OpenAI lures Google's top AI researchers with multimillion-dollar offers

OpenAI is wooing top artificial intelligence employees at Google with multimillion-dollar offers as the startup's employee stock sale nears completion, potentially tripling its valuation to more than $80 billion.

OpenAI's recruiters are targeting senior AI researchers at Google, offering annual compensation packages worth $5 million to $10 million, mostly in stock, The Information reported, citing insider sources.

OpenAI is reportedly specifically targeting Google employees who are developing Google's Gemini family of AI models. The startup has also given raises to some junior employees due to market conditions. Google has also recently poached some high-profile researchers from OpenAI, in part by offering higher salaries.

OpenAI and Google are locked in a fierce battle for talent, with both companies claiming to have more computing resources than the other. OpenAI CO Sam Altman said internally that Google will have a computing advantage until sometime next year when Microsoft makes more AI chips available, according to The Information.

Microsoft could bolster hardware support for OpenAI with new AI chips

Microsoft is reportedly developing its own AI chip, code-named "Athena," to compete with Nvidia's H100 GPU and join Google and Amazon in the AI hardware space. The Athena chip, which is expected to be unveiled at the Ignite developer conference in November, could be offered as a cloud solution for AI developers.

As major AI companies seek alternatives to Nvidia's dominance in the AI chip market, Microsoft is also working with Advanced Micro Devices (AMD) on its next AI chip, the MI300X, which is expected to be released in the fourth quarter.

Google has been developing and using its own AI chip, called the Tensor Processing Unit (TPU), since at least 2015. TPUs are designed for high-volume, low-precision computing, with more input/output operations per joule compared to graphics processing units (GPUs).

Google has used TPUs in various applications, including the AlphaGo vs. Lee Sedol Go game, Google Street View text processing, and Google Photos, as well as to train its current AI models. Since 2018, TPUs have been available to third parties through Google's Cloud TPU service.