Apple may be shopping for AI chip companies; Here's why

Apple is reportedly on the lookout for companies specializing in AI chip design as part of its strategy to handle demanding artificial intelligence tasks while reducing reliance on Nvidia. Historically, Nvidia's GPUs have been instrumental in training and operating advanced AI models, serving major tech firms' AI infrastructure needs. While Apple has begun conducting some AI processing within its own data centers, it still heavily leans on Nvidia’s GPUs, particularly those hosted on Google Cloud, for the more intensive AI workloads. Rumors suggest that Apple is working on an internal AI server chip, known as Baltra, which was anticipated to debut this year, although its launch has faced delays. According to a report from The Information, Apple is actively considering several semiconductor startups for potential acquisition. The tech giant has a history of acquiring smaller firms, including the recent purchase of Israeli AI startup Q.ai for nearly $2 billion, and is reportedly in discussions with PrismML as well. Apple is motivated by performance limitations within its current AI server infrastructure. Although it utilizes its in-house M2 Ultra chips, they fall short of the capabilities offered by specialized AI accelerators. Consequently, Apple relies on external infrastructures for its AI operations, with its Gemini-powered Siri handling complex AI requests using Nvidia GPUs through Google Cloud. In addition, Apple is on the hunt for AI startups focused on model compression technology, which helps advanced AI models run more efficiently on devices like the iPhone. Bloomberg reports that Apple is also developing an AI server chip based on the M5 Ultra, with a more powerful M7 Ultra chip expected to significantly enhance AI performance. This new generation of chips is projected to support up to 1.5TB of memory, although a server variant may not be available until 2029.