3 hours ago
1
Apple's choice of Google's Gemini to power the next version of Siri was just as expected as peanut butter mixing deliciously with chocolate.
Early this week, Apple and Google put out a joint statement saying that the iPhone maker and the online search giant would partner to bring Siri into the agentic AI era. What this means exactly is uncertain. What we do know is that Google's Gemini AI models will power the next generation of Apple foundation models. Knowing Apple, we likely won't see Google or Gemini branding all over iPhones, iPads and MacBooks. Instead, Apple is likely giving Google the ability to create a bespoke engine for Apple Intelligence. Apple, however, will be driving the car.
"If this is a rom-com, you would say, in retrospect, that this was always going to be the outcome -- it just makes too much sense," said Andy Tsay, professor of information systems and analytics at Santa Clara University's Leavy School of Business.
Apple's recent deal with Google follows years of the two companies working together to ensure dominance in their respective fields. Thanks to the Department of Justice's antitrust trial against Google, 2022 court documents revealed that Google paid Apple $20 billion to remain the default search engine across Apple devices. This disincentivized Apple from making a competing online search engine and Google was able to glean valuable data from Apple users. The deal also marks an end to Apple's bumpy road to creating its own AI models, at least for now. And Apple's current partnership with OpenAI for Apple Intelligence didn't yield the results fans were hoping for, although things have gotten better. As Google and Samsung continue to integrate AI deeply into their devices, Apple's partnership with Google ensures it doesn't remain behind for another year.
Google and Apple didn't immediately respond to requests for comment.
Don't miss any of our unbiased tech content and lab-based reviews. Add CNET as a preferred Google source on Chrome.
Why Google ultimately won out
It's likely that OpenAI and Anthropic heavily petitioned Apple to be the partner to power Apple Intelligence. Past financial partnerships and current technological innovation are why Google likely won out, observers said.
"In terms of technology, I think Google's technology is far superior," said Humayun Sheikh, an early investor in DeepMind and current CEO and founder of Fetch.ai, a company that creates and/or facilitates connections between systems and AI agents. Google bought DeepMind in 2014 for $650 million. "Google probably, I would say, in my opinion, has the most grasp of what's going on in how to train these models and how to put these guardrails," Sheikh said.
Sheikh acknowledged his bias toward Google as he's long been in business with the company. Fetch is currently working with Google to bring Gemini models to its Agentverse platform.
"Google has a lot more experience on mobile devices as well, so they know how to optimize all of that better than OpenAI does, because they have their own phones and have their own mobile operating system," Sheikh said.
Should iPhone users be concerned about privacy?
With data, there's no company that siphons it up better than Google. It has access to data from billions of users worldwide, thanks to Google Search, Chrome and Android. It then lets advertisers bid to perfectly place targeted ads against that data.
Apple users, however, are keenly attuned to privacy. The launch of Apple Intelligence put privacy as a top concern in a 2024 CNET study. Despite Apple users' privacy concerns, Google's long-term role as the default search engine has given it access to a lot of iPhone and iPad user data.
Likely, consumers will be worried about what Gemini-powered Apple Intelligence might mean for their most sensitive on-device information.
"It's unlikely that the user query will be directed to the Google server, I don't think that's the case," said Haibing Lu, professor and department co-chair of information systems and analytics at Santa Clara University's Leavey School of Business. Lu is also the co-founder and CTO of AIConform, a risk automation platform.
What's more likely, Lu said, is that Google will build the model for Apple, and the AI will live on Apple's servers and devices.
In essence, experts said, it matters less what AI models are being used and more how they're being deployed and whose server is hosting them.
Will regulators intervene? What's the long-term play?
Having the world's second- and third-most valuable companies partner on AI could mean further entrenchment by Big Tech, pushing out smaller players like OpenAI, Anthropic and Mistral. This might upset regulators worldwide, but at least in America, it's likely this type of deal will be allowed to go through.
Tsay said the Trump administration has been less keen to challenge Big Tech than the Biden administration was. "I take it as they've got no problem with this, whereas the previous administration probably would have had a big problem."
Ultimately, with AI, Apple wants to achieve the initial promise of Siri, a digital companion that can make life easier for users. This requires a retooling of the assistant so that it can integrate with AI to understand what you want it to do and execute on it effectively. Getting there, however, means leveraging the technology from other agencies and firms. Apple Intelligence, for now, might be powered by Gemini, but that may not always be the case.
"They may have bought a short-to-medium term solution," said Ashish Nadkarni, group vice president and general manager for worldwide infrastructure research at IDC, a global market intelligence and data company. "Remember Apple Maps? Google Maps was the default for a long time. Same with the search engine. It's Google for now. Could change."
The reinvention of Siri for the AI age will mean a different relationship between users and software. Apps will become less prominent, fading into the background as Apple Intelligence picks up the slack.
"Apple is seeing the fact that your reliance on multiple apps is going to come down, and you're going to be relying on one where agents can actually carry out the task," said Sheikh. "Hence the agentic system, because every agent needs to kind of be autonomous and have this cognitive ability to actually do something, understand something."
