![\"TOPSHOT-CHINA-TECHNOLOGY-XPENG-AI\"](\"https:\/\/www.europesays.com\/ai\/wp-content\/uploads\/2026\/08\/1786129690_965_0x0.jpg\")
<\/p>\n<\/p>\n
TOPSHOT – He Xiaopeng, cofounder and chairman of Chinese electric vehicle maker Xpeng, launches Xpeng’s next-gen Iron humanoid robot during AI Day press conference at its headquarter in Guangzhou, in southern China’s Guangdong province on November 5, 2025. (Photo by Jade GAO \/ AFP) (Photo by JADE GAO\/AFP via Getty Images)<\/p>\n
AFP via Getty Images<\/p>\n
The AI debate gets reported as optimists against pessimists. But in reality, there is much more nuance, layer and tension then what appears in the headline on tweet. <\/p>\n
The premise the loudest camps share<\/p>\n
Two narratives dominate coverage, and both sit on the same side of the only question that counts.<\/p>\n
The first holds that AI works and that is precisely the danger. Dario Amodei doubled down<\/a> in a January 2026 essay calling AI a \u201cgeneral labor substitute for humans.\u201d Citrini Research\u2019s February note<\/a> imagined agents lifting output per hour to rates unseen since the 1950s while employment and demand collapse.<\/p>\n The second holds that AI works and the gains spread. Cathie Wood rebutted Citrini<\/a> on X, writing that ARK forecasts \u201ca productivity boom, an acceleration in real GDP growth\u201d alongside lower inflation.<\/p>\n Each requires AI to be extraordinary. Neither argues against the trade. Technology historian Lee Vinsel calls the pattern criti-hype<\/a>, criticism that accepts an industry\u2019s grandest claims at face value so that warnings double as marketing.<\/p>\n On July 24, Anthropic\u2019s head of economics Peter McCrory published<\/a> an X essay finding no material AI effect on US employment, citing a June unemployment rate of 4.2%. \u201cI don\u2019t expect unemployment to be noticeably higher a year from now,\u201d he wrote. His chief executive had spent a year saying the reverse.<\/p>\n The bear case is four arguments wearing one label<\/p>\n Below the capability line the disagreement is real, and it fragments immediately.<\/p>\n Michael Burry\u2019s target is the accounting. He posted on X<\/a> that \u201cunderstating depreciation by extending useful life of assets artificially boosts earnings,\u201d estimating $176 billion of understated depreciation industry-wide from 2026 to 2028, with Oracle overstating earnings by near 27% and Meta near 21%. The counterargument<\/a> is that a chip cascades from frontier training into inference and cloud rendering across six years, which makes the longer schedules defensible, and auditors keep approving them. The dispute reduces to whether inference demand grows fast enough.<\/p>\n Bill Gurley questions revenue quality instead, flagging circular deals<\/a> in which a technology giant invests in a startup that spends the money back on the investor\u2019s cloud. Ed Zitron goes at solvency, calling OpenAI one of the largest<\/a> liabilities in recent economic history.<\/p>\n The deployment critique belongs to MIT\u2019s Project NANDA, which found that roughly 95% of enterprise generative AI pilots delivered no measurable profit-and-loss impact. That number gets repeated far more often than it gets read. The report describes itself<\/a> as preliminary findings drawn from around 300 publicly disclosed initiatives, 52 organizational interviews and 153 survey responses collected at industry conferences. What has held is the mechanism rather than the figure; enterprise systems that cannot retain feedback or adapt to context.<\/p>\n A VC investor can be right about depreciation and wrong about deployment, and most commentary treats the package as indivisible.<\/p>\n AI Narratives: market map<\/p>\n Josipa Majic PredinThe camp everyone misreads<\/p>\n The remaining position gets filed as AI skepticism and is something else. Arvind Narayanan and Sayash Kapoor\u2019s essay<\/a> argues AI is transformative in the way electricity and the internet were transformative, with tempo set by adoption rather than invention, and identifies the \u201ccapability-reliability gap\u201d as the barrier to working agents. <\/p>\n Daron Acemoglu moved between camps this summer. He was the economist whose modest estimates gave policymakers room to wait. In July he signed a letter<\/a> with more than 200 researchers, sixteen of them Nobel laureates, warning of white-collar disruption. Nothing in his book improves from that reversal, which makes it worth more than any founder\u2019s timeline.<\/p>\n What allocators should take from it<\/p>\n