Anthropic Surpasses OpenAI in Enterprise AI Market Share

AI vendors are now competing for market share in an environment where businesses increasingly refuse single provider commitment. New spending data from Ramp indicates Anthropic has expanded its lead over OpenAI among US businesses. In July, Anthropic achieved 43.5 per cent paid adoption, a rise from June, while OpenAI reached 39.7 per cent. This represents a notable shift in enterprise AI purchasing behaviour.

Credit: UNSPLASH

An earlier Ramp analysis revealed that 52 per cent of customers paying for either Anthropic or OpenAI were paying for both. Ramp referred to many of these as "free agents," signifying companies are hedging positions. This multi vendor approach grants organisations greater negotiating power regarding price, data retention, security terms, and performance. Businesses can allocate different workloads to various models.

But the longevity of this behaviour remains uncertain once artificial intelligence embeds deeply within corporate systems. Enterprise software traditionally fosters vendor lock in, as companies invest heavily in training, data connections, and workflow development. AI could follow this path, or buyers might treat models as interchangeable components, compelling vendors to compete.

Anthropic's rise has been sharp. In Mar., 30.6 per cent of businesses used Anthropic, compared with 35.2 per cent for OpenAI. By July, Anthropic pushed to 43.5 per cent, accelerating its market position. Ramp reported 43 per cent of Anthropic customers switched from another generative AI vendor. This contrasts with traditional software categories where changes are costly, according to Forbes.

And new Claude and GPT models cause marked shifts in spending patterns within months. For enterprise technology executives, this creates a procurement dilemma: a leading vendor today may not lead tomorrow. Buyers must decide which dependencies are justifiable, not seek a permanent leader. The overall market for AI services continues its expansion, with paid adoption reaching 55.7 per cent by July.

Ramp's Aug. report suggests buyers are selective about the price for advanced capabilities. Anthropic's Claude Fable 5, launched in June, was priced at USD 10 per million input tokens and USD 50 per million output tokens. Despite capabilities, access was interrupted in June, then reinstated in July. Fable 5 accounted for only 6 per cent of Anthropic tokens purchased and 11.4 per cent of money spent, suggesting benchmark superiority does not guarantee willingness to pay for costly models.

So, multi model serving platforms are also gaining traction. 5.8 per cent of AI spending businesses used these by June, providing access to open source and proprietary models. Among platform users, 85.8 per cent still used OpenAI, 93.2 per cent used Anthropic, and 96.4 per cent used at least one. This indicates AI systems are increasingly seen as interchangeable commodities.

Multi vendor purchasing introduces complexities like diverse APIs, pricing, security, and challenges in swapping models and maintaining workflow consistency. Enterprise AI usage remains in its early, experimental phase. More AI applications will accumulate switching costs as they connect to internal data and employee routines. Current low loyalty may not endure indefinitely.

Yet, Ramp's numbers suggest a firm point: enterprise AI purchasing has not settled into a single vendor, one technological stack, long relationship pattern. Anthropic can lead paid adoption and still share customers with OpenAI. Open source access can rise without displacing either company. No entity should assume permanent advantage or declare a definitive winner yet.

• Enterprise AI customers are adopting a multi vendor strategy, rather than committing to a single provider.

• Anthropic has surpassed OpenAI in paid adoption among US businesses, reaching 43.5 per cent in July compared to OpenAI's 39.7 per cent.

• This shift offers businesses greater negotiating power but introduces complexities in managing diverse AI systems.

• Despite benchmark superiority, customers are selective, often opting for less expensive models over the most advanced.

• The enterprise AI market remains in an experimental phase, with no definitive long term vendor lock in established.

Source: Forbes