According to reliable sources, ChatGPT developer OpenAI is promoting the development plan of its first self-developed artificial intelligence chip, which is expected to be completed and delivered to TSMC for chipping in the next few months. The project is planned to adopt TSMC's advanced 3nm process and is equipped with high bandwidth memory. The overall architecture is similar to that of Nvidia products.
This strategic move began in October 2023 to solve the problem of shortage of AI chip supply and high costs. Sources pointed out that OpenAI sees the training-focused processor as an important tool to enhance negotiation capabilities with other chip suppliers. The company's engineer team plans to continuously optimize processor performance in subsequent iterations.
If progress goes well, OpenAI is expected to achieve its mass production target in 2026. However, the road to chip development is not a smooth road: each slitting can cost tens of millions of dollars, the standard production cycle takes about six months, and the first slitting may face the risk of failure, requiring additional diagnostics and reslitting.
It is worth noting that although technology giants such as Microsoft and Meta have invested heavily in the field of self-developed chips, they have failed to make significant breakthroughs. At the same time, the market structure is undergoing subtle changes: DeepSeek, a Chinese AI startup, has lowered the hardware threshold through algorithm optimization, triggering discussions on the future direction of AI computing demand.
Against this background, the investment enthusiasm of technology giants remains undiminished. Microsoft, Amazon, Google and Meta have announced that it will invest a total of $320 billion in AI technology and data center construction in 2025, a further increase from last year's record spending.
This series of developments indicates that the competitive landscape in the field of AI chips may soon usher in major changes.