OpenAI Is Considering Making Its Own Chips Over Scarcity Of GPUs: Report

OpenAI, the organisation behind ChatGPT, is actively exploring the possibility of developing its artificial intelligence (AI) chips. According to a report by Reuters, OpenAI has even considered potential acquisition targets as part of its strategy.

No final choice has been reached, conversations inside OpenAI have focused on finding a solution to the organisation's dependence on pricey AI chips for its AI models, the report added. Custom AI chips, deeper collaborations with chip manufacturers like Nvidia, and supplier diversification outside of Nvidia are some of the options under discussion.

Sam Altman, the CEO of OpenAI, has ranked purchasing more AI chips as the company's top goal. Graphics processing units (GPUs), which are necessary for AI applications and are now dominated by Nvidia with over 80% of the global market share, are in short supply, and Altman has publicly expressed his concern about this.

READ: Bing Chat Now Gets OpenAI's New Text-To-Image DALL-E 3 Tool; Full Details

The lack of cutting-edge processors to power OpenAI's software and the high operating costs incurred by maintaining the required hardware are the two main issues driving the search for new AI chips.

Since 2020, OpenAI has been developing its generative AI technologies on a huge supercomputer with 10,000 Nvidia GPUs that were developed by Microsoft, one of its primary backers. However, the business is becoming increasingly concerned about the high running and scaling expenses of AI models like ChatGPT.

According to Bernstein analyst Stacy Rasgon, running ChatGPT is costly, with each query costing about 4 cents. A significant investment in GPUs would be necessary to scale up ChatGPT inquiries to a level equal to Google's search, initially costing roughly $48.1 billion and requiring about $16 billion in chips annually.

READ: ChatGPT Can Be A Good Psychotherapist; Here's What Research Suggests

OpenAI's investigation into creating its own AI chips is consistent with the desire of digital behemoths like Google and Amazon to create their own unique chips to cater to particular business requirements.

It's unclear, however, whether OpenAI would move forward with developing custom chips given that it would demand a sizable investment and level of skill. Success is not guaranteed, even if the company decides to invest resources in this project.