Siri Will be Replaced With Google Gemini
No, seriously…
My Medium friends can read this story over there as well.
What a move… Apple really did it.
They went to Google and basically said: “Hey, nice AI models you’ve got there. Mind if we run them on our servers?”
That’s the short version of a whole $1 billion custom Gemini deal that just took place in Silicon Valley. One of the largest and potentially most crucial deals in the last decade.
And honestly, it tells us pretty much everything about where Apple is in the AI race right now. Or does it?
Siri is… to put it nicely… a joke.
I use AI a lot. Daily. For work, for writing, for random “what is this thing?” questions.
And yet the most complex thing I trust Siri with is: “Wake me up at 6:30 (before I go go).”
So how did we end up with Apple is buying Google’s brain?
Apple tried to do it alone
It didn’t work. Obviously.
Apple isn’t poor. They’re sitting on enough cash to buy a small country. That’s an advantage. Duh…
They’ve already spent a ton of money training their own AI models. For years. But they don’t come close to Google. Or Meta. Or Anthropic.
That $1 billion price tag for the Google custom model is actually the cheaper option.
Apple looked at Anthropic. Looked at OpenAI. Looked at building more in-house.
And then went: “Yeah… just give us Google’s thing. That’s actually less painful.”
Classic Apple move, in a way. It’s not the fist time they did something like this.
They don’t like losing. But they really don’t like shipping something bad (again). And right now, their own models just aren’t good enough to fix Siri.
Tired of the iOS Liquid Glass Icons?
Here’s a fresh take! Try these glass icons* that bring back Mac Aqua vibes.
Apple did privacy too well
Which company do you trust more with your data: Apple or Google?
People pretend that’s a hard question. It’s not.
Google makes money with ads.
Meta makes money with ads.
Microsoft makes money by baking AI into everything.
Apple makes money by selling hardware and subscriptions.
Apple doesn’t really need your data to make money. That’s how it used to be. Which is why they’ve spent years locking it down, blocking trackers, and annoying Facebook’s entire ad business.
That’s great for us as users.
But for training massive AI models? It’s terrible.
Google, Meta, Microsoft, OpenAI, Anthropic… they all have huge piles of text, clicks, behavior, code, forum posts, whatever.
Apple has… your iCloud backup. Probably encrypted.
So while everyone else is scraping the planet, Apple basically said: “We’re good. We do privacy.”
And now they’ve noticed that “we do privacy” is very bad for AI. For now…
It’s all about usage data now
We’re moving from “train on the entire internet” to “train on what works.” Modern models aren’t just fed random web text anymore.
They’re fed:
successful tool calls
agent runs that didn’t crash into a wall
code suggestions people actually accepted
good answers that humans upvoted or didn’t bother editing
New AI tools are building private piles of “this is what good looks like” data. Not just data.
And again, Google, Microsoft, OpenAI, Anthropic — all have tons of this.
Apple, not so much. Because Siri never got good enough for people to use it like that in the first place.
So they’re missing legacy data and usage data.
Brutal combo.
Why Google is happy to help
So, if all that’s true, why the hell would Google help its biggest competitor?
Historically, the big tech rivalry was Apple vs Microsoft.
Then, for a while, Apple vs Google.
But today, I think Google doesn’t see Apple as the main threat in AI. Few do. Much bigger worries are:
Meta in ads
OpenAI in models
Anthropic backed by Amazon
Microsoft stapling AI onto every product like my daughter does glitter
Whatever the hell Elon Musk is cooking up
Apple, on the other hand, is kind of… useful for Google.
iPhone users make Google more money than Android users do. Google already pays Apple absurd amounts every year to stay the default search engine on iPhones.
So from Google’s perspective, this $1B deal is great:
First, they earn $1B.
Then, they keep Apple away from OpenAI / Anthropic (a bit more).
And they don’t lose any real strategic ground, because Apple isn’t going to become a model vendor.
Also, they’re not giving Apple their whole infrastructure. So Google gets money and leverage. And Apple…
Apple gets a brain transplant for Siri.
Win-win, right? Well…
But what does Apple actually want?
Long term, Apple doesn’t want to depend on anyone. That’s just how it is. Why would the most valuable company on the planet?
But short term, they need something that:
Doesn’t suck.
Can run on their Private Cloud Compute setup.
Keeps their privacy story somewhat intact.
And that private cloud thing is the key part of it. The idea is, your data goes to Apple’s servers, gets processed, but Apple never really “sees” it in a meaningful way. Even if it’s a Gemini model in the background.
If they can pull this off, they’ll be the only big player with a serious “cloud AI, but private” answers. Everyone else is kind of hand-waving their way around that.
Combine that with:
huge custom models in the cloud (Gemini-based)
smaller, tuned models on-device (locally, offline)
tons of iPhones out there with absurdly powerful chips
…and suddenly Apple could actually have a very strong AI story.
So… maybe that was the plan all along? Buying into AI?
Maybe.
Siri
Inside Apple, the AI+ML group has been such a disaster that people literally nicknamed it “aimless.” Apparently.
Now they’ve put Mike Rockwell (Vision Pro guy) in charge of Siri. He’s been openly frustrated with how bad it is, especially as a way to control Vision Pro (and other products).
Siri needs someone who’s not emotionally attached to its past and is willing to say: “This is not fine. Fix it.”
So, let’s try:
a working model (Gemini)
proper infrastructure (private cloud compute)
a mandate to stop shipping embarrassment
and a buttload of money
…and we might finally get a Siri that can do more than set timers.
Is this defeat?
No. On one hand, Apple is admitting: “We couldn’t get there alone. Not fast enough.” They’ve done that before. And still became a trillion $ company.
On the other hand, they’re doing it in a very Apple way:
They don’t send your data to Google.
They don’t become dependent on Google forever.
They use this as a bridge while they generate their own training data.
So, this isn’t Apple “giving up.” It’s Apple accepting, they were late and they’re privacy focus was an issue here.
They don’t have the data. Siri is a bad. And they want in on that AI gold rush.
What do you think: Smart short-term hack? Or a genius move for the future?
Brought to you by SparkLoop* – the #1 email recommendation tool. And it’s free!
The privacy angle is realy intresting here because Google basicaly gets paid while Apple keeps its data locked down. Its a huge validation of Gemini's capabilties when your biggest competitior comes knocking. This could turn into recurning revenue stream for Google from enterprise deals if they can prove the private cloud compute model works at scale.