Microsoft is ramping up its efforts to create its own artificial intelligence (AI) chips, reportedly codenamed "Athena", in-house since 2019 to both reduce its reliance on Nvidia GPUs and manage the costs of powering large-language models. This news comes after the already bustling AI sector has seen Google, Amazon, and Meta developing various machine learning chips to compete with their own ventures.
These high-priced chips are becoming increasingly harder to come by, such as the Nvidia H100 reportedly going for up to $40,000 on reseller services like eBay. With the production and unveiling of its own local chips, Microsoft hopes to control the cost of powering large-language models for its OpenAI projects, and also for their internal machine learning staff.
On April 17, news broke out of tech mogul and world’s richest man Elon Musk announcing the launch of TruthGPT, a “truth-seeking” large language model designed to tackle the heightening allegations of ChatGPT’s left-wing bias. This propelled the emergence of Amazon’s Bedrock AI platform, as the company recently debuted its first self-developed models.
The competition between companies in the AI field is steadily emerging, and the capability to manage costs of large-language models and OpenAI projects is an integral factor when it comes to successful development. This is why Microsoft’s initiative in developing their own AI chips may just be the competitive edge the company need, should the chips be successful.
Details remain scarce for now, as Microsoft hasn’t officially commented. But, already being tested out by Microsoft’s own internal machine learning staff and OpenAI developers, the result of their research and launching of their own proprietary chip could mean a re-thinking of the infrastructure and design to take the company from GPT-1 to GPT-4.
Only time will tell if the "Athena" chips will give Microsoft the boost they need in the AI race, and how the emerging ventures of Google, Amazon, Meta, and now Musk's TruthGPT will work to keep up.
Noah Davis
- 2023-04-18
Microsoft's AI Chip to Reduce Reliance on Nvidia GPUs
Microsoft is reportedly developing its own Artificial Intelligence (AI) chips that would reduce their reliance on Nvidia GPUs and manage the costs of large-language models for their OpenAI projects.
