Apple is breaking its own rules. The company plans to route some user queries for its new Gemini-powered Siri directly to Google Cloud. To pull this off, the tech giant is tapping into Google’s fleet of Nvidia Blackwell B200 data center chips, which are specifically built to process trillion-parameter AI models.
The news surfaced in a June 3 report from The Information. Apple built its entire modern brand on keeping data locked down inside its own ecosystem. Now, the sheer processing power required for next-generation artificial intelligence is forcing a radical change in how the company handles cloud computing.
Why Apple is looking beyond its own servers
Two years ago, Apple introduced Private Cloud Compute, a Mac-series chip-based server architecture built to keep everything securely in-house. But the hardware simply could not keep up with the demands of modern AI models. Apple’s proprietary server system was too slow to run Gemini, forcing the shift to Nvidia Blackwell B200 chips on Google Cloud, according to Digital Today.
Apple needed a way to use external servers without breaking its strict consumer privacy promises. Nvidia provided the exact fix. Apple officially approved the use of Nvidia’s “confidential compute” feature to bridge the gap. This hardware-level security protocol encrypts sensitive user data while it is actively being processed by the GPUs, keeping information blind even to the server hosts.
The timing is highly specific. Apple is scheduled to officially unveil the revamped Siri at WWDC 2026, which runs from June 8 to June 12. The broader context of Apple’s Gemini partnership for Siri and the anticipated WWDC 2026 AI announcements shows a company completely rethinking its hardware strategy, CNET recently reported. Following the June developer conference, the next-generation Siri is expected to launch to the public in September.
What the Google Cloud alliance means for the AI hardware race
By relying on Nvidia and Google, Apple is validating the reality of modern cloud computing. For decades, Apple maintained a strict walled-garden approach, controlling every single hardware and software ingredient that touched user data. Handing off core processing tasks to external Nvidia hardware changes that formula entirely.
It also signals that even the most well-funded tech giants cannot build competitive internal infrastructure fast enough to match Nvidia’s current B200 chip output. Apple heavily pushed its own silicon for servers over the last two years. If the company is now willing to outsource its most highly anticipated AI feature to Google Cloud, the barrier to entry for proprietary AI data centers is officially too high for almost anyone to conquer alone.