Siri’s Upgrade Could Depend on Google’s Infrastructure

Apple is preparing to run its next-generation Siri on Google’s servers. Facing cloud capacity strain ahead of its Gemini-powered upgrade, the company has asked Google to host infrastructure for the more advanced, conversational version of its assistant.

According to The Information, the discussions center on deploying privacy-compliant servers within Google data centers to support the rollout, marking a potential shift from model integration to greater reliance on infrastructure.

Scaling up before the spotlight

Apple’s cloud needs are starting to outgrow what it has on hand, The Information reports.

When Siri gets a complex request, it can hand the job off to Private Cloud Compute (PCC), Apple’s own server setup built on Apple silicon. The catch is that those chips were designed for consumer devices, not the heavy AI workloads that come with running large models like Gemini.

PCC is averaging about 10% usage today, and some hardware earmarked for the system still hasn’t been installed. That cushion may not last once the upgraded Siri lands and usage ramps, especially with more conversational back-and-forth and multi-step tasks expected to push far more queries into the cloud.

A door once kept closed

For Apple, the Google angle is notable because it wasn’t always on the table. Apple reportedly kept its AI teams away from Google’s cloud technologies for years over privacy concerns, and software chief Craig Federighi repeatedly vetoed Google Cloud for AI computing needs.

That posture began to change after Google updated its security systems in 2023 in ways that satisfied Apple’s requirements. Since then, Apple has started adopting Google’s cloud infrastructure for parts of its AI work, setting the stage for the kind of server-hosting request now under discussion.

The Information’s coverage also flags a practical pain point on Apple’s side: PCC is slower to update than other server setups, adding friction just as Siri’s next version nears.

The agreement behind the upgrade

In January, Apple and Google put their Gemini tie-up on the record, saying they’d entered a multi-year collaboration that would base Apple’s next-generation foundation models on Google’s Gemini models and cloud technology. The joint statement pointed to “a more personalized Siri” as one of the outcomes Apple expects to ship.

The deal isn’t just technical. One estimate put the annual cost of bringing Gemini into Siri at around $1 billion, with guardrails designed to keep Siri queries walled off from a user’s personal information. Apple is also expected to use Gemini models customized to its requirements.

Now Apple is also discussing where that Siri compute should run, a sign that the collaboration could extend from software into the servers that support it.

Google is drawing its robotics efforts closer, folding Intrinsic into its core AI units to connect advanced models with real-world automation.