The Commodity Trap: Why Enterprise AI Loyalty is a Myth
AI-generated image · US National Wire
New spending data suggests businesses are treating LLM providers as interchangeable utilities, pivoting based on raw model performance rather than ecosystem stickiness.
In the enterprise software world, 'stickiness' is the holy grail. It is the moat that protects margins and ensures predictable recurring revenue. But as the battle between OpenAI and Anthropic intensifies, the data suggests that generative AI is behaving less like a sticky platform and more like a volatile commodity.
As first reported by TechCrunch, data released by Ramp—a corporate credit card and expense management firm—reveals a striking lack of loyalty among U.S. businesses. In May, Anthropic surged ahead of OpenAI, capturing 41% of the market share compared to OpenAI's 39% among Ramp's paying business users. By July, that lead widened, with Anthropic holding nearly 44% and OpenAI sitting at nearly 40%.
However, the pendulum is already swinging back. Ramp economist Ara Kharazian noted on X that OpenAI is currently growing faster in the third quarter to date. This volatility is driven by a simple, ROI-centric calculation: which model performs best right now?
**Opinion:** From an operational lens, this indicates that enterprise buyers are not investing in 'ecosystems'; they are purchasing raw capabilities. When a new model drops that outperforms the previous leader, the cost of switching appears low enough that companies are willing to 'flop back and forth.' This transforms AI spend into a race to the bottom where the only sustainable advantage is being the absolute best performer at any given moment.
Performance is the only metric that matters, and it is a fickle one. Kharazian attributed OpenAI's recent growth to the fact that 'GPT-5.6 Sol' has become an increasingly preferred choice for developers. Conversely, Anthropic's 'Fable 5' model reportedly disappointed in terms of real-world application and adoption. The friction wasn't just technical; it was financial and regulatory. Fable is noted as being expensive, and Anthropic faced backlash after notifying users that it must retain data for 30 days.
While the overall market is expanding—with the percentage of Ramp customers paying for AI climbing from over 50% in March to nearly 56% by July—the instability of market share is a warning sign. Ramp's data, which covers more than 70,000 American businesses (skewing toward the tech industry), suggests that neither lab has secured a permanent win.
For investors and executives, this volatility should be a cause for concern. If enterprise AI spending isn't sticky, then the current growth isn't a sign of brand loyalty, but a reflection of a market in a constant state of audition. In this environment, the provider with the most 'sticky' product isn't the one with the best brand, but the one who can maintain a performance lead just long enough to keep the churn at bay.