Several points worth noting: 1) Once you experience the benefits of intelligence, it's addictive. OpenAI's free users use GPT an average of 7 times per day. However, when they start using the first subscription tier, the average daily usage doubles to 15 times; with the second tier, it triples to 21 times; and with the highest tier, it increases 11 times. This is the terrifying and addictive nature of the business models of these large-scale AI companies. 2) Computing power supply is still far from sufficient. "If you do not have compute, you do not have revenue." In 2026, OpenAI even proactively rejected some business opportunities due to insufficient computing power. Codex grew from 100,000 developers to 2 million in three months, with demand growing far faster than computing power supply. Only those on the front lines can truly appreciate how much you can do once you have computing power. For ordinary users, you can pay directly with subscriptions or APIs, but for enterprises and developers truly using AI at scale, in the long run, purchasing computing power is definitely a more cost-effective business than paying by token. The booming inference market has significantly extended the economic lifecycle of all computing hardware, increasing returns, and this process will continue. 3) Creativity and initiative are always scarce. Interestingly, however, OpenAI, an AI company, prioritizes people above all other resources. They are aggressively recruiting top researchers because computing power can be bought, and data can be accumulated, but people who can define the next generation of product direction and make groundbreaking breakthroughs cannot be bought. This proves that no matter how powerful AI becomes, its creativity and initiative still come from people.