Accounting audit of AI skepticism: Tech giants' numbers debunk shrinkage predictions
The critical review published by researcher and developer Dan Lo traced the predictive record of one of the most prominent skeptics of generative AI, Ed Zietron, and put his widely repeated claims to the test of explicit numbers and documented financial data. Zietron’s core thesis, formulated in late 2024, assumed that the largest tech firms, led by Meta, Google and Microsoft, were in a clinical death phase and were fumbling to embed AI into their products because they could not generate any organic growth, considering this rush a desperate attempt to rescue dying systems.
The accounting facts documented by Lo in his analysis show a reality completely different from those predictions, as the three companies continued to achieve record financial jumps in revenue and operating profit. Meta’s revenue rose from $135 billion in 2023 to $165 billion in 2024, then to $201 billion in 2025 and $117 billion in the first half of 2026, accompanied by operating profit that reached $83 billion in 2025. Alphabet, Google’s parent company, recorded a steady increase in revenue from $307 billion in 2023 to $403 billion in 2025 and $230 billion in the first half of 2026, while Microsoft posted $305 billion in revenue in 2025 with operating profit of $143 billion.
The analytical review finds that the skeptical discourse resorted to cherry-picking secondary indicators or inaccurate data from external sources such as SimilarWeb to promote a decline in Meta platform users, ignoring the consolidated financial figures and the continued growth in cloud computing services and YouTube.The failures were not limited to describing the situation of major companies, but also extended to the core of the technical and economic forecasts of the AI models themselves.The analysis identified a series of definitive forecasts issued by Zaitron, a research firm, since early 2024, claiming that the technology had reached its ultimate limits, that training data were exhausted, and that OpenAI’s growth had halted. These estimates have been repeatedly proven wrong by the development trajectory, as OpenAI has exceeded its revenue projections and introduced new inference models such as the O1 model.
An examination of the computational models used by Zytron, as presented by researchers Joho Senilman and Timothy Lee, revealed systematic, glaring errors in the spreadsheets, including the omission of time periods and the double-counting of days in Anthropic’s revenue forecasts, which produced misleading conclusions aimed at confirming pre-existing beliefs rather than offering a unbiased economic reading. This deconstruction highlights that turning technical doubt into a populist narrative built on faulty numbers gives followers a false sense of strategic reassurance at a time that demands the highest level of computational vigilance.
Considerable changes occur in the assessments of technology managers and digital transformation teams in the Gulf, Egypt and the Levant based on this documentary lesson.Reliance on claims that the AI boom is ending or that its financial systems are near collapse may cause regional institutions and startups to postpone building internal software capabilities, to slow investment in adapting models for local markets, or to hesitate in reserving the necessary computing power and cloud infrastructure. This scenario does not call for adopting marketing promises without scrutiny, but it does require decision makers and engineers not to use contentious theses as an excuse to stall their technical plans, since infrastructure-providing companies continually grow and expand their operating margins, and falling behind the pace of inference places the organization against competitors who move based on actual productivity metrics rather than media debate.