OpenAI wanted to protect itself against NVIDIA. It didn’t work out
NVIDIA is the undisputed leader in the AI hardware market. However, not all companies like this monopoly, and OpenAI had a plan to change that.
More text below the video
According to a TechCrunch report, OpenAI was considering purchasing the company Cerebrasa startup specializing in equipment for working with artificial intelligence. This would ensure the possibility of producing our own AI accelerators and reduce dependence on NVIDIA and their solutions.
OpenAI currently cooperates with Broadcom
This information comes from new court documents from the lawsuit Elon Musk against OpenAI. They reveal preliminary discussions between him, company executives and Tesla about a potential acquisition of Cerebras. However, this plan ultimately did not come to fruition, but the specific reason was not revealed.
He was the initiator Ilya Sutskeverco-founder of OpenAI, who in 2017 suggested that Tesla could be the acquirer of Cerebras. However, it noted that this purchase could create a conflict of interest because Tesla’s obligation to maximize shareholder returns was not consistent with OpenAI’s nonprofit mission.
Cerebras designs AI chips the size of an entire silicon wafer that the company claims outperforms NVIDIA hardware. The company is currently preparing to go public. Collected $715 million in financingand the valuation is expected to be approximately USD 8 billion. However, the largest customer raises doubts, G42 from Abu Dhabi, which is closely related to the Chinese company Huawei.
After abandoning the described plan, OpenAI has revised its strategy. Initially, they wanted to build a network of factories operated by entities such as TSMC and dedicated exclusively to the production of AI processors. Due to financial unprofitability, this idea was abandoned.
Currently, OpenAI works with Broadcom on designing its own AI accelerators to be produced by TSMC. The first chips may be ready as early as 2026which will allow the company to increase cost efficiency in training AI models, in line with its long-term strategic goals.
