ChatGPT-owner OpenAI is exploring making its own AI chips -sources
SAN FRANCISCO/WASHINGTON听–听OpenAI, the company behind听颁丑补迟骋笔罢,听is听别虫辫濒辞谤颈苍驳听尘补办颈苍驳听颈迟蝉听辞飞苍听artificial intelligence听肠丑颈辫蝉听and has gone as far as evaluating a potential acquisition target, according to people familiar with the company鈥檚 plans.
The company has not yet decided to move ahead, according to recent internal discussions described to Reuters. However, since at least last year it discussed various options to solve the shortage of expensive听础滨听肠丑颈辫蝉听that听翱辫别苍础滨听relies on, according to people familiar with the matter.
These options have included building聽颈迟蝉听辞飞苍听础滨听chip, working more closely with other chipmakers including Nvidia and also diversifying聽颈迟蝉听suppliers beyond Nvidia.
Open础滨听declined to comment.
CEO Sam Altman has made the acquisition of more听础滨听肠丑颈辫蝉听a top priority for the company.聽聽of graphics processing units, a market dominated by Nvidia, which controls more than 80% of the global market for the听肠丑颈辫蝉听best suited to run听础滨听applications.
The effort to get more听肠丑颈辫蝉听is tied to two major concerns Altman has identified: a shortage of the advanced processors that power听翱辫别苍础滨‘s software and the 鈥溾 costs associated with running the hardware necessary to power聽颈迟蝉听efforts and products.
Since 2020,听翱辫别苍础滨听has developed聽颈迟蝉听generative artificial intelligence technologies on a massive supercomputer constructed by Microsoft, one of聽颈迟蝉听largest backers, that uses 10,000 of Nvidia’s graphics processing un颈迟蝉听(GPUs).
Running听颁丑补迟骋笔罢聽is very expensive for the company. Each query costs roughly 4 cents, according to an analysis from Bernstein analyst Stacy Rasgon. If听颁丑补迟骋笔罢聽queries grow to a tenth the scale of Google search, it would require roughly $48.1 billion worth of GPUs initially and about $16 billion worth of听肠丑颈辫蝉听a year to keep operational.
CUSTOM听颁贬滨笔厂听ERA
An effort to develop聽颈迟蝉听辞飞苍听础滨听肠丑颈辫蝉听would put听翱辫别苍础滨听among a small group of large tech players such as Alphabet鈥檚 Google and Amazon.com聽that have sought to take control over designing the听肠丑颈辫蝉听that are fundamental to their businesses.
It is not clear whether听翱辫别苍础滨听will move ahead with a plan to build a custom chip. Doing so would be a major strategic initiative and a heavy investment that could amount to hundreds of millions of dollars a year in costs, according to industry veterans. Even if听翱辫别苍础滨听committed re蝉辞耻谤肠别蝉听to the task it would not guarantee success.
An acquisition of a chip company could speed the process of building听翱辫别苍础滨鈥檚聽辞飞苍听chip – as it did for Amazon.com and聽颈迟蝉听.
Open础滨听had considered the path to the point where it performed due diligence on a potential acquisition target, according to one of the people familiar with聽颈迟蝉听plans.
The identity of the company听翱辫别苍础滨听examined purchasing could not be learned.
Even if听翱辫别苍础滨听goes ahead with plans for a custom chip – including an acquisition – the effort is likely to take several years leaving the company dependent on commercial providers like Nvidia and Advanced Micro Devices聽in the meantime.
Some big tech companies have been building their聽辞飞苍听processors for years with limited results. Meta’s聽custom chip effort has been beset with issues, leading the company to聽, according to a Reuters report. The Facebook听辞飞苍别谤听is now working on a newer chip that will span all types of听础滨听work.
OpenAI‘s main backer, Microsoft, is also developing a custom听础滨听chip that听翱辫别苍础滨听is testing, The Information has reported. The plans could signal further distancing between the two companies.
Demand for specialized听础滨听肠丑颈辫蝉听has soared since the launch of听颁丑补迟骋笔罢聽last year. Specific听肠丑颈辫蝉, or听础滨听accelerators, are necessary to train and run the latest generative听础滨听technology. Nvidia is one of the few chipmakers that produces useful听础滨听肠丑颈辫蝉听and dominates the market. – Reuters
