作者表示，大型语言模型不应被视为只会检索并重复资料的搜寻引擎，而应被视为能够综合资料并据此推理的系统。他说，AI 的发展速度前所未见，对 AI 的需求已经存在且正快速成长，同时受制于供给不足。他也说，coding-model 业务在过去一年或两年以来一直以极快速度成长，并追问为何投资人没有在该产业近期股市重挫前，先把 AI 对软体的影响反映进价格。
他说，AI 能吸收比任何投资人更多的资料、记得更好，并且在辨识过去模式方面做得更出色；但它也缺少与新发展、主观判断、直觉以及非完全可靠性相关的投资面向。至于 AI 是否是泡沫，他说这项技术是真实的、已经有需求，且正被大规模应用。他表示，现在更多资金正投入用于部署模型的 inference capex，而不是 training capex，且这种转变与真实需求相关，并已在转化为庞大营收成长；同时他也指出，目前部分 AI 营收具有循环性，因为 AI 公司彼此互相采购。
他区分了 hyperscalers 与新创公司，表示 Microsoft、Amazon 与 Google 都是获利极高的公司，其当前价格可能偏高或偏低，但不太可能被证明是灾难性地过度昂贵。相对之下，他指出有些新创在尚未说明策略或宣布产品时就获得数十亿美元估值，这些只能被视为乐透券：多数最终一文不值，少数赢家则变得非常富有。他说，今天更可能是低估而非高估 AI 的潜力，但这不代表 AI 投资就是便宜货或定价公平；他建议采取中等部位，并保持选择性与审慎。
The writer says large language models should not be viewed as search engines that retrieve and regurgitate data, but as systems that can synthesise data and reason from it. He says the pace of AI development is unlike anything seen before, that AI demand already exists, is growing rapidly, and is supply constrained. He also says the coding-model business has been growing at warp speed for a year or two, and asks why investors did not price in AI’s software impact before the sector’s recent stock market rout.
He says AI can absorb more data than any investor, remember it better, and do a superior job of recognising past patterns, while also lacking parts of investing tied to novel developments, subjective decisions, intuition, and less-than-total reliability. On whether AI is a bubble, he says the technology is real, already in demand, and being applied on a large scale. He says more money is now going into inference capex to deploy models rather than training capex, and that this shift is tied to actual demand that is already translating into massive revenue growth, while noting some current AI revenue is circular because AI companies buy from each other.
He distinguishes hyperscalers from start-ups, saying Microsoft, Amazon, and Google are enormously profitable companies whose current prices may be high or low but are unlikely to prove ruinously excessive. He contrasts this with start-ups receiving multi-billion-dollar valuations, including some that have yet to describe strategies or announce products, and says these can only be viewed as lottery tickets where most tickets end up worthless and a few winners get very rich. He says AI’s potential is more likely underestimated today than overestimated, but that this does not mean AI investments are bargains or fairly priced, and he recommends a moderate position with selectivity and prudence.