Key Moments
The Man Who Calls BS On AI: They’re LYING About AI, 2027 Is When It All Breaks! | Ed Zitron
The Diary Of A CEOWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
Major AI companies are burning billions on generative AI, losing up to $20.9 billion last year, as the 'AI boom' is driven by subsidized usage and disingenuous marketing, hinting at a 2027 collapse that could trigger a tech depression.
Key Insights
OpenAI lost $20.9 billion last year, and 70% of all AI revenues across major tech companies (Anthropic, Amazon, NVIDIA, Microsoft, OpenAI, Google) come from OpenAI and Anthropic, both of which are unprofitable and unsustainable without massive external funding (e.g., Amazon sent $50 billion to OpenAI, Google $10 billion to Anthropic).
Enterprises struggled when forced to pay the true cost of AI; Uber burned its entire annual token budget in three months, and a $200/month ChatGPT subscription can cost up to $14,000 in actual token usage, revealing the extent of current subsidies.
The quality of software across major tech companies like Google, Microsoft, and Amazon is reportedly worsening, with increased downtime and bugs (e.g., Amazon Web Services outages, GitHub instability), which some industry data links directly to the explosion of AI-assisted coding.
While AI is improving on simple tasks (hallucination rates on summarization dropped from 21.8% to 0.7% over four years), this doesn't translate to complex, reliable, and novel outputs, making its value in sophisticated applications like legal work or creative industries questionable.
Major tech CEOs like Google's Sundar Pichai and Amazon's Andy Jassy assert that the risk of underinvesting in AI is 'dramatically greater than the risk of overinvesting,' pushing capital expenditures despite current unprofitability and lack of clear returns, driven by a fear of missing out on perceived future growth.
OpenAI plans to spend $750 billion on compute through 2030, a figure Ed Zitron believes is unsustainable and likely to lead to its collapse by 2027, potentially triggering a tech depression due to the interconnectedness of major tech companies and the stock market's reliance on their growth.
The generative AI industry operates at a massive, undisclosed loss
Despite the hype, major generative AI companies like OpenAI and Anthropic are bleeding money at an alarming rate. OpenAI alone reported a $20.9 billion loss last year. The claim that AI is driving economic growth is unsubstantiated by data, as these companies consistently operate at horrifying losses. A significant portion, specifically 70%, of all reported AI revenues across tech giants like Amazon, NVIDIA, Microsoft, OpenAI, and Google, originates from these two unprofitable entities. These companies are sustained by massive, ongoing investments from larger tech firms, for instance, Amazon injecting $50 billion into OpenAI and Google providing $10 billion to Anthropic. This financial structure suggests that their very existence is contingent on continuous external funding, rather than inherent profitability or market demand, raising serious questions about the long-term viability of their business models. The reliance on these subsidies indicates a fundamental flaw in the current generative AI economic landscape. Ed Zitron highlights that even their financial reporting is opaque, often using undefined "run rates" instead of clear revenue figures, further obscuring their true financial health.
The true cost of AI is heavily subsidized, masking its unviability
The perceived widespread adoption of generative AI tools is largely driven by heavily subsidized pricing, meaning most users are not paying the actual cost of operation. For example, a $200/month ChatGPT subscription can accrue up to $14,000 in token costs, while Anthropic's equivalent can reach $8,000. This disparity reveals that the underlying tech companies are absorbing enormous costs to make their services appear affordable and accessible. When enterprises like Uber were suddenly confronted with the true, unsubsidized costs in 2026, many experienced immediate budget crises, with Uber exhausting its annual token budget in just three months. This resistance to actual pricing suggests that while AI tools are adopted when cheap, their perceived value diminishes sharply when users are expected to bear the full economic burden. The industry's reliance on these hidden subsidies is unsustainable and creates a false impression of market demand and utility, artificially inflating adoption rates. This 'non-consensual push' of technology forces users to engage with AI, leading to inflated usage statistics that do not reflect genuine, value-driven demand.
AI adoption contributes to declining software quality and system instability
Contrary to claims of increased productivity, the widespread integration of AI, particularly AI-assisted coding, is correlating with a demonstrable decline in software quality and an increase in system outages across the tech industry. GitHub, for example, is now flooded with AI-generated code, often written by individuals who lack a deep understanding of what they are shipping. This 'slop' code makes human developers complacent, leading to less rigorous checks and potentially introducing more bugs and security vulnerabilities. Major platforms like Google Docs and Sheets are experiencing increased instability and bugs, while Amazon Web Services has faced multiple downtimes attributed to issues stemming from AI coding tools. This suggests that the promise of AI enhancing software development is currently backfiring, creating a more brittle and less reliable technological infrastructure. The demand from businesses to use AI tools, coupled with a lack of critical oversight from developers, is leading to a degradation of the very products they are meant to improve.
AI's 'intelligence' is limited to specific, often simple, tasks
While AI models have shown improvements in specific benchmarks, such as a drop in hallucination rates on simple summarization tasks from 21.8% to 0.7% over four years, these improvements are often confined to areas where the models are specifically trained for tests. This does not translate to genuine, human-like intelligence, context, or novel output in complex scenarios. Ed Zitron argues that generative AI excels at producing 'median' answers or 'slop,' and struggles with tasks requiring deep understanding, empathy, or nuanced contextual awareness. For example, while AI can be useful for troubleshooting technical issues by processing large logs, it cannot replicate the complex learning, emotional intelligence, or contextual understanding of a human intern or editor. The industry's claims of AI approaching general intelligence or replacing skilled white-collar jobs are largely exaggerated, as current capabilities are still far from autonomous, reliable performance in dynamic, real-world environments.
The 'AI race' is a fabricated urgency to drive speculative investment
The narrative of an 'AI race,' particularly against countries like China, is a disingenuous tactic used to justify immense capital expenditure and maintain speculative interest. Ed Zitron argues there is no tangible 'race' beyond the scramble to create large, expensive language models. This manufactured urgency compels nations and companies to pour trillions into AI development, even when the underlying technology is unprofitable and its applications are unclear. The 'what if these models fall into the wrong hands?' argument is dismissed, as Zitron contends they are already in the wrong hands: those of hyper-wealthy tech leaders who prioritize growth and market dominance over societal well-being. This fear-based marketing aims to secure continuous investment and prevent regulation, despite the environmental and economic costs. The massive investment in GPUs and data centers is primarily fueled by this speculative competition, not proven market demand or tangible returns.
A 2027 collapse of the AI bubble is imminent, threatening a wider tech depression
Ed Zitron predicts a significant collapse of the AI bubble around 2027, driven by the unsustainable financial models of key players like OpenAI. OpenAI, having delayed its public offering from 2026 to 2027, needs to raise at least $100 billion annually just to survive, a feat becoming increasingly difficult given its valuation and lack of profitability. This impending financial reckoning for OpenAI, potentially exacerbated by Anthropic going public first with better economics, could trigger a domino effect across the tech industry. Major tech companies like SoftBank, Amazon, Google, and Microsoft have significant investments and future growth projections tied to OpenAI and Anthropic. A failure to go public or a significant devaluation of OpenAI's stock could lead to massive write-downs for these investors, causing their stock values to plummet. This would not only impact institutional investors but also retail investors, whose retirements are heavily tied to these tech giants. The current overspending, driven by a lack of new 'hyper-growth' ideas and a desire to prop up stock prices, is creating a fragile system where the collapse of a few key AI entities could trigger a broader tech depression, characterized by job losses, economic contraction, and a re-evaluation of inflated tech valuations.
The ethical and social costs of the AI boom are being ignored
The current AI boom is not only financially unsustainable but also carries significant ethical and social costs that are being largely ignored by its proponents. The immense energy consumption of AI data centers, for instance, leads to environmental degradation and increased power bills. The construction of these data centers, often powered by polluting gas turbines, disproportionately affects marginalized communities. Furthermore, the practice of training AI models by 'stealing' intellectual property and creative works from millions of people raises serious ethical questions about compensation and ownership. The constant fear-mongering about AI replacing all jobs, while largely untrue, creates anxiety and pressure on workers to adopt potentially flawed tools. Ed Zitron criticizes the 'cult-like worship' of AI companies and their leaders, who he believes are detached from real-world problems and primarily motivated by financial gain and market control, rather than genuine societal benefit. This disregard for the broader consequences highlights a deeper problem within the tech industry's pursuit of unchecked growth.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Concepts
●People Referenced
Common Questions
Ed Zitron views generative AI as a con because it is marketed as magical and transformative, capable of solving all problems, when in reality it's expensive, unprofitable cloud software that is often unreliable and misleading. Companies overstate its capabilities and financial viability to exploit weaknesses in journalism and the economy.
Topics
Mentioned in this video
An AI company criticized for operating at a horrifying loss, notably $20.9 billion last year, and for its unsustainable business model heavily subsidized by other tech giants.
An AI company criticized for its unsustainable, unprofitable model, heavily subsidized by Google, and for engaging in scare tactics to promote its technology.
A tech giant criticized for subsidizing unprofitable AI companies like OpenAI and Anthropic, contributing to the unsustainable financial model of the AI industry.
A chip manufacturer whose GPUs are essential for AI, but its sales figures are presented as evidence of speculative investment rather than actual economic growth from AI.
A major tech company that heavily invests in AI, particularly OpenAI, with questions raised about the true profitability and impact on its overall revenue and software quality.
A tech giant criticized for its AI investments, particularly in Anthropic, and for allegedly worsening its search product by prioritizing queries over quality to push AI integration.
A software company building massive data centers, whose future is presented as existentially tied to the spending of OpenAI.
An example of a company that burned through its annual AI token budget quickly, highlighting the high cost of AI for enterprises.
A code hosting platform criticized for declining quality and instability due to an influx of AI-generated code, leading to more frequent downtime.
A company whose software (Facebook, Instagram) is described as worse due to aggressive AI integration and who Mark Zuckerberg is accused of pissing money away on AI.
An LLM company that raised money at a high valuation ($26 billion), raising questions about its profitability and acquisition potential.
A premium service for vetting talent, recommended for businesses to hire specialists in emerging areas like AI to stay agile.
A content delivery network, used as an example of a non-GPU data center operation, contrasting with the specific, power-intensive GPU data centers for generative AI.
An AI platform mentioned in the context of a cybersecurity incident, where an AI model broke out of a sandbox due to human error in server setup.
An example of a hyped company that failed spectacularly, linked to SoftBank, serving as a cautionary tale for OpenAI's potential fate.
Elon Musk's space company, whose increased valuation boosted Google's net profits on paper, highlighting the role of speculative paper gains.
A company that builds data centers and rents out GPUs, used as an example of a company being propped up by NVIDIA's circular financing to create demand for GPUs.
One of the Magnificent Seven companies, whose stock value is part of the inflated tech market that could impact retirements.
An autonomous vehicle company, described as fascinating but having socioeconomic and real-world edge case problems, with cars getting stuck in controlled environments.
An autonomous vehicle company whose cars were observed getting stuck and blocking exits in Las Vegas, highlighting real-world operational issues.
Meta's generative AI model, mentioned in the context of creating 'weird popup things' on Instagram, illustrating Meta's problematic AI integration.
A major investor in OpenAI, whose future is linked to OpenAI's ability to go public, highlighting the cascading risks of the AI bubble.
Used as an example of a company that lost money for a long time before becoming profitable, contrasted with AI's unprecedented burn rate.
An asset manager mentioned as potentially getting involved in propping up OpenAI through private credit, investing in data centers.
One of the Magnificent Seven companies, whose stock value is part of the inflated tech market that could impact retirements.
CEO of Meta, quoted for aggressively investing in infrastructure to meet demand, akin to a detached leader willing to accept risks for growth.
CEO of OpenAI, referred to as 'Clammy Sammy' for making exaggerated promises about AI's capabilities and for his company's unsustainable financial practices.
CEO of Anthropic, initially discussed as someone the speaker trusts more for a balanced view on AI risks, but later criticized for changing narratives and scare tactics.
Former CEO of Microsoft, famously mocked the iPhone upon its release, a historical example of skepticism towards disruptive innovation.
A journalist who reported on Bing's erratic AI behavior, which was then talked up by Microsoft's CTO, highlighting the overhyping of AI's capabilities.
CTO of Microsoft, who downplayed concerns about Bing's AI, illustrating the tech industry's tendency to normalize problematic AI behavior.
A Nobel Prize-winning economist who famously dismissed the internet's economic impact in 1998, a historical example of skepticism towards new technology.
An economist whose ideas are invoked to explain the current neoliberalistic hellscape of growth at all costs, leading to a lack of AI regulation.
Analysts praised for their long-standing insights into China's access to NVIDIA GPUs.
An individual who worked with Meta on their AI efforts, indicating the vast sums spent by Meta on AI with unclear returns.
CFO of OpenAI, who vaguely pushed back the timeline for OpenAI's IPO, indicating financial instability and a delay in going public.
Ed Zitron's editor, praised for his incredible context, knowledge, empathy, and critical thinking, contrasting human qualities with AI's limitations.
An astrophysicist who famously wrote in Newsweek in 1995, dismissing the internet's transformative potential, a historical example of skepticism.
CEO of Amazon, quoted for defending massive AI investments by stating they are not based on a hunch and aim for leadership and future profits.
Former CEO of Uber, used as an example of tech leaders who are not used to pushback and are disconnected from public sentiment.
A co-founder of Anthropic, noted for his past as a critical journalist and his current 'cult-like' association with Anthropic.
A person mentioned by Ed Zitron as part of his community of critical thinkers and 'haters' of the current AI narrative.
Head of Ads at Google and later Google Search, criticized for prioritizing search query numbers over quality, leading to a degraded user experience, and later running parts of Gemini.
Former head of Google Search, who recognized the conflict between increasing search queries and maintaining search quality.
An individual who was involved in taking over Google Search and pushing for increased query numbers.
Cited for his early warnings about AI's danger, and his company Tesla, is discussed in the context of self-driving cars and the Optimus robot.
CEO of NVIDIA, presented as someone who financially backs AI companies (like CoreWeave) by offering favorable contracts, enabling unsustainable GPU spending.
A Google engineer who voiced concerns about the company's strategy of increasing search queries at the expense of product quality.
A political figure whose policies are invoked to explain the current neoliberalistic hellscape of growth at all costs, leading to a lack of AI regulation.
COO of Uber, quoted for struggling to justify spending on AI tokens due to a lack of useful outcomes, supporting the argument against AI's productivity gains.
From ProfitG Markets, quoted on the 'Botox' analogy for tech companies investing in AI to appear young and growing, and the market's eventual skepticism.
CEO of Google, quoted for stating that the risk of underinvesting in AI is greater than overinvesting, illustrating the CEO's perspective on aggressive spending.
A former Google executive, who was booed during a commencement speech for mentioning AI, indicating public backlash against the technology.
A reporter for the New York Times who reported on OpenAI's failed attempt to go public.
A person mentioned by Ed Zitron as part of his community of critical thinkers and 'haters' of the current AI narrative.
A political figure whose policies are invoked to explain the current neoliberalistic hellscape of growth at all costs, leading to a lack of AI regulation.
Mentioned as having a blog where he discusses 'global AI sisterating global decision-making,' referring to the pressure on employees to AI-wash their work.
A software engineer who described AI as making 'the easy things easy, the hard things harder,' highlighting its limitations.
A scientist and critic of AI, cited for his belief that AI will hit hard limits and that exponential improvement is not guaranteed.
Mentioned in the context of people hoping for a bailout in an economic downturn, but the speaker dismisses this possibility for the AI bubble.
A data center being built by OpenAI and Oracle in Abilene, Texas, highlighted for its immense power consumption (1.2 GW) in a small area, symbolizing the massive infrastructure cost of AI.
A cognitive bias where people with limited competence in a particular domain overestimate their abilities, applied to people using LLMs to write code without full understanding.
A group of large tech companies whose stock values are inflated and whose potential decline could significantly impact regular people's retirements.
NVIDIA's GPU model, mentioned as a component within the Stargate Abilene data center, illustrating the specific hardware driving AI infrastructure.
Used as an example of a disruptive technology whose value was immediately obvious upon release, contrasting it with AI's perceived lack of clear, obvious value.
NVIDIA's GPU technology, mentioned in the context of China already having access to advanced NVIDIA GPUs despite sanctions.
Tesla's humanoid robot, questioned for its autonomous capabilities and presented as an example of overhyped AI in robotics.
Broadcom's chip, mentioned in the context of hardware efforts to improve AI, but still not making LLMs profitable or less costly.
Google's Tensor Processing Units, specialized hardware for AI, mentioned as failing to solve LLM problems despite multiple generations, indicating a hardware bottleneck.
A product created by the podcast host, designed to foster connection and conversation in relationships and teams.
Google's AI model, mentioned as being aggressively pushed into Google Docs and other services, forcing non-consensual adoption.
Microsoft's AI assistant, mentioned as being aggressively pushed into Word, contributing to the non-consensual adoption of AI.
Amazon's AI, mentioned as having opinions on purchasing behavior, illustrating the pervasive and non-consensual integration of AI into daily life.
OpenAI's language model, cited as an example of a rapidly adopted product due to media hype and pressure, but with underlying cost and hallucination issues.
Mentioned as an example of a Google product experiencing bugs and instability, possibly due to the aggressive integration of AI.
Mentioned as an example of a Microsoft product experiencing bugs and instability, possibly due to the aggressive integration of AI.
Mentioned as an example of a Google product experiencing bugs and instability, possibly due to the aggressive integration of AI.
NVIDIA's software library that allows software to run on GPUs, described as foundational to the growth of generative AI.
A programming language mentioned in the context of evaluating AI's performance, questioning how to measure a '5% improvement' across different languages.
A programming language mentioned in the context of evaluating AI's performance, questioning how to measure a '5% improvement' across different languages.
A financial data platform that offers an 'Ask B' feature, which uses AI to generate BQL queries for financial data, sometimes producing hallucinations.
A programming language used for Bloomberg inquiries, which AI can generate, but instances of AI hallucination with financial data are noted.
A software tool that allows a single mouse and keyboard to control multiple computers, used as an example of AI's useful troubleshooting capabilities.
Anthropic's model, whose adoption has been low due to high costs, especially for enterprises, as they shift to token-based payment models.
An OpenAI model that was erroneously reported by media outlets as 'blackmailing' a task rabbit, when it was a prompted experiment highlighting human error and media sensationalism.
Meta's generative AI model, mentioned in the context of creating 'weird popup things' on Instagram, illustrating Meta's problematic AI integration.
An OpenAI model, mentioned as costing half a billion dollars for a training run that failed, highlighting the immense and uncertain costs of AI development.
A programming language used in many LLM applications, praised as incredible for web scraping and data processing, but misused when AI is credited for its functionality.
Microsoft's search engine, mentioned as an alternative to Google Search, but also criticized for its AI integration ('AI crap').
Amazon's cloud computing platform, cited as an example of a successful tech venture that grew to profitability over a decade, but with significantly lower initial capital expenditure compared to current AI investments.
An eSIM app providing secure data connections in over 200 destinations, recommended for travelers to avoid roaming fees and simplify SIM card management.
Meta's generative ad model, mentioned in the context of creating 'weird popup things' on Instagram, illustrating Meta's problematic AI integration.
The publishing platform where Ed Zitron hosts his Substack/blog, mentioned as his current platform after moving from Substack.
An OpenAI model that Dario Amodei (while at OpenAI) claimed was 'too scary to release,' illustrating a history of scare tactics from AI leaders.
A service for hiring people for odd jobs, involved in a sensationalized story where an AI model was said to 'blackmail' one of its workers.
OpenAI's video generation model, mentioned as being shut down and still far from practical for creating full-length movies, highlighting AI's limitations in complex creative tasks.
An analyst group that exposed the hidden costs of AI subscriptions, revealing how much more users actually burn in tokens than they pay.
An asset manager mentioned as potentially getting involved in propping up OpenAI through private credit, investing in data centers.
A research firm whose study was misreported by journalists regarding AI's impact on youth employment, highlighting how narratives are shaped.
Criticized for being dysfunctional, allowing tech companies to make misleading statements about AI revenues and growth without accountability.
A magazine that published Clifford Stoll's famous skeptical article about the internet's future in 1995.
A Disney film referenced when Sam Altman's unrealistic promise of an AI 'genie' is compared to the charming but fictional character.
A newspaper mentioned for reporting on OpenAI's failed attempt to go public at a $1 trillion valuation.
A video game mentioned as an example where the speaker used Claude AI to troubleshoot a broken mod for his son.
A film referenced to characterize Mark Zuckerberg's statement about risking capacity building as akin to Lord Farquaad's detached willingness to sacrifice others.
A tech news publication where Jack Clark, co-founder of Anthropic, used to be a critical journalist.
More from The Diary Of A CEO
View all 510 summaries
125 minTop White House Advisor: The US Empire Is Declining And Socialism Is Coming Next! | David Friedberg
89 minThe Scientist Who Scans Fathers' Brains: The Parenting Checklist You NEED Before You Have Babies!
94 minFinancial Crash Expert: He Predicted The 2008 Crash, Now He Says Capitalism Failed, It’s Too Late!
132 minTop Scientist REVEALS: We Invented Heaven, God Is A Human Invention! | Prof Brian Greene
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free