The Real AI Moat: Why Models Are Cheap But Data Orchestration Is Priceless

Harsh Verma is the Principal Software Engineer - AI/ML at Palo Alto Networks, an IEEE Senior Member & BCS Fellow.

​For much of the AI industry’s tremendous growth over the last few years, companies believed their advantage came from owning better models, with the goal of investing heavily to own superior models.

The AI landscape looks much different today. According to the Stanford AI Index Report 2025, performance gaps between leading frontier models are narrowing across many benchmarks. Open-source alternatives can also now perform tasks that only specialized systems could perform earlier.​

Since companies have less ability to differentiate themselves by the models they choose, building better data orchestration becomes the moat for the next stage of the AI era. Data orchestration is the process of knowing which data to use to keep AI infrastructure running smoothly. ​If AI is the engine, data orchestration is the fuel system. ​​

The Commoditization Of Intelligence​

According to recent Fivetran research, enterprises spend about $29.3 million annually on data, yet 73% of enterprise data initiatives fail to see the expected returns.

These initiatives are crucial for AI success, which depends increasingly on reliable, real-time, well-governed data pipelines.​​

When Bloomberg developed BloombergGPT, for instance, what differentiated them was their access to decades of proprietary financial data, market information, news content and domain-specific knowledge. ​​

The project highlighted an increasingly important reality: Unique data creates more defensibility than unique algorithms. Bloomberg’s moat is not its AI capabilities but the ability to feed high-quality financial intelligence into AI systems continuously.

As models are becoming more interchangeable, data orchestration becomes the strategic capability. Organizations should now ask, “Which model integrates better with our data, workflows and compliance?” ​​