AI skepticism has become a useful counterweight to ambitious product roadmaps, massive infrastructure spending and claims of imminent transformation. But skepticism is only as valuable as its underlying forecasts.
A new assessment by engineer and writer Dan Luu examines the record of Ed Zitron, a widely circulated critic of the AI boom. Its central conclusion: several of Zitron’s biggest company-level predictions—and the reasoning used to support them—have not held up against subsequent financial performance.
That is not a verdict on every concern raised about AI. It is a reminder that executives should evaluate bearish narratives with the same discipline they apply to vendor projections and internal business cases.
The flagship claim: big tech was “dying”
Luu focuses on a November 2024 talk in which Zitron described Meta as a “dying company” and characterized Google and Microsoft as companies unable to find growth, pushing AI out of desperation.
The financial record presented in the assessment conflicts with that framing. Meta’s revenue rose from $135 billion in 2023 to $201 billion in 2025, while operating income increased from $47 billion to $83 billion. Alphabet’s revenue rose from $307 billion to $403 billion over the same period, with operating income reaching $129 billion. Microsoft’s revenue increased from $228 billion to $305 billion, and operating income rose from $101 billion to $143 billion.
The source also cites continued growth through the first half of 2026 for all three companies.
These figures do not establish that every AI investment will generate an attractive return. They do, however, undermine a specific causal story: that AI spending by these firms is principally a last-ditch response to stalled core businesses.
Why the distinction matters
It is possible for a company to overinvest in a technology while its core business remains healthy. It is also possible for search quality, ad load, software quality or user trust to worsen even as reported revenue and profit grow.
Those are different claims, with different evidence standards.
Luu argues that Zitron often moves from evidence of a localized product issue—such as concerns about Facebook usage estimates or Google Search quality—to broad conclusions about corporate decline. In the Meta example, Luu questions reliance on third-party traffic estimates rather than company-reported product metrics and financial results. In the Google example, he argues that even a material Search problem would not automatically mean Alphabet lacks other growth engines, including YouTube and cloud computing.
For operators, this is a familiar analytical failure: treating a visible operational problem as proof of an enterprise-wide economic outcome. A product may be deteriorating in a meaningful way without the business being near-term unviable. Conversely, a business can post strong results while accumulating long-term product or regulatory risk.
Treat commentary as a hypothesis, not an input to outsource
The assessment’s larger criticism is methodological. It says Zitron’s writing uses many figures and links, but does not consistently connect them into a sound causal argument. That criticism comes from one reviewer and should itself be read critically; it is not an independent, comprehensive audit of every statement Zitron has made.
Still, the review offers a practical checklist for leaders consuming AI commentary:
- **Define the claim precisely.** Is the argument about model capability, customer demand, margins, market structure or a company’s survival?
- **Check the primary metric.** For a claim that a public company cannot grow, start with revenue, operating income, cash flow and segment performance—not a single third-party usage estimate.
- **Test the causal chain.** A flawed product decision does not by itself prove that an AI strategy is desperate or that a company’s economics are broken.
- **Separate horizon from certainty.** A business can be strong today while facing a credible multi-year disruption risk.
- **Demand falsifiability.** Good forecasts state what would prove them wrong and when.
What to watch next
The important question is not whether AI boosters or AI skeptics win the discourse. It is whether companies can translate AI expenditure into durable revenue, productivity gains or defensible product advantages.
Watch for segment-level AI revenue, inference and infrastructure costs, customer retention, pricing power, and evidence that AI features change workflow outcomes rather than merely engagement. Those measures will say more about the economics of the cycle than declarations that AI is either inevitable or doomed.
Skeptical analysis remains necessary. But for boards, founders and builders, the useful form is specific, measurable and willing to be wrong.
