A second technology spending race is forming, and it’s centered around AI. The first race was paid acquisition, and from the outside, the two patterns look identical. However, one was mostly a spending mistake, while the other is turning into a structural one.
Ming-Yuan Xie, a Taiwan-based serial entrepreneur and founder of Meow Universe, spent heavily on paid acquisition while building e-commerce businesses that generated over NT$2 billion in cumulative revenue. The prevailing logic at the time was to bid up channels to avoid losing customers to competitors.
Xie now sees the same fear driving AI budgets in many companies. The conversation around AI budgets is often focused on not falling behind, rather than evaluating the unit of return on investment. This framing borrows the emotional logic of a growth-marketing race and applies it to a very different kind of expenditure.
A capital allocator, faced with any spending proposal, asks three key questions: what is the unit of return, when does it become measurable, and what commitment does the spending create beyond the dollars themselves. These questions are essential for evaluating AI budgets, but they are often overlooked in practice.
Related: Miley Cyrus Beyoncé join tributes to Dolly Parton
The first question, what is the unit of return, requires teams to name the atomic thing they are buying per dollar. This could be a resolved support ticket, an hour of engineering time returned, or a research task completed without human intervention. If a team cannot name this unit, they are not evaluating a capital allocation, but rather describing a hope.
Xie’s experience with paid acquisition informs his views on AI spending. He argues that companies should know what they are buying before committing budget, rather than after the first disappointing quarter.
The third question, what commitment does the spending create, is often skipped entirely. However, it is essential for separating a marketing spend race from a capital allocation decision. When Xie bought and sold companies, he had to figure out what was truly reversible in the business and what was structurally locked in.
AI spending has a hidden second layer, which is usually larger than the invoice suggests. Companies often restructure functions around assumed capabilities, staff functions to run on the assumption that a tool will keep performing, and rewrite job descriptions to remove work the tool is expected to absorb. These structural bets can be costly to reverse, and the cost is rarely visible in the AI budget line.
Related: A guide to how air freight works
Xie argues that companies should be able to answer what commitment they are making, alongside what they expect to get back and when.
This approach will help companies avoid making the same mistake that Xie made with paid acquisition, but on a much larger scale with AI. As companies invest in AI, they should consider the long-term implications of their spending.
It’s not an argument for spending less on AI, but rather for evaluating AI budgets with a heavier standard. By doing so, companies can make more informed decisions about their AI investments and avoid costly mistakes.
