Openai is planning to reduce its dependence on Nvidia by creating its own AI chip. The company plans to complete the design of its first custom chip in the coming months and send it to Taiwan Semiconductor Manufacturing Co (TSMC) for production.
Openai and TSMC declined to comment.
“The ChatGpt manufacturer will complete the design of its first in-house chip in the coming months and send it for production at Taiwan Semiconductor Manufacturing Co,” Reuters reported, citing the source.
In this news, Openai has joined the growth list of the high-tech giants that are planning to build their own in-house AI chips, including Meta, which is developing AI brain chips to power open source models. Masu.
Openai is released from Nvidia
This move will ensure Openai is on track to begin mass production at TSMC by 2026. The initial process known as “taping” usually costs tens of millions of dollars and takes about six months to produce the finished chip. If a problem occurs, Openai will need to diagnose the process and repeat it, adding time and money.
Openai takes steps to reduce Nvidia’s dependence
The chip is not just a technology project, but also a strategic move to give Openai more negotiation power with other chip suppliers. After the first chip, Openai plans to develop a more sophisticated processor with a wider range of features.
With the initial tape out smoothly, Openai will begin mass production of AI chips later this year, offering an alternative to Nvidia’s dominant hardware. This will be an important milestone. It’s an important milestone, especially considering how long it usually takes for other ship manufacturers to reach this point.
“If the first tape goes smoothly, ChatGpt manufacturers will be able to generate a large number of first in-house AI chips and test alternatives to Nvidia’s chips later this year. Openai sends designs to TSMC this year The plan shows that the startup has made rapid progress in its first design.
Tech giants like Microsoft and Meta have struggled for years to create reliable chips. Meanwhile, market reforms sparked by Chinese AI startup DeepSeek have raised questions about whether a strong AI model would require fewer chips in the future.
Openai’s chip development team and costs
Richard Ho leads the Chip Design team at Open. He joined the company from Google over a year ago. There it was part of a custom AI chip program. Ho’s team has recently doubled to 40 people and is working with Broadcom on the project. Reuters first reported on the partnership last year.
Openai’s team is smaller than chipping efforts at companies like Google and Amazon, but costs are still higher. Designing one chip can cost up to $500 million, and building the software and support system you need can double that amount.
Companies like Openai, Google and Meta have shown that as AI models become smarter, more chips need to be linked in large data centers. This demand solidifies Nvidia’s retention in a market that manages around 80%.
Meta is planning to spend $60 billion on AI infrastructure next year, but Microsoft is planning to invest $80 billion in 2025. Openai is also involved in the $500 billion Stargate infrastructure program announced last month by US President Donald Trump.
However, rising costs and reliance on NVIDIA have led companies like Microsoft, Meta and Openai to look for internal solutions and alternative suppliers.
Openai’s first AI chip handles both training and AI models, but is initially used at a limited scale, primarily to run models within Openai’s infrastructure.
To build a chip program that is as extensive as Google and Amazon’s, OpenAI will need to significantly expand its engineering team and hire hundreds of new professionals.
TSMC uses advanced 3-nanometer process technology to manufacture OpenAI AI chips. The chip features a similar systolic array architecture and high-bandwidth memory (HBM) as Nvidia’s chips, along with advanced networking capabilities.
Source link
