Sam Altman, CEO of OpenAI, has announced that the company is considering introducing advertisements into the free version of ChatGPT, while paying Plus subscribers would continue to enjoy an ad-free experience. The statement was made during the London stop of the OpenAI Builders Tour, where Altman stressed that adverts would only be introduced if they were “thoughtful and tasteful” and did not compromise the user experience. The move could form part of OpenAI’s wider revenue strategy, as the company is currently operating at an annual loss of around $2 billion, largely due to the immense costs of running ChatGPT’s computing infrastructure.
At present, OpenAI’s operations are funded through several revenue streams, including Microsoft’s $13 billion investment, the $20-per-month ChatGPT Plus subscription, and its Enterprise and Team plans, which are priced at $25 and $30 per user respectively. Nick Turley, OpenAI’s Product Lead, confirmed at the same event that the company is actively exploring different monetisation options and has not ruled out advertising, provided it enhances the user experience and makes the service more sustainable. Advertising revenue could help offset ChatGPT’s heavy operating costs, which some estimates put at $700,000 per day, even as the service now boasts over 200 million active users worldwide.
Introducing adverts could prove an important step in building a sustainable business model for OpenAI, particularly as competition intensifies in the generative AI market with rivals such as Google’s Gemini, Anthropic’s Claude and Meta’s Llama. Altman emphasised that before rolling out advertising, the company would carefully consider user feedback and only allow adverts that add value and relevance. OpenAI has already moved towards diversifying its income streams with the launch of the ChatGPT Store in January 2024, enabling developers to sell their own ChatGPT-based applications and sharing revenues with the company. Advertising would therefore represent another step towards stabilising its finances while keeping the core model freely available to users.
Sources:
1.
John-Anthony Disotto
2.
Paulo Montenegro
3.
Chance Townsend
