Ed Zitron Slams Generative AI: Questionable Business Value and the Growth Anxiety Behind Big Tech's Bets

A Public Challenge Aimed at the Heart of the Industry

Tech critic Ed Zitron recently appeared on CNBC to deliver a provocative broadside against the current generative AI frenzy. His core argument is simple and sharp: generative AI (GenAI) hasn't truly "worked," and the reason Big Tech is going all-in on AI is essentially that they've run out of new stories to justify "hypergrowth."

The remarks sparked widespread debate on Hacker News and other tech communities. As a commentator long known for cold-water-pouring, Zitron's views may not represent the mainstream — but he touched on a question that more and more people are quietly asking: Is the massive investment in generative AI built on solid commercial returns, or on capital markets' hunger for growth narratives?

Core Argument #1: Where Is Generative AI's "Commercial Deployment" Actually Stuck?

When Zitron says generative AI "doesn't work," he isn't claiming that models can't generate text or images — those capabilities clearly exist. What he's really questioning is commercial viability: whether these technologies create value commensurate with their cost.

The Inherent Tension Between Technical Nature and Cost Structure

Generative AI refers to AI systems capable of producing new content (text, images, code, audio, etc.) from input prompts. At its core are large language models (LLMs) built on the Transformer architecture. These models learn statistical patterns of language through pre-training on massive datasets, enabling them to generate coherent, contextually relevant content. However, training a GPT-4-class model is estimated to cost over $100 million, and every inference call consumes significant GPU compute. Companies like OpenAI have long been trapped in a "the more you sell, the more you lose" dynamic — a direct reflection of unclear unit economics, and the fundamental technical underpinning of the industry's commercial struggles.

The Vast Gap Between Revenue and Cost

This critique is set against an economic challenge the entire industry faces. Training and running large language models is extraordinarily expensive, requiring enormous investment in compute, energy, and data centers. Yet sustainable revenue growth directly attributable to generative AI remains relatively limited. Many AI features are bundled into existing products as "added value" but fail to drive incremental paying behavior.

The Chasm Between "Demo-Worthy" and "Production-Reliable"

Zitron's critique also implies a deep concern about reliability. Generative AI is stunning in demos, but in production environments demanding high accuracy and predictability, hallucination remains a major deployment obstacle. Hallucination refers to models generating content that sounds plausible but is factually wrong or entirely fabricated — a consequence of the model's fundamental nature as a probabilistic text predictor rather than a system with real-world knowledge retrieval capabilities. In high-stakes production environments like legal documents, medical advice, and financial analysis, hallucinations can render a model completely unusable. While techniques like Retrieval-Augmented Generation (RAG) and Chain-of-Thought prompting have helped mitigate the problem, they haven't eliminated it. The gap between impressive demos and difficult real-world deployment is the most compelling real-world basis for the claim that it "doesn't work."

Core Argument #2: Has Big Tech Run Out of Growth Engines?

Zitron's more striking claim is this: Big Tech is going all-in on AI because it has exhausted every other direction capable of sustaining hypergrowth.

The Capital Market Logic Behind the Hypergrowth Narrative

Hypergrowth — typically defined as annual revenue growth exceeding 40% — is the central narrative framework Silicon Valley venture capital and Wall Street use to value tech stocks. Big Tech enjoyed over a decade of hypergrowth during the mobile internet and cloud computing eras, leading markets to assign valuation multiples far above those of traditional industries. As those engines mature, the rationale for sustaining high valuations comes under fundamental challenge. Under this structural pressure, management and investors alike have strong incentives to find new "growth stories," even if commercial validation is incomplete.

Traditional Growth Engines Are Maturing

Looking back over the past two decades, the tech industry has ridden several clear growth waves — PC internet, mobile internet, cloud computing, social networks. Each brought explosive gains in users and revenue. But those engines have broadly matured: the smartphone market is saturated, cloud growth is decelerating, and social network user counts have plateaued.

Against this backdrop, Wall Street still expects trillion-dollar companies to maintain rapid growth. Zitron argues that generative AI arrived at exactly the right moment to serve as a "lifeline" for this narrative need. In other words, the AI boom may be driven more by capital market expectations than by the technology's own maturity.

The Hidden Risk of Narrative-Driven Investment Decisions

This perspective raises a troubling question: if Big Tech's AI investment is largely about proving to investors that "we're still growing," then investment decision criteria may have drifted away from genuine commercial return logic. When capital is abundant and competitive anxiety (FOMO) runs high, rational cost-benefit analysis is easily overwhelmed by growth anxiety.

How to Rationally Evaluate This Skeptical Voice

It must be acknowledged that Zitron is a well-known industry critic with a strong positional bias, and it's easy for his arguments to swing to the opposite extreme. Wholesale acceptance of his judgment would be equally irrational.

The Reasonable Core Worth Taking Seriously

Despite the sharp rhetoric, Zitron's critique does identify several genuinely real problems:

Unclear unit economics: Many AI products may currently cost more to operate than the revenue they generate.
No killer app has emerged yet: A "killer app" is one that is sufficiently essential and universal to independently drive mass adoption of an entire technology platform. The closest generative AI has come to this is coding assistance — GitHub Copilot is used by millions of paying developers, and AI-native IDEs like Cursor have quickly amassed large paying user bases. Yet coding assistance is fundamentally limited to a technical audience, capping its scale. For ordinary consumers, converting "active use" into "stable paying relationships" — the retention and monetization efficiency question — remains the unresolved core commercial challenge for the entire industry.
Market valuations disconnected from commercial reality: Current valuations embed heavy optimistic assumptions about the future that have not yet been validated by sufficient commercial data.

The Other Side Not to Ignore: The Historical Precedent of Cloud Computing

On the other hand, simply labeling generative AI as "not working" ignores the real value it has already created in specific domains. Code assistance tools have transformed many developers' workflows, and AI has genuinely improved efficiency in customer service, translation, and content drafting.

The journey from early investment to scaled profitability often takes years — cloud computing offers the most compelling historical reference. After Amazon launched AWS in 2006, it was for a long time viewed as a support department for the core e-commerce business, with massive upfront infrastructure costs and an unclear profit model. It wasn't until AWS disclosed its financials separately in 2015 that the world realized it had become Amazon's primary profit engine — a journey that took nearly a decade. Yet this analogy also implies the reverse risk: not every heavily funded technology can replicate AWS's path. Cloud computing had unique network effects and user lock-in; whether generative AI possesses equivalent structural competitive advantages remains to be seen.

Practical Implications for Businesses and Practitioners

Whether or not you agree with Zitron's assessment, his critique provides a necessary dose of sobriety for an overheated industry.

For businesses, the key question isn't "whether to invest in AI" but how to convert AI capabilities into sustainable commercial value — avoiding blind deployment in the pursuit of a compelling narrative. For practitioners, rationally distinguishing between "technical potential" and "current deployment maturity" is more valuable than chasing the latest trend.

Is generative AI the next world-changing technological revolution, or a growth bubble inflated by capital? The answer likely lies somewhere in between. And voices like Zitron's — however uncomfortable — are an important counterweight against the industry falling into collective mania.