Self-developed chips become a shining example! OpenAI is reportedly collaborating with Broadcom to create AI chips, with TSMC's production capacity scheduled for 2026.

To achieve chip supply diversification, the developer of the ChatGPT chatbot, OpenAI, is reportedly cooperating with Broadcom and TSMC to develop its first self-developed chip. Meanwhile, OpenAI is also increasing its use of AMD chips, in conjunction with Nvidia chips, to meet its surging infrastructure needs. According to Reuters, OpenAI has explored a series of solutions to achieve chip supply diversification and cost reduction, including the consideration of completely self-developed chips and fundraising to build expensive wafer fabs. However, due to the time-consuming and enormous cost of building the infrastructure, OpenAI has temporarily suspended its wafer fab plans and shifted its focus to internal chip design work. OpenAI is seeking to ensure chip supply and manage costs by leveraging industry partnerships, similar to the practices of its competitors such as Amazon, Meta, Google, and Microsoft. As one of the largest chip buyers, OpenAI has chosen to obtain supply from multiple chip manufacturers and develop self-developed chips, which may have a broader impact on the tech industry. OpenAI has been working with Broadcom to create its first AI chip focused on reasoning, and has secured TSMC's manufacturing capacity. It plans to produce its first self-developed chip in 2026, with the timetable subject to change. OpenAI is still evaluating whether to further develop its self-developed chips or pursue acquisition, and may engage with more partners. OpenAI has assembled a chip team of about 20 people, led by top engineers who previously established the Tensor Processing Unit (TPU) at Google, including Thomas Norrie and Richard Ho. Currently, Nvidia's GPU market share exceeds 80%. However, factors such as chip shortages and rising costs have led major customers like Microsoft, Meta, TSL, and OpenAI to explore self-developed chips or seek external alternatives. OpenAI is less focused on self-developed GPUs and instead focuses on creating specialized chips to execute operating software and respond to user queries, i.e., the inference process. It is worth noting that OpenAI is also planning to use AMD chips through Microsoft's Azure service, indicating that AMD's new MI300X chip is attempting to carve out a share in Nvidia's dominant market. AMD expects its AI chip sales to reach $4.5 billion in 2024, with the chip set to be released in the fourth quarter of 2023. The cost of training AI models and operating services like ChatGPT is extremely high. OpenAI estimates a loss of $5 billion and revenue of $3.7 billion for this year. Related reports: SoftBank rumored to invest $500 million in OpenAI, but concerns have sparked uncertainty Displeased with OpenAI's plan to transform into a for-profit company? CTO and two important researchers simultaneously resign The strongest AI tutor has arrived? OpenAI rumored to release ChatGPT advanced voice mode on 9/24 Self-developed chips are becoming a trend! OpenAI reportedly partners with Broadcom to create AI chips, with TSMC production capacity scheduled for 2026. This article was originally published on BlockTempo, the most influential blockchain news media in the dynamic block.