ChatGPT developer OpenAI will work with U.S. semiconductor company Broadcom and Taiwan’s TSMC, the world’s largest foundry (semiconductor consignment production) company, to develop its own artificial intelligence (AI) chips, Reuters reported on the 29th (local time).
OpenAI does not rely solely on AI chip leader Nvidia, but develops its own chips with Broadcom and entrusts production to TSMC.
Earlier in July, IT media Deformation reported that OpenAI is in discussion with Broadcom to develop its own AI chips.
“OpenAI has been looking at various options to diversify its chip supply and reduce costs,” the source said. “We also plan to use AMD chips in addition to Nvidia chips to meet surging infrastructure demand.”
OpenAI is known to be a major customer of Nvidia, and it intends to diversify its chip supply by using AMD chips as well as developing its own AI chips.
As a result, it is believed that it has withdrawn its original plan to build a global network to produce AI chips.
The sources said OpenAI has for now abandoned its foundry plan with the cost and time needed to build the network, with plans to focus on in-house chip design efforts instead.
OpenAI CEO Sam Altman has been working with global semiconductor companies to establish a network for chip manufacturing and considering ways to raise capital for this.
He visited Korea in June last year and January this year as he began building an ‘AI semiconductor alliance’ for his own AI semiconductor development.
In particular, during his visit in January, he toured semiconductor production lines at Samsung Electronics’ Pyeongtaek plant and met with top executives from Samsung and SK, and said he hoped to cooperate with the two companies in making AI chips.
“This is the first time OpenAI’s strategy has been disclosed in detail,” Reuters said. “OpenAI is using a strategy to secure chip supply and manage costs by mixing industry partnerships and internal and external methods like Amazon, Meta, Google, and Microsoft.”
EJ SONG
US ASIA JOURNAL
