The Man Who Calls BS on AI: The $1 Trillion Problem Nobody Wants to Talk About | Ed Zitron

Overview

Ed Zitron argues that today’s “generative AI boom” is fundamentally being misrepresented as a world-changing, autonomous intelligence revolution. Instead, he claims that—at least economically and practically—it behaves like an expensive, unreliable “cloud software” product sold through hype.

Core claims

Generative AI as a “con,” not honest software

Zitron says companies overpromise capabilities (and future outcomes like job replacement and major cures) while underdelivering on core functionality—such as:

• Search quality and basic usefulness
• The need for complicated “prompt gymnastics” rather than reliable autonomy

Business model relies on subsidization and undisclosed economics

He claims leading AI firms don’t earn revenue commensurate with their spending, asserting that:

• Reported AI revenue is concentrated in OpenAI and Anthropic, which he describes as unprofitable and dependent on continued funding from large tech partners.
• Major investors/partners (notably Amazon and Google) provide tens of billions, implying a transfer of capital rather than sustainable customer monetization.
• AI revenue disclosure is opaque, using vague “run rate” metrics or not clearly reporting AI revenue at all. He views this as a transparency problem for “the biggest software change ever.”

Cost and capacity expansion don’t match real value

Zitron emphasizes that AI development and deployment require enormous:

• Capex (data centers, GPUs)
• Ongoing inference costs (running GPUs based on demand)

He argues that costs keep rising even as pricing and profitability remain unclear.

Adoption vs. “value”

• Adoption is real but not necessarily voluntary or rational. He pushes back on the idea that massive user adoption proves value, arguing AI is increasingly non-consensually pushed through major platforms (search, docs, Office tools, shopping interfaces).
• AI usage is often “search-like,” but not well. Generative tools are frequently used similarly to search, yet can be unreliable or vague for “specifics.”
• “AI slop” as an output problem amplified by incentives. He argues generative AI helps scale low-quality content (“slop”), building on pre-existing “SEO slop” dynamics that reward ranking over human usefulness.

The token/cost argument

• AI service pricing masks real consumption costs. AI systems are billed via tokens (inputs + outputs + internal processing), while consumer pricing and enterprise rate limits can obscure true compute expense.
• Power users can be extremely costly. Heavy usage can cause the company to effectively subsidize that consumption by charging far less than the true compute cost.
• Enterprises resist paying true cost. He cites examples where enterprise pilots or transitions to cost-based pricing face pushback (e.g., companies allegedly burning through token budgets quickly), suggesting monetization doesn’t align with the pricing model.

Why it persists (and what he thinks is driving it)

• Market-share loss vs. eventual profitability. He frames the situation like past “growth-at-any-cost” cycles, but argues AI spending can’t be justified by current revenues.
• Venture/stock incentives keep funding flowing. He suggests funding continues because projects inflate metrics (number go up, stock pump) rather than because unit economics are working.
• The “AI bets paid off” narrative is questioned. Even when tech giants claim AI success, he argues it’s often not due to direct AI profit—while other revenue streams (ads, pricing changes, platforms) rise for separate reasons and AI spending remains massive.

Bottom line

Zitron concludes that the AI industry’s scale of investment (including training and inference infrastructure) is not matched by disclosed, sustainable financial returns. He argues hype, capital flows, and incentives keep the system going despite unfavorable or unclear economics—suggesting a possibility that powerful firms may not be operating as rational meritocratic winners, but rather chasing growth dynamics that investors tolerate.

Presenters / contributors

• Ed Zitron (presenter)