OpenAI chairman says training your own AI model is a good way to 'destroy your capital'

Bret Taylor, the chairman of OpenAI, recently shared a stark warning for entrepreneurs considering the development of their own AI models. In a discussion on the Minus One podcast, he likened the endeavor to burning cash, citing it as a financial risk that could rapidly consume millions of dollars. Taylor, drawing from his extensive experience with companies like Google, Facebook, and OpenAI, advised against the creation of new frontier AI models unless one is part of a major player in the field. He emphasized that for most founders, building such models is prohibitively expensive and would ultimately lead to industry consolidation. The high capital requirements have effectively stifled the emergence of an independent data center market, according to Taylor. Instead, he encouraged aspiring AI entrepreneurs to collaborate with established giants in the industry, such as OpenAI, which profits from the sale of API access tokens that developers can integrate into their applications. International competitors are testing this model, as seen with the launch of DeepSeek's R1 reasoning model and chatbot. By utilizing less advanced technology, DeepSeek managed to minimize costs, resulting in its application topping the App Store charts and sparking discussions about the spending habits of tech giants in AI development. Taylor also proposed alternative pathways for entrepreneurs within the AI landscape. He highlighted the potential of the AI tools market, likening it to the 'pickaxes' of a gold rush, but cautioned that this space could be precarious. Moreover, he predicted that what are currently SaaS applications will evolve into agent companies by 2030. Ultimately, Taylor warned that creating custom AI models is a surefire way to 'destroy your capital,' describing them as 'fast-depreciating assets' that could cost millions. He advised exploring options such as leasing or utilizing open-source models instead, as the financial risks of building from scratch are significant.