For the first time since the iPhone launch, Apple finds itself behind in a major technology shift — years after the shift was apparent. This is a major red flag for the company.
Leading up to WWDC 2025, most Apple pundits aligned around John Gruber’s article “Something Is Rotten in the State of Cupertino” where the defining concept being that the Apple Intelligence miss points to a more soured core at Apple, Gruber concluded:
When mediocrity, excuses, and bullshit take root, they take over. A culture of excellence, accountability, and integrity cannot abide the acceptance of any of those things, and will quickly collapse upon itself with the acceptance of all three.
Gruber was right with that, and he still is right. And nothing at all which happened at WWDC 2025 changed any of this. Pre-WWDC, there was not a notable Apple blogger-type disagreeing here, which is in itself rare, but points to how obviously bad things are.
WWDC was a shit show of half baked ideas, and mostly defensive bullshit, and a lot of time spent trying to make Craig look cool. Too much time.
Anyways, what we know post WWDC about Apple:
- Apple made foundation models available to developers (this great, but not in itself that interesting) which won’t ship for public use until end of Summer (at best) and is not anywhere near what people think of when they think of modern AI tooling.
- Apple continued to say Apple Intelligence, with no real, anything to show for it.
- Apple then came out and said “good” Siri will be shipped in 2026. A year from now? Well after the OS 26 releases, and likely several iterations of frontier level LLMs being shipped from now.
- We got a worse UI in Liquid Glass, but whatever on that for now.
Jason Snell at Six Colors has a very weird article which is partly making clear that Apple is not good at AI, partly showing they are using AI, and partly saying maybe they shouldn’t bother thinking about AI, this line sticks out to me:
Apple’s AI stuff needs to get better, but what the company really needs to be is a builder of platforms that are good for users, including those who want to use AI to perform tasks.
Yes, but also, no. This has been bothering me all week. I was going to spend time talking about how shit Liquid Glass is as a UI (a UI that seemingly Snell and Gruber both like somehow?). Instead, I can’t get this AI bullshit out of my head when it comes to Apple.
And I only saw this post from Snell because Gruber linked to it, agreeing with Snell’s conclusion — wild times.
What I am saying is that Apple is acting like Blackberry and Nokia at the launch of the iPhone.
Apple is one of the largest and most profitable companies in the entire world.
So while OpenAI caught most of the huge tech companies flat footed, and really only Anthropic (and to a lesser extent Mistral) seemed prepared — it’s not to say Apple had not been aware of, or even working on, LLMs — but they didn’t have it. No big tech company did.
Let’s take a look at how big tech reacted to ChatGPT 3.5:
- Alphabet (Google): seems to have done an ‘oh shit’, they put their heads down and invested in the work knowing they would get there soon enough (spoiler: they are there now).
- Microsoft: spent money like there was no tomorrow, partnered with OpenAI, bought companies, hired mountains of people, and built actually useable tools around the tech almost right away.
- Meta: spent mountains of cash, focused up, and started cranking out models. They are in the game for sure.
- Apple: publish some white papers, mostly throwing cold water, have some really limited models to run locally, and then lied about what they would ship and when. They still have not shipped, and are, in no way, ‘in the game’.
In billions of dollars, here’s the cumulative net income reported for those company from 2023-2024 to give you some perspective:
- Alphabet: $172.92
- Microsoft: $158.96
- Meta: $101.46
- Apple: $193.00
(Microsoft and Apple are estimated, waiting for reports, etc etc.) Those are all above the valuations of most of the prominent frontier LLM companies themselves.
This is also to say, that in no way, is Apple and underdog — they are the top dog. Does Apple primarily make platforms and devices? Kind of, yes, but Alphabet primarily does Search, and Meta is social media. That’s not really an excuse as to why Apple is behind, nor should it be a reason for them to “focus on the platforms” because those Platforms are not where AI is happening: web and NVIDIA are where it’s happening. And AI is already transcending platforms — this is Microsoft not shipping a web browser back in the mid-nineties.
It’s the first time since the iPhone, Apple’s not in the game in any major way. They may normally ship late, and ship great — but they are announcing early, and failing to ship. That’s not shade, that’s fact. It’s been years, still nothing.
They have the money. They have the talent, or at least the means (money) to get the talent.
It seems abundantly clear that Apple is lacking the vision. A dangerous spot for a company this large.
There are, as far as I can see, three paths to take (since 11/2022 at least): integrate AI tooling through partnerships, build your own LLMs, or do both. Microsoft did both. Everyone else went down building their own. Apple? They kind of are trying to do both, but they are failing horribly at both which is WILD.
Something I keep thinking about on this is Meta and Alphabet — two companies that I really dislike for privacy reasons — but two companies who were caught with nothing, then hit the accelerator so hard it’s been impressive to see.
I did some quick estimates, since ChatGPT hit mainstream in late 2022 here’s the (very rough, because I used AIs for this) model release estimates:
- Anthropic: 15 models
- Apple: 3 Models
- Google / Alphabet: 16 models
- Meta: 7 models
- Microsoft: 15 models
- Mistral: 19 models
- OpenAI: 10 models not counting vision/audio models
- xAI: 6 models
Depending on how you count, you’ll see that I left some out — there’s a lot of specialized models, I mostly focused on the larger releases. There are, some really interesting take aways from this:
- Mistral is quite a small company compared to every other company on the list, and is bound by EU regulations, and yet they have cranked out 19 models in that time — which is pretty fucking impressive if you ask me. And they are solid models.
- Microsoft has largely focused on partnerships, but with 15 models, they are no rookie — they just don’t have a frontier model people use for general purpose.
- Anthropic is also small, yet jockeys for top spot with quality all the time, and put out 15 models, while people swear by their models.
- Meta went from shit models, then the 3 (and later) Llama models came out, and surprise: they are good. Only privacy will stand in their way.
- Alphabet/Google: went from trash to 2.5 Pro being a signal that they are a top-tier frontier creator here. It’s a fantastic model. 16 releases is no slouch!
- xAI went from nothing, to utter shit, to Grok-3. While I don’t use Grok-3, it should not be discounted — yes it has “issues” — but by all accounts it’s a really good model.
- OpenAI: you could likely give them a higher release count, but they’ve been top all the damn way, and they are a fraction the size of many of the companies.
Then there is Apple. I found three releases, and to be clear: none of these are releases people actually use outside of the Genmoji. In other words: Apple off loads to frontier models quickly (OpenAI). They have a few small models built, mostly exploring on-device quality of smaller models. And that is the correct term: exploration models. It is possible to be leading in this time: Meta, Alphabet, and xAI all pulled off huge feats in the same time.
Perhaps Apple is simply exploring, but even there they would be very behind. Microsoft is very much cranking the same type of exploration models out, and they have 15 out there. Could it be that Apple has more, and they are hidden from the public? It’s absolutely got to be the case. Yet, that would also stand to reason they are not great models.
And this is the worry: Apple, by every metric I can think of, is lagging further behind with each passing month in the generative AI space. It’s not a space where you can hide the models, you need people using them so you can make them better — which is another business area Apple is not good at — so even if they have some great stuff tucked away, it almost is a fallacy it could be that great since they don’t have any real practical experience.
Google didn’t take Gemini from being crap to being leading by toiling away hidden. Nope, they put it out there, and iterated. As has every company on the list, except Apple, who has largely stuck to keeping the models decently small and on device and releasing research papers.
It’s not a happy path for Apple long term.
Even if partnership is the right way forward for Apple, in that space they are not even remotely being active. Microsoft is taking that approach and they have partnered, and launched with partners, end user level features and products at breakneck speed. While Apple has touched on some of this, it’s mostly not good. So even if Apple has decided to focus on these partnerships, they are seemingly well behind the curve.
The Damned Point
Apple is significantly behind no matter what their strategy is. Lagging partnerships with the integration Apple has done (Siri offloading to ChatGPT at different times) being pretty awful from a user experience standpoint.
On the LLM creation front, Apple is even further behind. The models they have released are very niche, and they’ve not really shown any reason to believe they’ll have competitive models soon.
Apple’s behind — and not by a little bit. I am not confident the current Apple leadership knows how to catch up, or are even willing to admit they are behind.
