OpenAI May Be Developing Its Own AI Chip

OpenAI is pondering its next big move in hardware as it eyes the potential development or acquisition of its own artificial intelligence (AI) chips.

Amid growing concerns over chip scarcity and the escalating costs of running AI systems, such a decision would not only echo the steps taken by some of the industry’s largest players but also mark a strategic pivot in the company’s trajectory.

The global chip market, notably in AI, is a challenging landscape. Nvidia, which controls over 80% of the global market for AI-optimized chips, has left many companies, including OpenAI, at the mercy of supplier limitations and soaring costs. 

With each ChatGPT query ringing up at approximately 4 cents, scaling to even a fraction of Google search levels could rack up billions in expenses for OpenAI.

OpenAI’s CEO, Sam Altman, has not been silent about these challenges. He stressed the importance of securing a consistent and affordable source of graphics processing units (GPUs) for the company. 

Microsoft, which serves as OpenAI’s largest investor, constructed a colossal supercomputer that OpenAI has utilized since 2020 to run its AI applications, using 10,000 Nvidia GPUs. According to The Information, Microsoft is also reportedly developing its own custom AI chip that OpenAI is testing.

While building a proprietary AI chip could be a game-changer for OpenAI, it wouldn’t be without its hurdles. Even with substantial investment, potentially reaching hundreds of millions annually, a successful foray into chip development is not guaranteed. Moreover, the pathway to custom chip production could stretch over several years, during which OpenAI would continue its dependence on industry giants like Nvidia and AMD.

Acquiring an established chip company might streamline the process, drawing parallels with Amazon’s strategic acquisition of Annapurna Labs in 2015. Though specifics remain under wraps, sources have hinted that OpenAI has been down this route, even conducting due diligence on a potential target.

As the race for specialized AI chips intensifies, triggered by-products like ChatGPT, OpenAI’s eventual choice — whether developing, acquiring, or even partnering — will be pivotal in determining the company’s ability to harness AI’s unyielding power. 

One thing is certain: in an industry where innovation is currency, the coming years will be anything but predictable.

Editor’s note: This article was written by an nft now staff member in collaboration with OpenAI’s GPT-4.