Revenue at AI-native companies that sell AI models or services is rising sharply, but a concentration among a small number of firms stands out.

On an annualised revenue basis, OpenAI and Anthropic account for nearly 85 percent of the top 10 AI-native companies. A recent report by The Information showed that, including Cursor, Cognition, ElevenLabs, Midjourney, Lovable, Suno, xAI and Perplexity in addition to OpenAI and Anthropic, the top 10 accounted for 94 percent combined among AI-native companies.

Some revenue is double-counted because coding AI tool Cursor and AI search service Perplexity pay model usage fees to OpenAI and Anthropic. Even so, some AI app developers are drawing attention as fast-growing AI app companies emerge in areas beyond coding AI.

Over the past 6 months, 13 app developers passed the milestone of $100 million in annualised revenue. Suno, a music-generation app for individual users, has now exceeded $200 million in annualised revenue, and coding AI tool Cognition has entered a $400 million track. Cursor, one of the best-known coding AI tools, has surpassed $1 billion in annualised revenue in recent months.

Voice AI startup ElevenLabs recorded $330 million in annual recurring revenue as of the end of 2025.

ElevenLabs CEO Mati Staniszewski said in an interview with Bloomberg, "We reached $100 million ARR within 20 months of founding in 2022, and then surpassed $300 million within 5 months." ElevenLabs is expanding beyond AI voice generation and voice agent technology into music production features. Last year it also entered the entertainment market by producing AI voice content in collaboration with well-known actors including Michael Caine and Matthew McConaughey.

Business models used by AI companies to generate revenue also appear to be expanding beyond subscription-based approaches, and more recently into advertising.

· Beyond subscriptions and usage-based pricing... the AI services industry stirs with ad-based revenue models · OpenAI steps up ChatGPT ad tests, including prioritising sponsored information in answers

OpenEvidence, a medical AI chatbot service provider, sells chatbot ads to doctors, pharmaceutical companies and others, in a way similar to how Google provides ads in its search engine. Its annualised revenue was estimated at $150 million as of the end of last year. The Information said OpenEvidence generated that level of revenue while selling only one-tenth of its ad inventory. It means annualised revenue could exceed $1 billion if the remaining ad space is fully utilised.

Founded in 2022, OpenEvidence is known as "ChatGPT for doctors". Its AI chatbot is trained on authoritative medical journals including the Journal of the American Medical Association (JAMA) and the New England Journal of Medicine (NEJM), providing reliable information that can be used in actual patient care. Co-founder Daniel Nadler said it is designed to refuse answers when it is not confident, to prevent "hallucinations" as much as possible.

A New York Times report in October said OpenEvidence supports about 15 million clinical consultations a month. That was about double 8.5 million in July in less than half a year, and it is being used at more than 10,000 hospitals and medical institutions across the United States. OpenEvidence raised $200 million in a funding round conducted late last year, and its company value was assessed at $6 billion.

While the number of fast-growing players is increasing, uncertainty surrounding AI companies remains in terms of sustainability. Even as revenue grows, many still cannot escape losses. AI coding startups seen as doing well also face the uncomfortable situation of competing with OpenAI and Anthropic, which lead the AI model market.

Anthropic's AI coding tool Claude Code in particular has emerged as a threatening presence in the AI coding market, with annualised revenue exceeding $1 billion. Not only engineers but also some non-developers use Claude Code for actual development work, and even engineers at Google, a rival to Anthropic, are also said to use Claude Code.