In 2026, why does capital prefer AI applications that “make money”?

On the eve of WAIC this year, Dark Side of the Moon released Kimi K3, which broke the circle immediately after its release. But the capital market’s attention falls more on another matter. In the past six months, the valuation of this star model company has increased sixfold, with a target of US$30 billion, and it is simultaneously promoting its IPO in Hong Kong. In a letter to all employees on New Year’s Eve at the end of 2025, founder Yang Zhilin wrote: Focus on Agent in 2026 and not target the absolute number of users. While their valuation increased 6 times in half a year, they also took the initiative to abandon DAU, an indicator that has been chased by the capital market in the past two years. It is enough to prove that although user scale is still important, it is no longer the only hard currency for the valuation of AI applications. The new narrative line has returned to the long-term sustainable commercialization of product forms. The benchmark for capital evaluation AI has changed: in the past two years, computing power subsidies have been popular in exchange for scale, but book DAU can only support valuation, not gross profit. The capital market has discovered that the enclosure strategy in the Internet era does not work in AI. The computing power cost and retention curve have "revealed their true colors" one after another, and popular products that once had high hopes have also begun to withdraw from the market in batches. This is the turning point of the industry. Capital no longer pays for the scale itself, but turns to commercial quality verification. Under the new logic, companies with solid business models have more opportunities to remain as "good money" and obtain better valuations. For example, Haiyi, which recently completed over 100 million yuan in Series B financing, is a stable company that is favored by capital: it is from Chengdu, facing the world, and takes "born for the C-end" as its business philosophy. In the past few years, it has continued to make popular AI applications, and its financial data has also been relatively good. After the tide recedes, what’s more worth asking is: What kind of AI applications can “make money while standing”? Commercial quality verification, the turning point has reached the AI ​​application layer. The mainstream narrative in the past two years is that capital has joined forces with companies to compete with Silicon Valley to burn money, so scale is prioritized, computing power is subsidized, opponents are eliminated first, and the gross profit issue is "left for later." This style of play works well during the capital easing period, but in 2025