OpenAI has pulled ahead of Anthropic in winning over business users, according to new data, though both AI labs face a persistent problem. Enterprise customers switch between the two providers depending on model releases rather than staying loyal to one platform.
This volatility undermines the stickiness assumption that has fueled investor enthusiasm for both companies. When Anthropic rolls out Claude improvements, businesses test the new capabilities. When OpenAI ships GPT updates, they migrate again. The pattern repeats with each major release cycle.
OpenAI's GPT-4o and GPT-4 Turbo versions currently dominate enterprise adoption metrics. Anthropic's Claude 3 family gained ground after launch but hasn't converted switching behavior into lasting commitments. The churn reflects how businesses evaluate AI tools on performance benchmarks and cost rather than integration depth or switching costs.
For investors in both companies, the data signals a market structured around transactional relationships rather than network effects. Neither provider has achieved the kind of platform lock-in that makes SaaS businesses defensible long-term. An enterprise using Claude for document summarization will shift to GPT if OpenAI releases a superior model for that task.
The competition also intensifies as open-source models from Meta and others improve. Businesses now benchmark against free alternatives, further weakening enterprise pricing power.
OpenAI's lead reflects execution velocity and model performance rather than customer acquisition superiority. The company ships faster. Its models consistently rank higher on academic benchmarks. That advantage compounds over release cycles.
Anthropic positioned Claude around constitutional AI and safety, but enterprises prioritize capabilities over ethics in purchasing decisions. The positioning hasn't translated to differentiation at the budget line.
Both companies face pressure to build moat-like features beyond raw model quality. Agentic AI and specialized vertical models offer paths to stickiness. So do platform integrations