Oracle's New AI Models Could Unlock 10%-Plus Cloud Apps Growth-If Trust Turns Into Usage

Oracle's real upside is a richer slice of enterprise software spend

Oracle's new AI models matter only if they move usage from demos to daily workflows. If that happens, cloud applications can grow meaningfully rather than merely hold the line. In the latest quarter, cloud applications revenue grew 10% year over year. That is useful, but it is not the full bull case.

The stronger argument is that applications sit closer to customer workflows than raw infrastructure. If AI becomes part of how users do work inside those apps, Oracle can capture a richer, stickier part of enterprise software spend.

Oracle already has demand; the question is where it shows up

Oracle has plenty of evidence that enterprise AI demand is real. Its remaining performance obligations rose $85 billion sequentially to $638 billion, and management said most of that increase came from large-scale AI contracts in which customers prepaid for GPUs or supplied them themselves. Oracle's co-CEO Mike Sicilia also said customers have moved past the experiment stage and are now focused on enterprise-grade deployment.

The bull case is that those commitments spread into applications, not just compute. The bear case is that AI demand is still concentrated in infrastructure, with broader application monetization arriving later than investors hope.

Why Oracle's built-in AI approach could matter more than raw model capability

Oracle's strategy is less about having the flashiest model in isolation than about placing AI where business work already happens.

Built-in AI is easier to govern-and easier to adopt

Oracle's approach is built-in, not bolted-on, with AI agents installed in the same environment as its prebuilt workflow agents. That fits Oracle's broader pitch of unifying trusted data, business context, and governed execution. For enterprises, that matters because adoption usually depends on more than benchmark performance: it depends on governed data, consistent business logic, and controls users can trust.

Oracle's own framing draws a clear line between task-level help and process-level automation. It describes everything from task automation and process automation to system automation and enterprise automation, including use cases such as candidate ranking, sales lead qualification, supplier risk scoring, and churn prediction. That is the practical difference: AI inside the app is not just assisting with a document; it is embedded in a workflow.

Integration lowers the friction to real usage

Oracle says Fusion AI can work with up to 50GB of non-Fusion data per tenant. That matters because enterprises rarely want to rewrite everything before getting value. The ability to connect private data that sits outside Fusion while keeping execution inside Oracle's governed stack makes adoption more practical.

That setup can also strengthen retention over time. The more processes that run through the system, and the more AI execution stays inside governed Oracle workflows, the higher the switching costs become.

The next test is whether application growth catches up with infrastructure growth

Oracle no longer needs to prove that AI demand exists. Last quarter already showed 21% total revenue growth and 47% total cloud revenue growth, with infrastructure driving much of the acceleration at 93% cloud infra growth. The key question now is whether the application layer can grow fast enough to justify the buildout and make the business look less like a capex-heavy story and more like a higher-quality software platform.

Cloud applications are growing, but not yet fast enough to settle the debate

Oracle posted 10% cloud apps growth. That shows demand is real, but it is not yet the kind of acceleration investors usually reward on its own.

Watch for two signals:

• Bull signpost: cloud applications keep building from 10% while application commitments continue to improve, showing that AI is becoming an upsell engine inside existing workflows.
• Bear signpost: infrastructure stays hot, but application growth stalls and the revenue mix remains dominated by compute.

Repricing depends on where the value accrues

As long as Oracle's growth is driven mainly by cloud infrastructure, the company remains exposed to concerns around data center build-outs, including delivery timing, funding, and margin pressure. If applications start contributing more of the growth mix, the business looks less like a data-center developer and more like a software platform positioned closer to customer workflows.

What would confirm the thesis-and what would break it

The scoreboard is straightforward: measure Oracle by whether AI is changing daily usage inside the apps customers already pay for, not by AI headlines alone. Oracle can point to most of the RPO increase in Q3 and Q4 came from large-scale AI contracts, and management has said customers are ready to implement enterprise-grade complete agentic solutions. The next few quarters should show whether that translates into broader workflow usage inside the applications stack.

Signals that support the upside

• Buyers are talking about implementation, not just exploration, after management said customers have moved past the experiment stage.
• Oracle's integrated approach is designed to support the full automation spectrum, from task automation through enterprise automation.
• Recent 10% cloud apps growth shows the application layer is not stagnant.

Signals that would weaken the case

• AI still looks stronger in demos than in everyday use, which matches Oracle's own warning that AI initiatives impress in demos but stall at scale.
• Users adopt assistants for light tasks but still resist letting AI run core process steps within defined controls.
• Infrastructure keeps doing almost all the heavy lifting while 10% cloud apps growth fails to broaden into a stronger application-led growth story.

If usage deepens, Oracle's new AI models can make the application layer smarter, stickier, and easier to monetize. If not, the quarter's bigger story will remain the cost and complexity of building the infrastructure to serve AI demand.