OpenAI is reportedly in talks with Broadcom to build the AI chips they need. They are now in the process of building their own AI chips to reduce the cost of buying hardware from parties like NVIDIA.
Previously in January, we reported OpenAI CEO Sam Altman had held meetings with companies such as Intel, Taiwan Semiconductor Manufacturing Company, Samsung Foundry and SK Hynix for collaboration on the development of this processing chip.
Lately OpenAI has started hiring former Google employees involved in the production of Tensor chips used in AI data centers. This week TSMC's CEO said the AI chip supply issue is expected to remain until the end of 2025. A year later than the original forecast of 2024. The extremely high demand for NVIDIA's AI chips that TSMC uses is causing the issue to linger longer.
The cost of developing more powerful and complex AI models is increasing. Only a few companies can afford the development costs with all AI companies yet to show a return on capital.