Source: DeepTech On February 10th local time, Reuters reported that OpenAI is developing its first generation of internal AI chips. According to sources, OpenAI will complete the design of its first internal chip in the coming months and plans to send it to Taiwan Semiconductor for manufacturing. Taiwan Semiconductor will use 3nm technology to produce the OpenAI chip, which is expected to undergo testing by the end of 2025 and begin mass production in 2026. The chip is anticipated to feature "high bandwidth memory" and "extensive networking capabilities." At the same time, OpenAI expects to deploy its internal chips on a "limited scale," primarily for running AI models. By manufacturing its own chips, OpenAI will no longer need to rely on NVIDIA's chips to train and run AI models. NVIDIA's chips currently hold an 80% market share. Meanwhile, tech giants like Microsoft and Meta have also begun developing their own chips, making OpenAI's move in line with industry trends. OpenAI's chip design team is led by former Google TPU engineer Richard Ho, and in recent months, the team has grown from 20 to 40 members. Image | Richard Ho (Source: LinkedIn) Public information shows that Richard Ho is a "chip veteran" in Silicon Valley, holding a master's and Ph.D. from the University of Manchester in the UK and Stanford University in the US. His work experience at Google is a significant highlight of his career, where he stated on LinkedIn that he was responsible for "functional design verification across multiple Google chip projects (including TPU, VCU, and IPU)." During this time, he was responsible for implementing and verifying the application of machine learning (ML) technology to accelerate chip design. He initiated the use of ML for chip design, achieving cross-organizational and cross-functional collaboration, and developed tools for TPU production design, with related papers published in Nature. In fact, in 2024, foreign netizens discussed "whether OpenAI would launch its own chip first or NVIDIA would launch its own large model first." A year later, this question quickly received an answer today. OpenAI's self-developed chip has long been in the works. In 2024, according to the Financial Times, OpenAI CEO Sam Altman was negotiating with some of the wealthiest and most influential individuals in the Middle East, including the UAE's national security advisor, several national investment funds, and AI company chairman Sheikh Tahnoon bin Zayed al-Nahyan. At that time, media reports indicated that Altman was seeking funding for his ambitious new project, which involved developing and manufacturing chips for training and building AI models. In 2024, Altman also discussed potential collaboration with Taiwan Semiconductor for chip production. At that time, neither OpenAI nor Taiwan Semiconductor commented on the matter At the same time, Bloomberg reported in 2024 that Ultraman is discussing chip investments with Middle Eastern investors. With the rise of DeepSeek, Ultraman seems increasingly restless. Recently, he revealed new developments regarding OpenAI's large models during lectures in Germany and Japan, and on February 9 local time, he wrote a blog post stating his insights on AGI, followed by rumors of plans to manufacture chips. Chip manufacturing requires substantial financial resources, which aligns with OpenAI's recent efforts to seek new funding. As for the final effectiveness of the chips, it may become clear next year