Unleashing the Power of Siri’s Cloud Brain: The Nvidia Connection

Apple made waves two years ago at its Worldwide Developers Conference (WWDC) with the announcement of “AI for the Rest of Us.” Among the promises was a more intelligent and personalized Siri, powered by a unique cloud-based infrastructure built on privacy. While speculation suggested Apple would use its own silicon chips to power this infrastructure, recent developments indicate a shift in strategy.

Reports hinted at Apple building servers with M2 Ultra chips for its Private Cloud Compute (PCC) infrastructure. However, it seems Apple has recognized the limitations of its silicon chips for cloud-based AI compared to other established options. Following the delay in delivering on the Siri promise, Apple underwent executive changes and sought outside help, leading to a partnership with Google for its Gemini models.

Apple’s decision to lean on Google’s foundation for AI construction raised questions about where these models would run. It was revealed that the new PCC infrastructure would not be powered by Apple silicon but by Nvidia Blackwell B200 chips, the same chips used in Google’s powerful Gemini servers.

While Apple remains committed to on-device processing, some user queries to Siri will run on Google Cloud using a licensed version of the Gemini model. To ensure user privacy, Apple tested and approved Nvidia’s Confidential Computing technology, encrypting data and AI models during processing.

Despite the shift away from Apple-designed servers, Apple plans to retain the Private Cloud Compute branding. This move is seen as a necessary compromise, prompted by the upcoming WWDC where Apple will unveil the new Siri and other Apple Intelligence features.

It is important to note that the information provided in this article is speculative and has not been confirmed by Apple. As with all tech rumors, it is advisable to approach them with caution.