The AI moat test: what Oracle and Adobe need to prove

Key takeaways

• Oracle must prove huge AI spending can turn contracted demand into durable cash flow.

• Adobe must prove AI strengthens its creative workflow rather than making professional software easier to replace.

• For investors, AI growth matters less than the capital required and the durability of what customers pay for.

On 10 September 2026, Oracle and Adobe report earnings on the same day. They offer one of the cleanest comparisons in the artificial intelligence boom.

Oracle sells databases, business software and cloud computing. Its challenge is physical: build enough data-centre capacity for booming AI demand without letting investment swallow the economics.

Adobe sells creative and document software such as Photoshop and Acrobat. Its challenge is almost the reverse. AI makes content cheaper and easier to create. Adobe must show that easier creation makes its ecosystem more valuable, not less necessary.

Oracle needs AI demand to become big enough. Adobe needs AI creation to avoid becoming too easy.

Oracle's problem: growth that eats cash

Oracle's AI story is already enormous. Fiscal 2026 revenue rose 17%, fourth-quarter cloud infrastructure revenue rose 93%, and its contract backlog reached USD 638 billion.

Demand is not the obvious problem. The bill is.

Oracle spent about USD 55.7 billion on capital expenditure in fiscal 2026 against roughly USD 32 billion of operating cash flow. Capital spending therefore equalled about 174% of operating cash flow, leaving free cash flow negative.

That spending may still create value. The question is whether future returns justify today's cheque.

Customers have prepaid for, or directly supplied, around USD 75 billion of graphics processing units tied to large AI contracts. That reduces Oracle's funding burden. Investors still need to watch how quickly backlog becomes revenue and how much cash remains after new capacity is built.

Adobe's problem: a moat AI can redraw

Adobe starts from the opposite position. In its latest quarter, it generated about USD 2.2 billion of operating cash flow while spending just USD 58 million on property and equipment.

That is attractive software economics. But AI attacks scarcity instead.

Photoshop became valuable partly because high-quality digital creation required specialist tools and skills. Generative AI lets almost anyone produce a credible image from a sentence. Canva, Figma and AI-native tools are lowering the barrier further.

Adobe's defence is becoming clearer. Rather than relying only on Firefly, it increasingly lets customers use outside models, including OpenAI's, inside Adobe workflows. The moat can shift from owning the best model to owning the place where professionals create, edit, approve and publish.

Adobe's AI-first annual recurring revenue, meaning subscription revenue linked to newer AI products, exceeded USD 500 million in the second quarter and had tripled from a year earlier. Yet it remains small beside Adobe's overall subscription base.

The test is monetisation. Does AI bring in users, improve retention and support pricing, or simply become another feature Adobe must include?

Adobe also enters earnings with a leadership transition. Anil Chakravarthy becomes chief executive on 1 December. Adobe shares closed at USD 266.51 on 4 September, down 6.7% that day and roughly 18% lower in 2026.

Same boom, different moats

The broader lesson reaches beyond these two companies.

AI infrastructure businesses often need enormous upfront capital. Their moat can come from scale, access to chips and power, contracts and high utilisation. Software needs less physical capital, but AI can make once-special features easier to reproduce.

Growth alone therefore tells investors little. A better question is what a company must spend to produce that growth, and what stops competitors from taking the revenue.

Oracle must prove the economics of demand. Adobe must prove the durability of differentiation.

Risks

For Oracle, warning signs include slower backlog conversion, data-centre delays or another sharp rise in funding needs.

For Adobe, watch retention, pricing and AI subscription growth. If AI usage rises but customer spending does not, the product may improve faster than the economics. Leadership change adds execution risk.

Investor playbook

• Separate revenue growth from cash generation. Fast growth financed by faster spending deserves different treatment from capital-light growth.
• Watch Oracle's backlog conversion, not backlog size alone. Contracts matter most when they become revenue and cash.
• At Adobe, track whether AI improves subscriptions, retention and pricing rather than simply adding features.
• Across AI, ask what remains scarce: chips, power, workflow, distribution, data or customer trust.

Two tests, one lesson

Oracle and Adobe report on the same evening, but they are solving opposite versions of the same problem. Oracle has customers lining up for AI capacity and must prove it can build enough without sacrificing too much cash. Adobe already produces abundant cash and must prove AI does not make its core value abundant too.

That is the useful lesson for investors. A moat is not simply fast growth, a famous brand or an AI label. It is the ability to earn attractive returns while competitors struggle to copy the economics. Oracle needs AI to become big enough. Adobe needs AI not to become too easy. Thursday's numbers matter, but those two questions matter more.