OpenAI’s Hardware Ambition: Crafting the Future of AI Chips?

OpenAI, the AI research hub, is considering crafting its own AI-specific chips, given the rising demand and a shortage of the required hardware. Ever since the release of its AI chatbot, the clamor for these special chips has skyrocketed.

To tackle the shortage, OpenAI is thinking about a few solutions. They’re contemplating buying a chip-making company or collaborating more tightly with various chip manufacturers. This includes not just the industry leader, Nvidia, but other players too, insiders have shared with Reuters.

OpenAI’s head, Sam Altman, has previously voiced his worries about the limited availability of graphics chips (GPUs) and their rising costs. Securing more of these AI chips is now high on his list, as per those in the know.

Even if OpenAI decides to design its own chip – a step they haven’t taken yet – they’d probably still rely on companies like Nvidia for a while, given Nvidia dominates with an 80% share in the AI chip market.

The tech driving OpenAI’s AI system is anchored to a supercomputer packed with 10,000 Nvidia GPUs, courtesy of one of its main supporters, Microsoft. The cost of running their AI assistant isn’t peanuts – each interaction sets them back by about $0.04, as mentioned in a report.

If OpenAI dives into making their chips, they’d join the elite club of tech behemoths like Google and Amazon, who’ve already ventured into custom chip design, recognizing its pivotal role in their operations.

Meta, the company behind Facebook, also dabbled in processor development for their AI assistant. They faced hitches, discarded a few early designs, but are now forging ahead with a refreshed model.

Lastly, Microsoft, a staunch supporter of OpenAI’s chatbot initiative, is also experimenting with its custom AI chip, which is currently under testing by the creators of ChatGPT, as highlighted in a report by tech news outlet, The Information.

source