Apple's New Siri Will Run on Google's Nvidia Fleet
The overhauled assistant launching this month relies on Google's Nvidia Blackwell B200 data centre chips, with user data encrypted through hardware confidential compute.
Apple's overhauled Siri will run on Google's fleet of Nvidia chips — specifically Blackwell B200 data centre processors — with user data encrypted using Nvidia's hardware-based confidential compute feature.
Three companies in one sentence
The arrangement is worth stating plainly: Apple's assistant, running on Google's infrastructure, using Nvidia's silicon and Nvidia's security hardware.
For a company whose identity has rested on controlling its own stack — designing its silicon, writing its software, running its services — this is a substantial departure.
Why Apple made it
The economics of frontier AI infrastructure are unforgiving. Serving a capable assistant to Apple's installed base requires data centre capacity at a scale that takes years and enormous capital to build.
Apple's alternatives were to build that capacity and ship late, or rent it and ship now. It rented.
The privacy engineering
The confidential compute detail is the part that makes the arrangement possible at all. Hardware-based confidential computing encrypts data while it is being processed, not merely in storage and transit, with the encryption enforced by the processor itself.
In principle that means the infrastructure operator cannot read what passes through its own machines. It is how Apple can run user queries on a competitor's servers without abandoning its privacy positioning.
The guarantee is only as strong as the hardware implementation, and confidential computing systems have had vulnerabilities disclosed before. This is a considered trade-off, not an absence of risk.
What it says about the market
Even a company with Apple's resources has concluded that building its own frontier AI infrastructure is not the right use of capital. The compute layer has consolidated around a small number of providers, and nearly everyone else is a customer.
Elsewhere in silicon
Nvidia and MediaTek announced a deepened collaboration on next-generation AI computing platforms spanning AI infrastructure, edge devices and cloud ecosystems — pairing Nvidia's data centre position with MediaTek's low-power mobile and edge expertise.
Apple's new chief executive has separately teased a "huge launch". Given where the assistant now runs, the interesting question is how much of it Apple will own.