OpenAI Gains Ground on Anthropic in Enterprise Spending, Ramp Data Shows

OpenAI has begun reclaiming market share among U.S. businesses after losing its dominant position to Anthropic earlier this year, according to new spending data published by Ramp, a corporate credit card and expense management platform. The data tracked more than 70,000 American companies that collectively spend billions through Ramp’s bill pay and corporate card products, offering rare visibility into how enterprises allocate their artificial intelligence budgets across competing vendors.

Rapid shifts in market dominance

OpenAI held what appeared to be an unassailable lead in enterprise AI adoption throughout 2024 and early 2025, but that advantage evaporated this year. In May, Anthropic crossed a significant threshold, capturing 41% of AI spending among Ramp’s paying customers while OpenAI’s share fell to 39%. For the first time since generative AI gained broad enterprise adoption, OpenAI did not hold the top market position. The company has not recovered that lead since.

By July, nearly three months into the apparent shift, the gap had narrowed slightly but Anthropic maintained its advantage. Anthropic’s share stood at nearly 44% while OpenAI climbed to nearly 40%. Ramp economist Ara Kharazian noted in a post on X that OpenAI is currently growing faster than Anthropic among Ramp’s business customers during the third quarter to date, though he cautioned that the trend could shift with a month remaining in the quarter.

The data carries important limitations that temper sweeping conclusions. Ramp’s customer base skews heavily toward the technology industry and excludes large enterprises that rely on spend-management systems from established providers like American Express. The company declined to release actual dollar amounts, offering only percentage-based market share figures. Still, the trends within this dataset of over 70,000 businesses suggest genuine shifts in how enterprises evaluate and allocate their AI spending.

Model releases reshaping purchasing decisions

Read also: Ramp Launches Router, Its Own AI Model Routing Service · Binance Opens Trading to AI Agents With User-Controlled Limits · Calendly Launches Note-Taker Focused on Post-Meeting Automation Work · Micro1 reaches $500M annual run rate in data-labeling boom

GPT-5.6 Sol driving OpenAI’s recovery

The fluctuations in market share appear tightly coupled to model releases and their reception among developers and businesses. Kharazian attributed OpenAI’s recent acceleration to GPT-5.6 Sol, which he characterized as exceptionally capable and increasingly the preferred choice for developers working on complex tasks. The model’s stronger performance metrics apparently convinced some businesses that had recently switched to Anthropic to reconsider their spending allocation.

Fable 5’s struggles with adoption

Conversely, Anthropic’s high-end model tier, Fable 5, appears to have encountered resistance in the market. Kharazian noted that Fable 5 disappointed both in adoption rates and real-world application performance, despite Anthropic’s substantial engineering investment. Two factors contributed to this reception: the model’s premium pricing and regulatory-imposed data retention requirements.

Anthropic implemented a 30-day data retention policy on Fable users to comply with regulatory requirements, a decision that triggered backlash among businesses when the policy became widely known. While Fable is intentionally architected for specific, high-value use cases rather than general-purpose chatbot applications, the combination of higher costs and mandatory data retention created friction for enterprises trying to optimize their AI spending. Some businesses apparently decided that the additional privacy constraints and expense were not justified by incremental performance gains over less expensive alternatives.

Enterprise vendor selection as continuous evaluation

The data demonstrates that the enterprise AI market has entered a phase where businesses treat vendor selection as an ongoing process rather than a one-time decision. Companies are not yet locked into multi-year contracts or exhibiting the brand loyalty typical of mature enterprise software categories. Instead, they monitor new model releases, benchmark performance on their specific workloads, and adjust spending accordingly when they perceive better value or capability elsewhere.

Market expansion absorbs competitive intensity

The market share battle between OpenAI and Anthropic unfolds against a backdrop of rapidly expanding overall demand for AI services. While the two companies compete aggressively for a larger slice of enterprise spending, the total pie itself continues to grow. Among Ramp’s business customers, the percentage paying for AI services of any kind reached 50% in March. By July, just four months later, that proportion had climbed to nearly 56%.

This expansion means OpenAI can achieve revenue growth and profitability improvements even while losing share to Anthropic. The AI spending market is expanding fast enough that companies can switch vendors without the overall market contracting. This dynamic benefits both companies in absolute terms, though it creates anxiety for investors trying to assess the long-term defensibility of their competitive positions.

What enterprise adoption patterns reveal

The Ramp dataset provides a window into a critical phase in enterprise AI adoption. The market is no longer in an early stage where a single supplier dominates and customers face limited alternatives. Yet it has not reached the maturity of established enterprise software categories where switching costs are high and migration requires substantial investment and planning.

Businesses appear willing to test different models, run performance benchmarks against their specific workloads, and reallocate spending based on results. This fluidity presents both opportunities and uncertainties. For AI companies, it means abundant room for growth and rapid market expansion as more enterprises adopt AI-powered tools. But it also suggests that today’s market leaders cannot assume their position is secure. A better model release, a pricing reduction, or a superior feature set from a competitor can prompt significant customer migration within months.

Investor implications of market stickiness

The volatility revealed by Ramp’s data raises a fundamental question for investors in OpenAI, Anthropic, and other AI companies: How sticky is enterprise AI spending? In mature software categories, customers invest significantly in integration, training, and customization, creating high switching costs. A business that has built its workflows around one vendor’s API is unlikely to migrate to a competitor unless dissatisfaction becomes extreme.

The current AI market shows limited evidence of that kind of stickiness. Businesses appear to treat AI vendor selection as a tactical decision, reassessing every few quarters or even every few months when new models arrive. This creates both upside potential—the winner in any given quarter can capture substantial share—and downside risk—leadership positions can evaporate quickly if execution falters.

For OpenAI and Anthropic, the Ramp data suggests they are competing in a market where leadership can shift rapidly based on model performance and pricing. Neither company can afford complacency about its competitive position.