Which Economy Are You Building For?

Oracle did not miss the AI moment. It paid for it.

In fiscal 2026, Oracle reported record revenue of $67.4 billion, up 17%, with GAAP earnings per share up 34%. On the surface, that is a strong year. The company is not pretending demand exists. The demand is there. Its remaining performance obligations reached $638 billion, driven in large part by AI cloud contracts.

And still, the stock fell.

Not because the business looked small. Because the cost of serving the opportunity looked enormous.

Oracle's capital expenditure reached $55.7 billion in fiscal 2026. Free cash flow was negative $23.7 billion. The company raised $43 billion in debt financing during the year, along with $5 billion in equity financing, and said it expects to raise about another $40 billion through debt and equity in fiscal 2027. (Oracle notes that a portion of its AI contracts are prepaid or customer-supplied hardware, which reduces the capital it must raise — but the direction of travel is unmistakable.)

That is the AI infrastructure economy. It is real. It is growing. It is being built with debt.

This is not a criticism of Oracle. It is a useful signal. AI at scale is not weightless. It depends on data centers, power, chips, land, cooling, contracts, credit, and a level of capital intensity that belongs closer to heavy industry than software myth. The interface may feel frictionless. The economics behind it do not.

For creative businesses, that matters because it clarifies a category error — one I hear in the questions clients now ask.

A year ago, the question was whether to use AI at all. It has changed. Now it is quieter, and more serious: how do we use it without making the brand feel cheap, legally exposed, or culturally synthetic? The novelty has worn off. What is left is governance — rights, provenance, disclosure, and the question of who actually authored the work.

The economy building cloud capacity, GPU clusters, and AI compute is not the same economy that commissions a Chanel campaign, a Pirelli calendar, a beauty editorial, or a still-life series with a particular hand behind it. These markets may touch. They may use some of the same tools. But they do not create value in the same way.

One economy borrows to build capacity. The other compounds through taste, trust, relationships, authorship, and cultural fluency.

The market is already pricing this in.

Adoption is settled — that argument is over. The IAB reports that 83% of advertising executives now use AI somewhere in the creative process, up from 60% two years earlier. The tools are in the building. What has not settled is trust. In the same research, 82% of those executives believe younger consumers feel good about AI-generated advertising. Only 45% of those consumers actually do. That distance — between how confident the industry feels and how the audience receives the work — is precisely where a brand gets hurt.

Premium brands have already found the edge of that tolerance. Luxury's first serious experiments with AI-generated campaigns met enough public backlash that the more careful houses stepped back. The model that survived is not "hand the archive to a machine," but bespoke tools built and kept under creative direction — AI as an instrument, not the author. At the same time, the strongest brands have begun treating "made by a person," "artist-led," and a traceable creative chain as commercial signals, not sentimental ones.

That is the same line the law is drawing and provenance is making legible. Work with a human author behind it can be owned, licensed, and defended. Work without one cannot, quite. Content Credentials and the standards around it are turning authorship into a record a client can point to — which is why it is moving into the brief itself, not a footnote to it.

When AI companies and infrastructure providers raise tens of billions to serve demand, the lesson for a boutique creative firm is not to imitate scale. That would misunderstand the signal. The lesson is to know which economy you are actually building for.

Belle & Company does not need to become more like an infrastructure business. It needs to become clearer about the value that does not behave like infrastructure.

A photograph is not valuable because it can be produced at scale. A photographer is not valuable because they can generate endless options. A representation firm is not valuable because it can move faster than the market's loudest tools.

The value sits elsewhere. In selection. In conduct. In the ability to know what should be made, who should make it, and why the result deserves to stand apart from everything else being produced.

This is where the boutique argument becomes harder, and more useful. It is not about resisting AI. It is not about sentiment. It is not about pretending the tools will not change the work.

They will. But the tools belong to a different economic logic.

The AI infrastructure cycle rewards capital, capacity, and utilization.
The serious image-making economy rewards judgment, authorship, and trust.
Those are not interchangeable.

When the leverage cycle turns, some of the infrastructure bets may prove brilliant. Some may not. That is for investors, lenders, and operators to sort through.

For creative firms, the question is simpler. Which economy are you building for?