Keynes’s Beauty Contest: If Everyone Is Bullish, Where Is the Customer?

When enthusiasm for a technology becomes widespread, a basic question can become surprisingly difficult to answer: who will keep paying for it?

A rising asset price can reward its owner. A service that customers keep using may sustain a business. These processes can support each other, but they do not establish the same thing. A willing buyer in the market is not necessarily a willing customer for the product. That distinction is what I want to bring from Chapter 12 of Keynes’s The General Theory of Employment, Interest and Money into today’s technology discussion.

It is not an argument for dismissing everything popular. A valuable new market may begin with incomplete evidence. The harder task is identifying whether our confidence rests on future business income or on someone else eventually offering a higher price.

Whose preferences are you predicting?

In Chapter 12, “The State of Long-Term Expectation”, Keynes describes a newspaper beauty competition. Success depends on anticipating the preferences of the other contestants. Choosing the faces you personally find most attractive is therefore a different exercise from predicting which faces the group will favour.

Applied to markets, a belief in a business’s long-term earning power differs from a prediction that other people will like its shares more next month. The latter can be sophisticated and profitable. It still does not, by itself, tell us whether a product solves a worthwhile problem.

Following the crowd need not mean failing to understand this distinction. Someone can recognize uncertain long-term prospects while also knowing that their work is judged against a short-term ranking. Individual intelligence may then be directed towards anticipating other people’s moves.

Yet calling everyone irrational would miss much of the chapter. Keynes also asks how limited knowledge about the future interacts with the way investment markets are organized.

Spending is happening; the eventual return is another question

Recent data make the distinction concrete. The U.S. Bureau of Economic Analysis’s October 29 advance estimate for the third quarter of 2026 identifies investment as a contributor to growth. Its technical notes say that information-processing equipment, mainly computers and peripherals, led the increase within equipment investment.

These are observations about expenditure. The advance estimate remains subject to revision, and not every computer investment is an AI investment. Nor does buying equipment establish that the eventual customer demand will justify the expenditure.

The IMF’s October 2026 World Economic Outlook includes both potential gains from AI investment and productivity, and downside risks from a reassessment of AI expectations. Those are outlook scenarios, not a report card for any particular business.

Investment, technical progress and rising asset prices can all be real. They still leave room for questions about who pays, why they return and what it costs to serve them.

Give each source of income its own explanation

Imagine a company introducing an AI service that attracts considerable market attention. This is a hypothetical example, not a claim about a particular company.

A shareholder might benefit because another investor offers more for the shares. A customer might pay because the service saves time, improves quality or makes something newly possible. Resale income and operating income play different roles. Financing can give the business time to develop customers, so the two are connected without becoming interchangeable.

I would separate the discussion into a few concrete questions. Is the valuation argument mainly about the next buyer or about future business earnings? If it is about earnings, what does the customer gain, and why might they continue using the service? What computing, support, human review and other costs are necessary to deliver that income?

Retention, renewals and delivery costs are my contemporary extension of the question. They are not a checklist written by Keynes. Research, infrastructure and a mature subscription service also develop on different timelines; a single revenue threshold cannot sensibly judge them all.

My earlier essay on who pays for GPU and data-centre investment follows the movement of costs. This reading asks a different question: which imagined source of income supports the confidence embedded in a price?

The option to leave can make it easier to enter

Stopping at “focus on long-term value” would leave out an important tension in Keynes’s argument.

An individual’s ability to sell an asset may encourage them to take a risk they would otherwise avoid. But everyone cannot simultaneously turn an industry’s factories or other committed resources back into cash. An asset changing hands does not undo the resources already committed to the enterprise.

Resale is therefore not inherently a problem. Making exit much harder could also make genuine businesses harder to finance. The concern is what happens when anticipating short-term market preferences becomes the dominant influence on the direction of an enterprise. How much attention remains for the people who will actually use its products?

My position has a cost, too. Giving greater weight to customer evidence can mean missing a price rally or underestimating demand that has not yet matured. I cannot ask for patience while pretending that patience carries no opportunity cost.

Evidence can strengthen confidence; optimism still matters

If paying customers continue using a service, delivery becomes more reliable and the costs of earning income become clearer, confidence in the business should increase. A previously overheated story does not make later progress irrelevant.

Keynes also gives spontaneous optimism a place in enterprise. Calculations alone cannot settle every decision about an uncertain future. Waiting for complete knowledge can prevent worthwhile activity from ever beginning.

I find it more useful to keep two questions together: what makes people willing to commit resources, and whose work or life becomes better enough to justify paying for the result?

The next time someone says that everyone is bullish, ask which kind of expectation is being discussed: a price someone hopes to obtain, or a reason a customer might stay. Clarifying that difference may reveal whether the people in the room agree at all.

Listen to the conversation

Listen to the Mandarin conversation with M.K., Dorami and a reconstruction of Keynes for another way into market expectations, liquidity and customer value. The historical guest is reconstructed from The General Theory.

Frequently asked questions

Does Keynes’s beauty contest mean all investing is gambling?

No. The analogy describes how participants can turn towards predicting one another’s preferences. The same chapter distinguishes speculation about market psychology from enterprise concerned with lifetime earnings, and examines liquidity’s benefits as well as its problems.

Why are market prices and customer value different?

Prices incorporate expectations about the future and about other traders. Customer value concerns why a product or service is worth using and paying for. A price can reflect expectations of that value without proving that it has already materialized.

Does a new technology without stable revenue have no value?

That does not follow. Research, infrastructure and new markets can take time. The argument is to identify the basis of confidence while acknowledging the opportunities that waiting may miss, rather than imposing one revenue test on every early enterprise.

Do recent investment figures establish adequate returns from AI?

No. The BEA release describes economic activity, while the IMF outlook includes forecasts and risk scenarios. Neither independently establishes a particular company’s customer demand, costs or future earnings.

Original text and sources

• Keynes: The General Theory, Chapter 12, “The State of Long-Term Expectation”
• IMF: World Economic Outlook, October 2026
• BEA: GDP advance estimate, third quarter of 2026, released October 29

Related reading: How AI productivity gains reach people’s lives.