The real reason why OpenAI spent $6.5 billion on Jony Ive's AI startup

• OpenAI bought Jony Ive's io to enhance product distribution and user reach.
• Generative AI competition is now more about distribution rather than technology.
• Google's vast distribution network poses a challenge for OpenAI. Ive's gadgets could help.

OpenAI's decision to buy Jony Ive's gadget company, io, is about distribution.

The generative AI race has entered a new stage. It used to be about creating the best AI models, but they're all pretty similar these days.

Who ultimately wins will depend a lot more on distribution, and less on the quality of the underlying technology.

Getting ChatGPT and other OpenAI models and products into the hands of users — that's what really counts. Without that direct relationship, these products won't be used as much, or it will cost a lot to get the offerings to consumers indirectly.

Ive designed the ultimate distribution tool for technology, Apple's iPhone. Then he left and started work on io, a new type of device company built for the AI era.

AI could spark a wave of new gadgets

While phones still dominate, generative AI could change that. We might all wear smart glasses with AI chatbots built in. Meta, Apple, Google, and others are working on this. We could have a little clip thingy attached to our shirts, so we can converse constantly with AI models and chatbots. Who knows what else might work in this new era?

Either way, if you're Sam Altman running OpenAI, you don't want to have Google or Meta or Apple standing in between you and your users. You know what happens then? You end up having to pay for distribution. Mark Zuckerberg despises being an app on Apple's mobile device platform, which takes a juicy 30% fee from many developers. Even Google pays Apple roughly $20 billion a year to have its search engine distributed on iPhones and other Apple gadgets.

Does Altman want to pay Apple $20 billion in a few years? Does he want to give Tim Cook 30% of the revenue OpenAI generates from ChatGPT paid apps on iPhones? Of course not.

One solution is to hire the original iPhone designer to build OpenAI's own gadgets. This quote stood out to me in the OpenAI announcement about the deal with Ive and his io company.

"It became clear that our ambitions to develop, engineer and manufacture a new family of products demanded an entirely new company," Altman and Ive wrote.

Even if this hardware journey costs billions of dollars, it could be cheaper than paying other tech giants for distribution. And at least you control your own destiny and have that direct relationship with users.

io vs IO

I'm writing this from Google IO, the internet giant's annual conference. Altman loves to crash this party. He did it last year, and he's doing it again today. It's all the more perfect because Ive's device company is called io.

My interpretation of this party-crashing is that Altman could be pretty worried about Google.

Despite that very expensive deal with Apple, Google is a master of distribution, and it's using everything at its disposal to quickly get its new AI products and tools into as many hands as possible. These are the offerings that compete directly with ChatGPT.

Here are some examples of Google's distribution power, taken partly from this week at Google IO and partly from all the work Google has been doing for the past 20 years or so. For a startup like OpenAI, this must be terrifying.

• Android, Google's mobile operating system, supports more than 3 billion devices. The company is prominently displaying its Gemini AI chatbot service on as many Android gadgets as possible.
• There are millions of Pixel devices and Chromebooks out there. And guess what? Google is weaving Gemini into many of these phones and laptops.
• Chromebooks, Android, and Pixel are mainly ways to distribute Google technology to users directly. Chrome is the default browser on these Googley gadgets. This week at IO, Google showed off how Gemini is now baked into the Chrome browser. That's suddenly more than 1 billion users who will see Gemini every day, and they could end up using this chatbot, rather than ChatGPT.
• Then there's the big kahuna: Google Search. The company announced several ways that its new AI technology is being weaved into Search. There's a new AI Mode that launched across the US on Tuesday. Suddenly, roughly 250 million people will see — and probably use — AI Mode regularly. Because it's baked into the top of the Search page prominently. There are about 1.5 billion daily users of Google Search.

This is the type of distribution power that startups can only dream of. If OpenAI really wants to compete with tech giants, it badly needs Jony Ive's new gadgets — and a host of other distribution.

Distribution = data = better AI

Why is this so important? Well, AI products only really get better when a lot of people use them regularly. Google cofounder Larry Page used to call this the toothbrush test: If your product isn't used twice a day, forget about it.

If that happens, AI companies can collect mountains of data on how users are behaving. That information informs product updates. But in the generative AI era, this data is also incredibly valuable for developing new AI models and related products. This data can be pumped into training new models, with certain user permissions. It can also be used for fine-tuning and other AI development techniques.

The more you have, the better. Again, this data feedback loop only works if you have distribution—massive distribution.