Chatbot Siri, powered by Gemini, to use Nvidia's powerful Blackwell B200 GPU

Siri could get a new brain from the Wizard as soon as this coming Monday.

Image shows page from Siri's set up process.
Siri 2.0 to use powerful Nvidia chip. | Image by PhoneArena
There is a reason why Nvidia has a market capitalization over $5 trillion. The company's GPUs, using parallel processing, use thousands of cores to take a complex task, break it down into thousands of smaller, simpler tasks, and solve them all at the same time. Compare that to CPUs, which run sequentially and use a smaller number of faster cores that work on one task, finish it, and then work on the next one. 

Apple WWDC 2026 kicks off this Monday


Apple iPhone users can feel the excitement. On Monday, Apple's annual WWDC developer conference will kick off and Siri 2.0 is expected to be discussed at length. And then, once the keynote comes to an end, if tradition holds, Apple will release iOS 27 beta 1 for developers.

Will you use AI more after Siri upgrade?
0 Votes

The hope is that the beta will include all of the changes to Siri that we've been hearing about since WWDC 2024. According to a report from The Information that was run a few days ago, 
some queries that you ask chatbot Siri will run on the custom 1.2 trillion parameter Gemini model that Apple is reportedly paying Google $1 billion a year to use. 

The huge custom Gemini model will replace the current 3 billion parameter generative AI model that is used today for Siri to help with summaries, basic text rewriting, and improved language understanding. The user can send more complex questions to ChatGPT for a response.

Apple could be using the Nvidia Blackwell B200 GPU


According to The Information, when Siri uses the new custom Gemini model, it will be using a privacy technology from Nvidia. That led some people familiar with the matter to come to the conclusion that Apple will be using Nvidia's AI chips when answering queries posed to chatbot Siri in Google Cloud.

Recommended For You

The report goes on to say that Apple will likely be using Google's holdings of Nvidia’s Blackwell B200 data center chips that will play into Apple's demands for privacy. Using Nvidia's confidential-compute feature, data is encrypted as it is processed on the chip. The Blackwell B200 is made up of two GPU dies built on TSMC's 4NP, the second generation of the foundry's 4nm process technology.  

Nvidia's GPU could sport a whopping 208 billion transistors compared to 25 billion to 30 billion for the A19 Pro


Part of the speed of the Blackwell B200 comes from the blistering-fast 10 Terabytes-per-second (TB/s) interconnect that puts the two GPU dies together. For the software running over the B200 chip, the two dies feel like one GPU and the combined dies sport a combined 208 billion transistors. To put that in perspective, the A19 Pro application processor (AP) designed by Apple for the iPhone 17 Pro, iPhone 17 Pro Max is equipped with 25 billion to 30 billion transistors.

The iPhone Air has a "binned" A19 Pro AP with only a 5-core GPU (compared to the normal 6-core GPU design).

If all of this sounds unusual for Apple, that is because it is. The company usually loves to have control over every aspect of a product. Keep in mind that the only iPhone models currently in use that will be receiving the new update that makes Siri smarter will be the following:

"Siri, what time do I need to leave to pick up my mom from the airport on time?"


It will be interesting to see not only chatbot Siri do her thing and become a genius, but also to see Personal Siri go through your native iOS apps to find out what time you should leave to pick up your mother from the airport. Siri will go through apps like Messages, Mail, Photos, Notes, Calendar, Apple Wallet and others to find info about the flight your mom is going to be on.

Once Siri has this information, it can go to a third-party flight tracker and get up-to-date arrival time data. Siri then goes to Apple Maps and, working out your trip to the airport, figures out what time you should leave your current location in order to pick up your mom on time.
Get Visible as low as $20/mo for 1 year. Limited time offer with code: FRESHSTART
$20 /mo
$25
$5 off (20%)
Offer Ends 6.1.2026 at 11.59pm ET. New members get $5/mo off the $25/mg Visible plan, $35/mo Visible+ plan, or $45/mo Visible+ Pro plan for the first 12 months. Promo code FRESHSTART required at checkout.
Buy at Visible
Recommended For You
COMMENTS (0)
Latest Discussions
by readdriver • 2
by ECPirate37 • 2
by menooch18 • 2