Ed Zitron: AI Boom Is a 'Con,' 2027 Crash Imminent

He points out that OpenAI, for example, incurred a $20.9 billion loss in the previous year. According to Zitron, these companies cannot provide a clear and convincing path to profitability.

He emphasizes that the industry is primarily burning through capital rather than generating sustainable returns. This pattern suggests a fundamental flaw in the business model, where the expenditures far outweigh any demonstrable income or future profit potential.

Zitron argues that despite the hype, the financial realities paint a grim picture, indicating that the current model is unsustainable and relies on continuous capital injection rather than organic growth.

Big Tech companies are intentionally opaque about their AI-specific revenues, often obscuring the truth with vague metrics. Ed Zitron notes that these companies rarely disclose their actual AI earnings to investors. When they do, they frequently employ metrics like 'annualized run rate,' which he dismisses as meaningless.

This lack of clear disclosure is identified as a major red flag for investors, indicating an unwillingness to provide concrete financial performance data. Zitron suggests that such tactics are used to mask poor performance and maintain an illusion of success, rather than presenting a factual account of their AI divisions' profitability.

He observes that AI features are being forcibly integrated into widely used software, such as Google Docs, Microsoft Word, and Amazon's platforms.

Users are being conditioned to believe that failure to adopt these AI tools will result in them 'falling behind.' This marketing tactic creates a sense of urgency and necessity, rather than allowing users to evaluate the tools based on genuine need. Furthermore, Zitron notes that AI is often being used as a substitute for traditional search functions, particularly due to a perceived decline in Google's search performance.

Ed Zitron highlights that on a $200 per month ChatGPT subscription, a user can burn through $14,000 worth of compute tokens, and on Anthropic's service, $8,000 for the same price point.

This discrepancy means that even on a $20-a-month service, users can consume $400 worth of tokens. Enterprises reportedly panicked in 2026 upon realizing the true costs of AI usage when faced with actual bills, leading to widespread concern.

Zitron asserts that companies are currently subsidizing users heavily, effectively masking the true cost of AI inference and delaying the inevitable reckoning with these unsustainable expenditures. This subsidy model allows for artificially high adoption rates but is not financially viable long-term.

Microsoft's financial performance serves as a stark example of the vast disparity between heavy capital expenditure on AI and the actual returns generated. Ed Zitron points out that Microsoft spent $115 billion in capital expenditures in a single year and intends to spend $175 billion the following year.

He argues that the math simply doesn't add up, as total AI revenue is minimal compared to the colossal cost of GPU infrastructure and energy required to run these systems. Zitron describes the current spending trajectory as a directionless 'egregor of capitalism,' signifying a massive, undirected investment without clear, profitable outcomes.

This highlights a critical financial imbalance where enormous sums are being poured into AI development without a corresponding increase in revenue, raising questions about the sustainability of such investments.

This framework helps analyze AI's current situation.

Historically, disruptive innovations like the automobile, which replaced horse carriages, initially faced high costs and unreliability. Despite these drawbacks, they eventually transformed industries. AI, in its current state, exhibits similar flaws to early automotive technology, being costly and often unreliable.

However, the dilemma highlights that such technologies, if they eventually mature, can reshape industries despite their initial imperfections. The question for AI is whether it is merely a flawed early-stage innovation or fundamentally uneconomical.

He argues that AI relies on 'harnesses' and feeding files, rather than true learning or contextual understanding.

Zitron explains that human collaboration, such as the editing process, provides more than just the final text; it offers unique insights and mutual learning that generic large language model (LLM) outputs cannot replicate. He illustrates this by referencing his editor, Matt Hughes, whose value comes from incredible context and knowledge, not just the ability to process information.

This highlights a perceived limitation of AI: while it can produce text, it cannot offer the nuanced understanding, personal context, or collaborative depth that human professionals bring to a task, suggesting a fundamental gap in its capabilities for certain types of work.

He states that consumers are not paying the 'honest cost' of these services, which are heavily subsidized by tech giants.

Zitron posits that if AI were sold at its actual, unsubsidized cost—if users had to pay the real per-million token rate or API costs—the demand and adoption rates would likely look significantly different. This suggests that the perceived value and utility of AI are artificially inflated by the massive investments and subsidies from major tech companies, obscuring its true market viability.

He refers to a point made by an individual who noted that before the iPhone's release, there were thousands of presentations detailing how shrinking GSM, Bluetooth, and Wi-Fi radios would inevitably lead to a device like the iPhone.

In contrast, Zitron argues there is no such clear path for AI. He points to a Goldman Sachs report that also notes the lack of a clear trajectory toward the promised outcomes of AI. Unlike the iPhone, which had a demonstrable path of component miniaturization, AI lacks a similar, predictable development roadmap.

Zitron concludes that companies have promised world-changing disruption without a proven business model or a clear, inevitable path for how AI will achieve these grand ambitions.

He suggests this investment is driven by a fear of missing out rather than genuine, diverse customer demand.

Zitron argues that these companies observed each other's massive investments in GPUs and concluded that such spending must be for a valid reason, leading to a circular system of mutual validation. This collective action props up a few unprofitable AI startups, creating an illusion of widespread industry growth.

He posits that the con is illustrated by the fact that everyone saw these tech giants making huge purchases, then assumed they wouldn't do so without reason, thus fueling further investment without critical examination of underlying profitability.

This suggests that while AI tools aim to boost productivity, they may also be introducing new forms of instability.

AI-generated code is increasingly found to be buggier than human-written code, leading to more system failures and requiring significant debugging efforts. Furthermore, the massive amount of AI activity, including training and inference, is physically straining the underlying infrastructure.

This strain contributes to a higher incidence of outages and reduced reliability across the tech industry, challenging the narrative that AI inherently improves software quality and efficiency.

He explains that these leaders are not objective observers but rather 'talk their book,' meaning they advocate for scenarios that directly benefit their companies.

Zitron notes that the narrative around AI constantly shifts based on what best serves the company's valuation at any given moment. For instance, a CEO is unlikely to express doubts about AI's viability; instead, they will champion its transformative potential, even if unproven.

He argues there is a significant lack of actual evidence to support the claims of widespread productivity gains related to AI. This strategy of constantly promoting an optimistic, often exaggerated, future for AI creates a self-fulfilling prophecy in the market, drawing in more investment despite a lack of concrete returns.

He states that every scam and con throughout history begins by rushing the target, emphasizing the need to act immediately.

CEOs and AI promoters employ fear-based marketing, warning potential customers and businesses that they will be 'left behind' if they don't adopt AI now. This creates immense pressure to make quick decisions without thorough critical thinking or evaluation.

Zitron suggests that this insistence on speed is a deliberate strategy to bypass careful consideration, allowing companies to push products and narratives without adequate scrutiny. He sees this as a hallmark of manipulative sales tactics rather than genuine, value-driven innovation.

Ed Zitron argues that the perceived danger of AI falling into the 'wrong hands' is misdirected; he believes current AI models are already controlled by the wrong people: corporate leaders like Mark Zuckerberg, Sam Altman, and Dario Amodei. His concern is with the inherent risks posed by those currently steering these companies.

Zitron emphasizes that security risks, such as hacking, are frequently exacerbated by human error within these organizations, rather than solely by the autonomous capabilities of AI itself. He contends that focusing excessively on external threats, like 'Chinese access,' diverts attention from the systemic risks posed by the domestic AI firms and their leadership.

This perspective shifts the conversation from hypothetical future dangers to present-day, institutional issues within the tech industry, highlighting a fundamental distrust of current corporate governance in the AI space.

Zitron argues that much of what is branded as 'agentic AI' is merely basic software scripts performing automated tasks, rather than truly intelligent systems. He also claims that industry leaders deliberately avoid discussing the actual harms caused by generative AI, such as environmental damage from vast data centers and potential data theft through large-scale training.

He maintains that true value still resides in human taste and judgment, asserting that these qualities cannot be commoditized or replicated by AI-generated content. He believes that the real benefits of automation are often conflated with the grand, unfulfilled promises of generative AI, which he sees as detrimental.

He uses LinkedIn posts as a prime example, arguing that posts generated by ChatGPT or Anthropic's Claude become generic and indistinguishable because everyone else is using them.

Zitron posits that the most valuable output now is 'irreplaceably human' work—content that stands out precisely because it avoids AI-generated blandness. He suggests that a truly great LinkedIn post, for instance, is one that clearly comes from a unique human perspective.

While admitting to using AI for specific technical troubleshooting, Zitron differentiates this practical application from the industry's grand promises of AI transforming creative work. He believes that for creative and insightful content, human input remains paramount.

He claims that the actual practical outcomes derived from AI are not materializing as promised, indicating a slowdown in meaningful progress.

Zitron points to examples like video generation, specifically mentioning Sora, noting that while impressive, its practical application is not yet commercially viable for professional film production. He suggests that the technology, despite its advancements, is still far from being integrated into established, high-stakes creative industries effectively.

He also asserts that building truly 'agentic' workflows—systems capable of autonomous action—requires fundamental structural engineering that extends beyond mere training data. This suggests that simply feeding more data into models will not solve the underlying challenges of developing truly intelligent and practical AI systems, reinforcing the idea that a significant barrier has been encountered.

Ed Zitron presents his 'dot-com bubble theory,' suggesting that major companies are investing heavily in AI and GPUs because they have run out of genuine hyper-growth ideas. He argues that this spending spree on GPUs served as a signal to the market, indicating that companies were 'doing AI' and driving stock growth without necessarily having a concrete strategy.

According to Zitron, this behavior mirrors the speculative investments seen during the dot-com era, where companies poured money into internet-related ventures simply to be perceived as innovative. He concludes that the current economy is being fueled by an 'AI bubble,' similar in nature to past tech bubbles.

This theory posits that the GPU spending is not a reflection of genuine, sustainable innovation but rather a tactical move to inflate market perception and valuations in the absence of other viable growth avenues.

He argues that this progress is directly tied to constant capital injection from venture capitalists, Amazon, and Nvidia, rather than organic, self-sustaining growth.

Zitron warns that once this money stops flowing, the progress will inevitably halt. He highlights that the scale of losses in current AI companies is unprecedented when compared to past tech giants during their formative years, indicating a fundamentally different and more precarious financial structure.

This perspective suggests that the AI boom is not built on a foundation of sound economics or profitability but rather on a continuous influx of speculative capital, making it highly vulnerable to any disruption in funding.

He emphasizes that such a breakthrough is not currently happening, despite ongoing efforts from various companies.

Zitron is critical of the reckless and damaging way data centers are being built, citing their environmental impact and the massive capital expenditure involved. He suggests that the current venture capital model favors these unsustainable data center investments over more traditional, stable businesses.

This argument underscores the immense economic inefficiencies inherent in the present AI infrastructure. Without a radical shift in hardware costs, Zitron believes the industry's reliance on ever-increasing compute power will remain unsustainable.

He notes that enterprises are increasingly moving towards token-based pricing models, which makes the true costs of AI usage transparent.

Uber's COO, Andrew McDonald, reportedly stated the difficulty in linking money spent on tokens to actual useful business outcomes. This sentiment reflects a broader trend where companies are realizing that the promised productivity gains from AI are not easily quantifiable or consistently achieved.

Zitron views this struggle to justify AI expenditure as a critical indicator that the AI bubble is not merely a sectoral issue but poses a systemic risk to the broader economy. The inability to demonstrate clear ROI could lead to a significant slowdown in enterprise adoption and investment.

Zitron highlights that VC returns since 2018 have been historically poor, indicating a systemic issue beyond just the AI sector. He explains that corporate accounting tricks allow companies to inflate their value by pegging it to speculative AI holding prices, creating an artificial sense of growth and success.

This reliance on speculative valuations rather than concrete profits or sustainable business models creates a dangerous bubble. Zitron suggests that this financial engineering masks underlying weaknesses and could lead to significant write-downs when the market eventually corrects.