“Winning” in AI is a little more complex than you may think.

In our Winners and Losers of 2025 article, we talked a lot about things that seemed like Impossibilities in the tech world coming true. Xbox having a PSP-like handheld in the ROG Xbox Ally, Airdrop now being available on Android phones, things that basically would have seemed unthinkable two years ago. Well it seems like Google and Apple decided to make a trifecta of Impossibilities, because last week, they both announced that Apple’s new Siri and Apple intelligence efforts, (which were essentially failing) are now going to be rebuilt and revitalized by being based on Google’s Gemini models, essentially meaning that the heart of Siri and every other Apple intelligence feature you’ll use, will literally just be Gemini. This was a move a lot of people saw coming late last year, but it does beg the question here: with Google making their biggest competitor admit and concede to Gemini, while starting to leapfrog OpenAI in some areas, have they basically won the AI race?

Well, for starters, let’s start by delving into the Google and Apple deal a little more. Stated in a blog post, it specifically states that the foundation of the upcoming next generation Siri and Apple intelligence will be based on “Gemini’s foundation models”. A confusing statement to say the least, but it’s probably in part because Apple, being Apple, would want to just outright say “hey, we suck at AI, Google’s kind of better at it, so we’re letting them do the work”. It’s not a position the company would like to be in, but it is one they’ve ended up in after almost 3 years of announcing a big AI shift with Apple intelligence and Siri that never happened. It hasn’t completely derailed Apple, of course, but with AI being the next major frontier in the tech industry, it’s something they have to fix, no matter how many iPhones they can sell. As such, a deal that is in some ways similar to Microsoft OpenAI’s situation is emerging, with Apple specifically hiring Google to employ their Gemini models to be the building blocks of Siri and Apple intelligence, similar to Microsoft and Copilot, while not owning part of Google like Microsoft owns part of OpenAI. Instead, Apple will be paying Google a $2 billion fee per year while they apparently learn from Google’s work to eventually build their own models wholesale. This is the tech industry equivalent of, for example, Mercedes hiring BMW to make an AMG engine because their current options just cannot compete, and well, yes, it has been extremely embarrassing for Apple, while showing that Google’s current victorious streak in AI genuinely shouldn’t be underestimated. That being said, it doesn’t outright spell a “win” in the AI wars because when that starts to be considered, three key areas should be focused on.

The cost and environmental problem

Now, if you’ve taken more than 5 minutes to read about AI products and how they run and get to your device, you know for a fact that never in the past decade or so has the tech industry rushed into something so costly and, depending on who you ask, wasteful. The AI race has led to multiple companies, venture capitalists, and huge wells of both technical knowledge and just straight-up cash to be poured into developing better, faster, and more capable AI tools and products. The problem is that, for all that cost, a loop of what looks like diminishing returns has started to emerge. Don’t get us wrong, ChatGPT, Gemini, and other products like Cursor and Claude Code have rightfully gained the love of many and kicked this movement into high gear. But year after year, the achievements and capabilities have seemed less mind-blowing and more “well, of course it should do that” or “of course it fixed that major flaw”. All while on top of all that money put into data centers, research labs, and think tanks, AI has become the most environmentally wasteful technical endeavor of maybe the past 50 years, with everything from wastage of water to increased carbon gases becoming a byproduct of the energy, cooling, and compute required to run all these AI tools. As such, one of the biggest ways an AI winner will be determined is whichever company is able to run cutting-edge models in the cheapest and most sustainable way possible. Especially if the latter brings us much closer to an environmental cataclysm. This is why models like DeepSeek have gained so much infamy in the tech world, as the original R1 model was able to run on much older GPUs and required much less money and resources, technical and environmental. Now, does Google have an edge in this part of the race?

Well, actually yes, as it’s custom-made Tensor TPUs (essentially custom-made GPUs), are quite a bit cheaper to run, manufacture and maintain than the NVIDIA GPUs everyone is clamoring after in the field. It still doesn’t negate the major problems the whole industry faces, however. And so while Google’s progress here shouldn’t be undermined, it’s still not quite enough to completely ignore everything going on, as a leapfrog from a chipmaker partnering with a competitor is still very much possible.

The Distribution Race

It could be argued that this point, more than all others, is where Google has genuinely, definitely won. Ever since the smartphone wars began, one very key truth became clear to anyone paying a passing interest in the tech world: having the most users often means having much more success. Sure, there may be exceptions like in high value hardware devices, but when it comes to software and any other digital goods, the term critical mass became a euphonism for “we’re a success”. When your app or service is in everyone’s hands, you can literally do anything with it, whether that’s introducing subscriptions, advertising or even pivoting into whole new businesses like digital stores or direct in-app purchases. On top of that, there’s the very important brand and ubiquity win that comes with it. You become the standard, and everything/ everyone else has to be compared to you in order to become successful.

In the AI Race, distribution is probably the biggest battleground for average consumers, as that dominance is required at first in order for any of these companies to more directly make money out of the users to begin with. ChatGPT obviously won the early advantage and brand phase, as now all of these bots are compared to it off the bat, whether they want to or not. Meta even has a secret advantage with how much people use Meta AI, especially in developing nations through WhatsApp. But think about this: every Android phone you see probably already uses Gemini right now, whether the user knows they are directly engaging with it or not. It’s seamlessly keeps being woven into other Google apps like mail, Calendar and of course replaced Google Assistant, so it’s already a key way to engage with key Google functions and products.

Now add the fact that EVERY IPHONE that can run Apple intelligence is essentially going to be powered by Gemini at its core, and you start to see why this is a big deal. Google would have turned the majority of Android phones and assumingly, a key chunk of iPhones, iPad and even Macs to run on its infrastructure. It’s a huge win for distribution, even if it doesn’t immediately bring up some of the same distribution advantages. For starters the average iPhone user likely won’t know what is actually powering Siri unless they dig into some settings. And while Google’s biggest gain from these kinds of deals is usually data, Apple did seem to point out how it will still handle the customer data and basically make sure Google doesn’t get to directly handle it. But still in a world where Microsoft has abandoned all common sense to chase the stock prices it’s AI moves give it, Google gaining some good PR from this probably helps too.

The killer app problem

And finally, there’s what arguably continues to be the most important problem when it comes to winning the AI race: a killer app, AKA an application or use case that makes it all seem worth it. Every major step in technological innovation needs one. For smartphones, it was instant quality internet and social media. For the cloud, it was a huge shift in how websites and apps run ( not to mention persona computing and corporate infrastructure). Even for Electric Vehicles, while currently unsustainable in certain markets, they have a genuine environmental and sustainable reason for existing. When it comes to AI, though, those reasons to exist get iffy depending on who you ask. Especially at a daily consumer level. Most people still regard AI, and its uses, to mostly be focused around finding information, specifically asking  ChatGPT a question and it answers, and while that is genuinely interesting and useful, it’s arguable that the billions of companies are sinking into this and causing tons of problems that aren’t worth what’s essentially “Google but better” for most people. Hence, the need for an undeniable killer app exists.

Everyone is, of course, attempting one in their own way. Apple and Google are both ironically competing by leveraging their ecosystems, pushing Apple intelligence and Gemini as all-knowing assistants that can help you manage everything from your emails to your everyday commute, even your spending and saving patterns in Google’s case. They both want to basically create JARVIS from Iron Man, and while Google has started an early version of this, they still haven’t exactly cracked the code to the point where a person has their whole life managed by Gemini. On that same note, companies like OpenAI have pushed out all sorts of products from their own riffs on a personal assistant to the much more well-known Sora video generators and even their own take on a web browser or search engine. Again, all these products, while having great uses, haven’t quite hit the mark. And the same could be said for multiple similar products from competitors as well.

This is why the one proven use case for AI tools right now is basically SAAS subscriptions and enterprise solutions. Do I really need AI looking into my private spending habits? Probably not. But does it make more sense when that AI manages my multinational company’s expense spreadsheets? Definitely. Same goes for even AI bots that help coordinate corporate calendars and project management tasks. The closest to a killer app AI has gotten in the enterprise sphere, however, is definitely when it comes to coding and other software programming/development tasks. Even back when GitHub Copilot first came out, people saw the potential, and now in a world of Cursor or Claude, code companies genuinely have restructured their whole software teams around such products. In fact, let’s take a moment to just appreciate the work Anthropic does as a company because it’s arguable that, between regular Claude, Claude Cowork and Code, they’ve actually started to build an ecosystem of AI work tools that more and more people trust and keep using. Credit can also be given to Microsoft here, but well with how they shove Copilot down the throats of regular Windows users, we don’t feel as compelled to appreciate them.

One thing may be noticeable about everything I just wrote though: none of these products have broken through to everyday consumers, while those that have only done so in an “it’s for work” sense. For many other technological pushes, that’s probably enough, but for the AI space it doesn’t seem to be, as again, with all the costs, nothing less of an all-encompassing monopoly win can satisfy these companies. And Google needs that maybe more than anyone else that isn’t OpenAI.

So with all that said, has Google won the AI race? Well hopefully after reading all this you’ve realised the “race” isn’t as easily won as most people would like to quantify it. But doing a general analysis of where everyone stands, Google is in a much better place than it was years ago. I’d argue only Anthropic has a cleaner track record in the AI space right now and that’s arguably just because of a smaller scale approach to building software than Google. They certainly don’t have the Apple deal that started all this noise after all. But this is still a very dynamic and drastic space, where a lot can happen. And that’s why it’s so much fun to observe. Here’s to hoping those “fun to watch” activities don’t end up drying up the ocean, though, fingers crossed.

Related Stories: