TLDR: Technology critic Ed Zitron presents a detailed case arguing that the generative AI industry is an unsustainable bubble, characterized by inflated valuations, circular funding schemes, and a profound lack of genuine demand. He highlights NVIDIA’s central role in propping up this ecosystem, OpenAI’s unrealistic financial projections and massive burn rate, and the widespread failure of AI products like Microsoft 365 Copilot to gain significant traction. Zitron contends that Large Language Models (LLMs) are inherently unreliable, cannot replace skilled knowledge workers like software engineers, and are plagued by uncontrollable costs, all while executives and the media perpetuate myths about their capabilities.
In a comprehensive and scathing critique, technology commentator Ed Zitron has laid out a detailed ‘Case Against Generative AI,’ asserting that the industry is in the midst of an ‘egregious bubble’ poised for a ‘fiery apocalypse.’ His extensive analysis, published in his ‘Where’s Your Ed At’ newsletter on September 29, 2025, dissects the financial, technical, and human factors contributing to what he describes as one of the ‘dumbest shit that the tech industry has ever done.’
Zitron’s central argument revolves around the unsustainable economics of generative AI. He estimates that the entire generative AI industry will generate approximately $61 billion in revenue in 2025, a figure dwarfed by the ‘hundreds of billions’ in capital expenditures and venture capital poured into it. He points out that, outside of NVIDIA, hyperscalers like Microsoft, Amazon, and Google, and OpenAI itself, there is ‘less than a billion dollars in AI compute revenue.’
NVIDIA’s Central Role and Circular Funding
A significant portion of Zitron’s critique targets NVIDIA, which he claims has a ‘virtual monopoly’ on the GPUs essential for generative AI. He alleges that NVIDIA is engaging in ‘extraordinary, dangerous measures to sustain growth,’ including investing in ‘neoclouds’—specialized GPU cloud providers like CoreWeave, Lambda, and Nebius. This creates a circular funding mechanism where these neoclouds raise billions in debt, collateralized by their GPUs and customer contracts, to purchase more NVIDIA GPUs. Zitron highlights that CoreWeave, for instance, has $25 billion in debt against estimated revenues of $5.35 billion, with NVIDIA being a major investor and customer, even signing deals to buy unsold capacity.
OpenAI’s Astronomical Costs and Unrealistic Projections
OpenAI, often seen as the face of generative AI, comes under intense scrutiny for its financial instability. Zitron projects that OpenAI will burn $115 billion over the next four years and needs ‘more than a trillion dollars’ for operational expenses and data center construction. He dismisses OpenAI’s internal projections of reaching $200 billion in annual revenue by 2030 as ‘ridiculous,’ questioning how its CFO, Sarah Friar, could approve such figures. He notes that OpenAI’s promised deals, such as a $300 billion commitment to Oracle and a ‘bullshit’ $100 billion ‘investment’ from NVIDIA (which is largely contingent on OpenAI building $125 billion worth of data centers), are financially unfeasible. ‘OpenAI cannot afford the $300 billion, NVIDIA hasn’t sent OpenAI a cent and won’t do so if it can’t build the data centers, which OpenAI most assuredly can’t afford to do,’ Zitron states.
Lack of Demand and Product Failures
Contrary to widespread hype, Zitron argues that there is a severe lack of genuine demand for generative AI products. He cites the dismal adoption of Microsoft 365 Copilot, which, as of August 2025, had only ‘around eight million active licensed users’ out of 440 million Microsoft 365 subscribers, a mere 1.81% conversion rate. This translates to approximately $2.88 billion in annual revenue for a product category that generates $33 billion per quarter for Microsoft. Zitron suggests that Microsoft stopped reporting its AI revenue in January 2025 because ‘they clearly don’t have any good news to share.’
He also points to the struggles of AI software businesses like Replit, which shifted to ‘effort-based’ pricing for its ‘Agent 3’ product, leading to user complaints of spiraling costs without concrete results. Similarly, Anthropic’s Claude Code, despite its popularity, generates only around $33 million a month in revenue, with users often burning compute costs far exceeding their subscription fees. Zitron highlights a ‘viberank’ leaderboard where users compete to see how much they can burn, with one user costing Anthropic $51,291 in a month on a $200 subscription. ‘That’s not a real business! That’s a bad business with out-of-control costs,’ he asserts.
The Myth of AI Replacing Workers
Zitron vehemently refutes the narrative that Large Language Models (LLMs) are replacing software engineers. He quotes engineers who describe coding LLMs as akin to a ‘slightly-below-average Computer Science graduate fresh out of school’ or a ‘clever but woefully inexperienced intern who can’t learn.’ Nik Suresh, a software engineer, notes that while LLMs can write code rapidly, their high error rate means he still has to ‘read and verify every line, which is frequently much slower than writing them myself.’ Colt Voege, a Principal Engineer, emphasizes that LLMs ‘do not understand problems; they digest and generate language tokens probabilistically,’ making them unreliable for complex software engineering tasks. Zitron concludes that ‘coding LLMs don’t actually replace software engineers, and never will, due to the inherent unreliability of Large Language Models.’
He argues that the jobs being ‘replaced’ by AI are primarily those in fields like translation, art direction, SEO, and copy editing, where ‘shitty bosses with little regard for their customers’ are eager to slash contract labor. Brian Merchant is quoted stating that ‘[AI hype] has created the cover necessary to justify slashing rates and accepting ‘good enough’ automation output.’
The ‘Business Idiot’ Phenomenon and Media Complicity
Zitron attributes much of the AI bubble to what he calls ‘Business Idiots’—executives who are ‘entirely disconnected from labor and actual production’ and view work solely as ‘outputs’ that can be automated. He states, ‘Generative AI exists for two reasons: to cost money, and to make executives look busy.’ He also criticizes the media for its complicity, accusing journalists of ‘dutifully writ[ing] it down and publish[ing] it without a second thought’ when executives make unsubstantiated claims about AI’s ‘transformational’ impact or job replacement. ‘The only thing ‘powerful’ about generative AI is its mythology,’ Zitron declares.
The Inevitable Collapse
Also Read:
- OpenAI’s $300 Billion Oracle Cloud Deal Ignites ‘AI Bubble’ Concerns Amid Massive Infrastructure Expansion
- Elon Musk Accuses OpenAI of Betraying Founding Mission, Declares it ‘Built on a Lie’
Zitron concludes that the generative AI bubble is ‘inevitable’ to burst. He highlights that venture capital could run out in six quarters, with OpenAI’s massive funding demands threatening to ‘sap the dwindling funds available from other startups.’ He dismisses comparisons to past tech bubbles, stating, ‘This is nothing like anything you’ve seen before, because this is the dumbest shit that the tech industry has ever done.’ He warns that the continuous perpetuation of the AI myth by executives and the media will ultimately lead to ‘retail investors and regular people’s 401Ks’ suffering when the bubble collapses.
