Apple Finally Agrees to Fix Siri
Siri sucks. You know it. I know it. You'd think Apple would know it given that it has been the butt of jokes for nearly 15 years running. But their actions (and marketing) have always suggested otherwise. Siri leads to fun moments of Zooey Deschanel whimsically dancing around her home, you see – not Larry David swearing at his phone in his car. You're simply holding it wrong, or something.
But in the past 18 months, something finally woke Apple up from the drunken slumber, dreaming of Steve Jobs' Siri promises. That something was the company completely and utterly shitting the bed when it came to new promises for Siri in the Age of AI. Granted, given the track-record, we had no reason to believe them, but many did, because they showed the new functionality on stage at WWDC! Well, sort of. Virtually. Maybe. Maybe not. Probably not – at least not anything that was actually close to shipping.
A complete and utter embarrassment. Commercials had to be pulled back. Heads rolled. That's what it seemingly took to wake up Apple to the reality we've all been living with for over a decade. Again, Siri sucks.
And now, at last, we may have an actual solution. It's still a very much "I'll believe it when I see it" situation, but there is one key difference this time: Apple is finally giving Siri a brain transplant. Some of us have been calling for this for a while – again, we're living in the Age of AI, so Siri sucking this hard is all the more unacceptable – and now it finally seems on the verge of happening.
Here's Mark Gurman reporting for Bloomberg:
Apple Inc. is planning to pay about $1 billion a year for an ultrapowerful 1.2 trillion parameter artificial intelligence model developed by Alphabet Inc.’s Google that would help run its long-promised overhaul of the Siri voice assistant, according to people with knowledge of the matter.
Following an extensive evaluation period, the two companies are now finalizing an agreement that would give Apple access to Google’s technology, according to the people, who asked not to be identified because the deliberations are private.
While this news isn't exactly new – it has been known for a while that Apple was doing a sort of "bake off" internally to potentially find a third-party partner to help fix their AI efforts – and we even knew that Google was likely in pole position, it's still reassuring to read a report with actual numbers and potential timetables. That suggests a deal that is indeed at the finish line, if not already over it. And it makes a new Siri – an actually working Siri – possible in the Spring of 2026.
Halle-fucking-lujah.
I know, I know, I really shouldn't get ahead of myself here. But I put a lot of faith in Google – and not because I worked there for a decade-plus, I swear. I'm still an Apple guy through-and-through – and was during my entire tenure within Google! They simply still make the best devices. And I want to use the best devices. But what they don't make the best of is AI, as I concluded by post on the possibility of this news a couple months ago:
I sort of love this for them. Both of them. Imagine an iPhone largely powered by Gemini. The Pixel team can’t love this notion, but Jimmy Fallon will be totally stoked regardless. And it’s the right move for Google. The iPhone is still the iPhone. The Pixel is not the iPhone. And for Apple. Siri is still Siri. Siri is not Gemini.
We are so back. They are so back.
I kicked off that piece noting the long, illustrious-turned-contentious (to say the least) history of Google and Apple. From the best of friends to "thermonuclear" war and back again. The only constant being the Google Search deal. Once the courts kept that intact despite Google losing an antitrust case, the stage was set for this new deal.
Apple may not have wanted to pay Anthropic $1.5B a year to use Claude but $1B a year to a partner that is paying you $20B+ for that Search deal? That can be just an in-kind deal! "Google, that $25B you own us this year? Make it $24B, but we'll take a custom build of Gemini. Deal?"
Deal.
Speaking of that model, back to Gurman with a few more details:
Under the arrangement, Google’s Gemini model will handle Siri’s summarizer and planner functions — the components that help the voice assistant synthesize information and decide how to execute complex tasks. Some Siri features will continue to use Apple’s in-house models.
The model will run on Apple’s own Private Cloud Compute servers, ensuring that user data remains walled off from Google’s infrastructure. Apple has already allocated AI server hardware to help power the model.
There are probably a few other interesting wrinkles in there – one of which may be Apple's willingness to do this because a lot of their cloud infrastructure is already running on Google Cloud. So this may not be as heavy of a lift and as big of an ask as it may seem on the surface. And while the "walled off" aspect is clearly a must for Apple here, you could also imagine that the company may be willing to share some data – fully anonymized, of course – back to help constantly improve the model. And that may speak to why Google would want to do this deal (well, that and the money). Apple has devices in the wild at a scale that basically no one can match. Maybe Samsung, but this unlocks a totally new user base.
While the partnership is substantial, it’s unlikely to be promoted publicly. Apple will treat Google as a behind-the-scenes technology supplier instead. That would make the pact different than the companies’ Safari browser deal, which made Google the default search engine.
You could see why Apple might request this – it would be an embarrassing admission that they can't fix Siri themselves, that they need Google – but given this news already circulating, I also wouldn't be shocked if they do talk about it more than Gurman is suggesting. If nothing else, it would help assure those of us who have lost faith in Siri that she really may be "fixed" now. My advice to Apple would be to suck it up and own it. They always tout wanting to give their user base the best services – long an explanation for continuing to use Google Search – and Gemini is, if not the best, damn close.1
Speaking of, this might also be slightly awkward because Apple was last seen on stage (well, on video) at WWDC a couple years ago touting OpenAI as having the best AI tech and product with ChatGPT. It's a bit odd they're not partnering with them here given that partnership already in place, but everyone knows the scaling issues OpenAI faces – and moving over a billion potential Siri users would seemingly be a big problem, for both sides. Not to mention the whole Private Cloud Compute element. This is an operation that Google is just uniquely qualified to handle.
The agreement is also separate from earlier talks about integrating Gemini directly into Siri as a chatbot. Those discussions came close to fruition in both 2024 and earlier this year but ultimately didn’t materialize into a feature. The partnership also doesn’t weave Google AI search into Apple’s operating systems.
This is causing quite a bit of confusion, at least in my social media feeds. This new deal with Google to help power Siri would be different from the aforementioned OpenAI deal where Siri "falls back" to ChatGPT if it doesn't know an answer. (And how Apple had previously been talking – in public! – about using Gemini as another option here.) Presumably that functionality will remain intact, but it will probably be used even less given that Gemini-powered Siri should be more than capable for most requests!
There's also seemingly some confusion around the 1.2T parameter model cited above. While such a model would be far ahead of Apple's current best-of-breed models for Siri – perhaps by an order of magnitude! – it's not some magical new threshold that Apple will get access to first. Google doesn't disclose how large their current Gemini models are, but there's some thought that Gemini 2.5 is already this large, or close to it. Ditto with GPT-5. Perhaps even larger.
Further, it's widely known that Google's next model, Gemini 3, is right around the corner. You can bet that will be at least 1.5T when it comes to parameters – in fact, that may be what this bespoke model for Apple is based on. That's just a guess, but it seems like a reasonable one.
At the same time, per the report, Apple is working on their own 1T+ models internally with the notion of eventually taking over more of the Siri workload again one day. Google will be going into this partnership eyes wide open this time, but they also must know that Apple – in the midst of a very real, and very problematic brain drain on their models team – may struggle to match the state of the art from the current leaders in AI. Which is to say, this might still be a "many year" proposition for Google to help Apple here.
And I think I like the hybrid approach – where Gemini can handle complex queries while Apple can maintain their current AI systems for, say, setting timers in iOS – more than Amazon's early attempts to totally rebuild Alexa on the fly. In a way, it seems like they eventually ended up at the same place, with Amazon using their deep partnership with Anthropic to help get Alexa+ up to snuff (sort of).
One unknown here remains the promised personalized features of Siri from back at WWDC 2024. Is Apple going to be able to use Gemini for the always important getting-grandma-home-from-the-airport task? Given that Google itself touted such functionality at their own I/O conference, perhaps. Or maybe Apple is still working on their own models for this work. Seemingly we'll see soon enough!
Regardless, I'm just excited to have a Siri that knows who won which Super Bowl.
M.G. Siegler
M.G. Siegler
M.G. Siegler
1 One other wrinkle per the reporting is that Apple may have actually thought Anthropic's models were the best –better than both OpenAI's and Google's – but simply couldn't figure out the right kind of deal to make it work... ↩