OpenAI's ambitious plan to develop its own custom AI chip is reportedly encountering significant funding challenges. According to The Information, the initial phase of this project, in collaboration with chip designer Broadcom, carries an estimated cost of around $18 billion. Broadcom is reportedly unwilling to proceed with production unless Microsoft commits to purchasing approximately 40 percent of these custom chips. This arrangement effectively leverages Microsoft's creditworthiness to mitigate financial risks for Broadcom.
Internally, OpenAI manager Sachin Katti has characterized this reliance on Microsoft as "financially unattractive" and a potential long-term impediment. Despite these concerns, OpenAI is reportedly pressing forward with the project due to its strategic advantages, though Katti clarified that this specific funding structure would only apply to the first iteration of the chip.
The comprehensive chip initiative, codenamed "Nexus," is designed to support 10 gigawatts of data center capacity and could incur costs up to $180 billion solely for chip production. The inaugural chip, named "Jalapeno," is engineered to deliver superior efficiency for OpenAI's models compared to existing Nvidia hardware. However, its anticipated availability is not until 2027.
This situation marks the second instance where OpenAI's private company status has complicated a major financial transaction. Previously, SoftBank reportedly had to reduce a planned loan—originally intended to be backed by OpenAI shares—from $10 billion to $6 billion. This reduction was reportedly due to lenders' difficulties in reliably valuing a private entity like OpenAI.
