more details: https://saily.com/doac ⛵

Key Insights

  • OpenAI lost $20.9 billion in a single year, with nearly all revenues from AI coming from just two unprofitable companies funded by the same three giants that are building data centers
  • When enterprises were forced to pay actual token costs instead of subsidized subscriptions in March 2026, they immediately cut AI spending dramatically, with Uber exhausting annual budgets in three months
  • Microsoft spent $115 billion on capital expenditures in one year but only generated $34 billion in total revenue, with just $24 billion from OpenAI, revealing fundamentally broken unit economics
  • The AI industry has created a circular funding loop where Amazon, Google, and Microsoft feed money to OpenAI and Anthropic, which spend it on Nvidia GPUs, then those companies point to GPU sales as proof of demand
  • Hallucination rates on benchmark tests have improved dramatically (from 21.8% to 0.7%), but these are simple, specifically-designed-for-the-model tests that don't reflect real-world reliability
  • Tech executives deliberately use fear-mongering about powerful, dangerous AI to justify massive spending while knowing the technology is actually unreliable and limited
  • Hardware companies like Nvidia and Broadcom, despite being the most resourced semiconductor firms, haven't achieved the cost-reduction breakthroughs needed to make AI profitable
  • OpenAI needs to raise at least $100 billion annually just to survive, making it impossible for the company to ever become profitable at current spending and revenue levels
  • If OpenAI can't go public, SoftBank's $100 billion paper valuation in the company becomes worthless, triggering cascading failures across venture capital and investment funds
  • Venture capital has returned less than $1 per dollar invested since 2018, meaning the entire VC ecosystem is actually destroying wealth while celebrating paper gains
  • The data centers built for AI generate immense environmental damage through gas turbines in low-income communities, yet this isn't being regulated or stopped despite known harms
  • Companies like Meta could only justify AI spending by showing 15 basis points (0.15%) of improved retention after spending billions, yet continue claiming AI is transformative
  • The tech industry deliberately avoids disclosing actual AI revenues, instead using undefined metrics like 'annualized run rate' that can mean month times 12, 13, or last four weeks times 13
  • Unlike disruptive innovations like AWS, there is no technical roadmap showing how generative AI will become cheaper, more autonomous, or more reliable despite trillion-dollar spending
  • Retail investors' retirements are heavily concentrated in mega-cap tech stocks where Nvidia alone represents 7-8% of the S&P 500, creating severe downside risk if the AI bubble collapses

Topics

Generative AI as a financial conCircular financing between tech giants and AI labsProfitability crisis in AI companiesToken economics and real vs. subsidized costsPredictions for 2027 economic collapseSoftBank's financial exposureStock market valuation dependency on techVenture capital returns crisisEnvironmental and community damage from data centersHallucination benchmarks and real-world reliabilityComparison to dot-com bubbleHardware limitations and scaling challengesMedia and analyst complicityExecutive fear-mongering tacticsInequality in access to capital

Transcript

If you're running a business, people have probably told you to use AI or get left behind. And I get that urgency. Still, AI doesn't fix a messy business. It exposes one. Because AI can only work with the information it can see. So if you ask AI something about your business, about stock or sales or customers or about your team, that information needs to be connected to the AI to be useful. That's the idea behind NetSuite by Oracle, the sponsor of this episode. NetSuite is the, the sponsor of this episode. NetSuite is the AI-powered business management suite that securely connects your financials, your inventory, your commerce, your HR, and your CRM into a single source of…

Full transcript available for MurmurCast members

Sign Up to Access

More from The Diary Of A CEO with Steven Bartlett

Get AI summaries like this delivered to your inbox daily

Get AI summaries delivered to your inbox

MurmurCast summarizes your YouTube channels, podcasts, and newsletters into one daily email digest.