Google Gemini Could Be the Ceiling on Apple's AI Ambitions

[MacRumors]

Apple analyst Ming-Chi Kuo says the real test of today's WWDC keynote is whether Apple can deliver better AI experiences than Google using the same Gemini models.

Apple is using Google's Gemini to underpin the revamped version of Siri and new Apple Intelligence features. The key takeaway from WWDC, Kuo argues in a new post on X, will not be the short-term market reaction after the event. It will be whether Apple, using the same Gemini models, can deliver better AI applications, agentic workflows, and on-device and cloud hybrid experiences than Google.

If the answer is yes, it would help extend Apple's "bull" case. If the answer is no, the implication is that Apple's ceiling is set by a model it does not control. Kuo raised the point arguing against the market sentiment that "Even if Apple is temporarily behind on AI, it will ultimately catch up and come out ahead."

Nevertheless, Kuo believes Apple's business momentum will stay strong through year-end based on his latest supply-chain checks, which he expects observers to spin as "If Apple is doing this well without AI, just imagine once it has AI." Other reports suggest Apple's longer-term advantage could lie in on-device AI, with the company expected to show how its custom silicon lets it process more AI queries directly on the device rather than in the cloud.

Kuo expects today's announcements to have little bearing on the direction of Apple's stock price in the second half of the year. Regardless of what Apple says at WWDC, he argues, the positive second half of 2026 share-price trend is unlikely to change as long as the core narrative stays intact.

The longer-term risk is more pointed. Should Apple fail to outdo Google with Gemini, Kuo says the stock would not necessarily turn bearish, but the assumption that Apple "will ultimately come out ahead" would begin to face growing scrutiny. How much longer the bull narrative can last beyond 2026, in his view, is what makes the keynote worth watching closely.

Article Link: Google Gemini Could Be the Ceiling on Apple's AI Ambitions

 

[The Game 161]

where apple al with gemini models should shine would be with app developers taking al within apps better. we know android developers just don't put the same care or effort into their apps. it's why samsungs now bar has been rather pointless as hardly any apps support it

 

[szw-mapple fan]

I feel like we’ve pretty much hit the ceiling of what can be done with LLMs functionality-wise. The specs/scores of different models will keep getting better but I don’t think it matters much what model Apple ends up using longer term. It feels like we’ve already figured out the most common consumer grade use cases and none of them really require much more than an adequate model.

 

[datsdatcharlie]

agreed

I feel like we’ve pretty much hit the ceiling of what can be done with LLMs functionality-wise. The specs/scores will keep getting better but I don’t think it matters much for consumer grade AI since it seems like we’ve already figured out the most useful use cases

 

[jclardy]

I think Apple is in a good position here, mainly because what most people would find useful is a proper harness. OpenClaw became popular because it finally brought things full circle, you didn't just ask something and follow directions - it could actually just do everything.

On an iPhone if Apple can securely expose data and functionality (Which is possible because of the App Intents system used for shortcuts) and build a solid harness, they essentially are agnostic to the model provider.

 

[acader]

Short answer NO. If Google were to provide the exact functionality of Gemini to Apple and Apple does a much better job, then Gemini on Android/ChromeOS will be stunted. We had seen in the very first version of Google Maps when Apple Maps did not exist. Google Maps was a little inferior and lack 100% of the functionality of Google Maps.

 

[the future]

If the answer is no, the implication is that Apple's ceiling is set by a model it does not control.

If the answer is no, the implication is that Apple is so smart as to be model-agnostic so they a) don‘t spend gazillions on AI like the other guys and b) just hop to the model that is ahead at the time (if necessary).

 

[BSoares]

Apple is far behind and the amount of money they would need to invest to catch up doesn't make sense since Siri is free and Apple doesn't force a lot of ads on us so they will never get that money back. It makes more sense to pay Google who has lots of reasons to continue investing in the area and just use their expertise.

 

[svish]

Think Apple's Siri and Apple Intelligence will heavily depend on Gemini models. However with Apple's focus on privacy and a massive number of active devices, it will have an edge. Waiting to see the improvements later today.

 

[DavidLeblond]

I feel like we’ve pretty much hit the ceiling of what can be done with LLMs functionality-wise. The specs/scores will keep getting better but I don’t think it matters much for consumer grade AI since it seems like we’ve already figured out the most useful use cases.

I'm honestly astonished when I read people throwing their entire life into AI. I have the coding helper plugged into my development environments and the most used feature of it is the "snooze" button because when it gets stuff wrong it will repeatedly shove the wrong answer in my face so much it kills my productivity. And I have used it in another area of my job in a place that made it very useful... until one day something went off with the model and it started hallucinating like crazy so now I'm back to my old processes. It went from being useful to a burden.

 

[kalsta]

Kuo raised the point arguing against the market sentiment that "Even if Apple is temporarily behind on AI, it will ultimately catch up and come out ahead."

It’s a bit unclear who that quote is from — is it Kuo or ‘the market’?

 

[DavidLeblond]

It’s a bit unclear who that quote is from — is it Kuo or ‘the market’?

I assumed that quote was from Kuo but he was talking about the market sentiment. So, both!

 

[AppleFan91]

My guess is Gemini is just a multi-year stopgap, similar to Google Maps and The Weather Channel (or on the hardware side, working with Intel/Qualcomm, etc), until their own in-house model is ready.
This buys them time to make their in-house model a true competitor, even if it takes years behind the scenes. It removes the current pressure to put out a potentially half-baked product (again).

 

[MacATDBB]

It shouldn't be impossible to polish Gemini Pro's results a bit. Of the three major players (Gemini, Chat, Claude), I find Gemini Pro most frequently has to walk back replies or admits to hallucinating results. If apple simply added the follow-up question "Are you confident about that?", then parsed that output before returning a result, they could probably make a notable improvement akin to Claude's "extended thinking" mode (but at some resource cost).

 

[Wizec]

Gemini has been hopelessly wrong for me for weeks now. I’ll ask it a question, then press it because it’s wrong and the next response is, “you’re absolutely right, thanks for catching my mistake…” this does not bode well for the short term.

 

[marblesbarkley]

Gemini quality varies A LOT. Using their new 3.1 Flash Lite gives awful answers. Using Flash Extended is much much better.

Wonder which Siri will get. Difference is night and day.

All versions hallucinate of course because AI is fundamentally flawed and we’re all supposed to ignore it because investors told us to.

 

[ghostface147]

Gemini does do silly things from time to time. I’ll ask it to list out my stack and it gets it wrong consistently. I tell it to forget past info and use the current info given and it forgets again. Let me ask it now real quick. Yup, had to ask it three times to get it right.

It makes routine math mistakes when adding up troy ounces. That’s pretty much the brunt of it, but I’ve gotten to the point of asking it to double check itself routinely. However overall, it does well. Just a few things here and there.

 

[Devyn89]

It’s bizarre to me that no matter how many times consumers say we don’t care about AI it keeps be talked about as this thing that will make or break Apple in the future.

 

[mikebenton]

I may be in the minority, but I'm not particularly concerned about whether Apple uses Gemini, ChatGPT, or something else behind the scenes. Up to this point, most of what has been marketed as AI on smartphones has felt more like a collection of features than something truly transformative. Image generation, custom emojis, and playground-style tools are interesting demos, but I don't think that's where the long-term value is.

The real opportunity is contextual AI. An assistant that understands my emails, calendar, messages, documents, photos, travel plans, habits, and preferences—and can connect all of that information together in a useful way. That's where AI stops being a gimmick and starts becoming genuinely valuable.

What's interesting about Apple is that they are one of the few companies positioned to do this while keeping much of the context on-device. If that vision is the goal, then the bigger question isn't whether Apple uses Gemini for some requests. The question is where the line is drawn between local intelligence and cloud intelligence.

Personally, I don't care if Apple uses Gemini, ChatGPT, or another model for complex reasoning tasks that need cloud-scale compute. What I care about is Apple building an assistant that understands the context of my digital life and can act on it intelligently. That's the part that could fundamentally change how we use our phones.

We're still very early. The industry has spent the last two years showing us what AI can create. The next phase will be showing us what AI can understand. I think that's the more important milestone.

 

[Biro]

I’ve thought for a while that the main reason Apple is behind on AI is because it was trying to make either all or most of it work on device in the interest of privacy. It’s clear that approach isn’t going to work - at least not yet. But it might in the years ahead. So they’re using Gemini for now and that makes sense. It is quite possible they’d can still offer users a bit more than Google can with the same basic LLM underneath.

 

[HouseLannister]

Apple can't surpass Gemini because Google is iterating on this stuff too fast for Apple. Apple launches products and services and then gets around to updating them after a year or more. I am still waiting on App Library to get good from its initial launch in iOS 14. (6 years later and NO changes.) Apple isn't an AI company and has no interest in being an AI company. Whatever they launch today will be what we get. It could be good enough, but it will already feel behind and it will only get worse with time. Come back in two years and we will be complaining about how dumb Siri is all over again because Gemini lapped it several times.

 

[Biro]

Gemini does do silly things from time to time. I’ll ask it to list out my stack and it gets it wrong consistently. I tell it to forget past info and use the current info given and it forgets again. Let me ask it now real quick. Yup, had to ask it three times to get it right.

Kind of like spellcheck autocorrect.

 

[DogDogDogFood]

We’ll probably know more about this in a few hours, but nothing about the January statement suggests that Apple will be “using the same Gemini models”. It sounds more like Apple will be using Gemini for distillation or with a bunch of custom post-training or something like that.

 

[Shalev Lazarof]

What's kue or whatever his name is engineering background?

 

[ProbablyDylan]

Gemini not being very good doesn't bode well for something heavily based on it.

They really should've just bought Anthropic before their valuation blew up