I haven’t written much lately. Much of the interesting news has been deep tech AI that I’m not an expert on, and I’m conflicted about covering a large swath of companies who partner with me through my gig at Google. This is an AI topic but more of a product/business question than a tech question
OpenAI App Store
From the Information, OpenAI is contemplating an app store of sorts to enable the buying and selling of AI models. I’m very confused about this idea, even if on the surface it sounds plausible enough. I will concede we lack full details here that may change the analysis.
Smartphones are effectively a duopoly of Google and Apple. That isn’t for lack of trying, many companies have given it a go over the last decade. Even Amazon and Microsoft have been utterly humbled here. One major driver is app stores. The iOS and Android ecosystems have accrued a decade of development and advantage. Customers assume the existence of their favorite apps. Developers assume the existence of massive, global user bases. Any platform starting from scratch needs to solve a massive cold-start problem.
While this is superficially a reason why an App Store strategy could work elsewhere, phones are very different than other situations.
Value accrues to the phone by becoming a better device. There may be some carryover in the usage of OpenAI’s services. Will developers share data to help make GPT5 better? I doubt it. If this is just a marketplace that doesn’t improve GPT, it doesn’t seem strategic.
OpenAI would in many respects be in direct competition with their customers here. Apple has a very uneasy relationship with many developers, and this would be even worse.
There seems to be a tension here. If training a model is hard, its a strategic asset to be guarded. If its easy to train/tune a model, there will be little sustainable value here.
Big companies and open source seem to be the two poles of the AI landscape. My model for B2B marketplaces is for a small-ish-sized company more than either.
The bigger issue to me is that a multi-modal world seems to be devastating to the value of OpenAI. In fact, this is already a risk with open source. They are pouring gasoline on that fire with this if it succeeds. OpenAI’s GPT4 would be one of many models, limiting their ability to build an API/service business powering other companies. ChatGPT would be competing not on differentiation of better LLM, but on scale and ability to accrue users. That is a game that Google, Meta, and Microsoft have tremendous advantages playing.
Ben Thompson writes often about the different types of companies. This seems like potentially a third company within OpenAI. ChatGPT is a consumer-facing entity trying to reach scale before Google. Their API business is attempting to be THE model of choice for developers. This marketplace concept seems to be a third-way hedge that isn’t aligned to either.