Previously confirmed with NVIDIA, AMD and OracleCooperationLater, OpenAIAnnounceCollaborating with Broadcom to build its own customized AI acceleration chips and systems, deployment is expected to begin in the second half of 2026 and all planned projects will be completed by 2029.
This news further confirms the previous market claim that OpenAI and Broadcom are collaborating.rumor, will use the customized AI acceleration chip built with the help of Broadcom for OpenAI's infrastructure and its partners' data centers.
OpenAI explained that the collaboration involved chip orders worth "billions of dollars" and a computing power of 10 gigawatts. At the same time, if we refer to Broadcom CEO Hock Tan's previous statement that he had received an order of up to $100 billion from an unnamed customer, it is obvious that this is a collaboration project with OpenAI.
Prior to this, OpenAI had already signed partnerships with NVIDIA and AMD. The NVIDIA partnership involved a $1000 billion investment and the provision of 10 gigawatts of AI infrastructure, while the AMD partnership involved 6 gigawatts of computing power and a multi-billion dollar investment in exchange for approximately a 10% stake in AMD.
In addition, OpenAI signed a cooperation agreement with Oracle in July to obtain 4.5 gigawatts of data center computing resources through the Stargate Project.
OpenAI CEO Sam Altman has revealed to employees that he hopes to build a total of 250 gigawatts of computing power within the next eight years, a significant increase from the 2 gigawatts of computing power projected for the end of this year. However, based on current estimates, building such a massive computing capacity would cost approximately $10 trillion, so Altman has stated that achieving this goal requires new funding.
However, even with the investment and cooperation from industry players such as NVIDIA and Microsoft, it is still difficult to support such a huge capital demand. At the same time, considering that OpenAI's expected revenue this year is only $130 billion, it is probably difficult to support such a huge expenditure. Therefore, OpenAI may have to find a way toMore financing optionsGet funding.
In addition, since OpenAI has begun building its own customized AI acceleration chips and systems, it is clear that it also plans to reduce its dependence on companies such as NVIDIA, and use them to meet its own artificial intelligence technology computing needs. In the future, it may even be able to provide such customized computing power systems to external users, thereby creating greater revenue opportunities.
However, whether the current development of AI technology applications can truly generate greater market revenue still seems to depend on the actual application development model. At the same time, OpenAI must also face competition from industry players including Google and Meta. Therefore, how OpenAI will generate greater revenue from its AI technology applications and support its massive computing power construction costs may become a greater challenge.
