OpenAI: The Side Quest Problem
TL;DR: OpenAI built the most recognised AI product in the world. Then spent 2025 launching video, hardware, a browser, shopping, and ads in what looked like an ecosystem play but felt more like building to be seen building. The Sora shutdown isn't just a product failure. It's what happens when investor signalling starts driving product decisions and nobody asks whether the parts actually add up to something.
If this resonated, the newsletter is where I keep working through these patterns properly. And if you know someone navigating the AI platform landscape right now, feel free to share it their way. Thank you :)
I’ve been watching the OpenAI product story unfold with genuine interest over the past twelve months. Not because of the hype cycle. Because of what it reveals about a pattern I recognise, that moment a company stops building for users and starts building for the story it needs to tell.
Sora’s shutdown on March 24 was the most visible sign yet that something has gone wrong. But Sora itself isn’t the interesting part.
What the numbers actually show
Sora launched as a social video app in late 2025 and did genuinely impressive things early. Downloads peaked at around 3.3 million in November, according to mobile intelligence firm Appfigures. It hit the top of the App Store due to real, early interest.
Then the numbers started moving the wrong way. By February 2026, monthly downloads had fallen to just over 1.1 million. A 66% decline in three months. Monthly active users peaked in December and kept falling. KeyBanc Capital Markets analyst Justin Patterson put it plainly in a research note: “Even with all of OpenAI’s resources, Sora could not attract and retain an engaged audience”.
As fun as it was to use and play around with (even frustrating at times), the numbers make that case on their own. OpenAI was estimated to be spending around $15 million a day to run the product, per Forbes reporting from late 2025. Lifetime in-app revenue before shutdown: approximately $2.1 million.
OpenAI also inked an arrangement with Disney in December to license over 200 characters, with a reported $1 billion investment. Whether this was a strategic content play to secure exclusivity or simply a move to manage copyright exposure, nobody really knows. Either way, three months later Sora was gone, the deal was gone, and according to subsequent reporting, it's unclear how much of that $1 billion investment, if any, was ever funded.
The ecosystem that wasn’t
Sora wasn’t a one-off experiment. It was part of a broader 2025-26 product sprint. They built video generation capability. A proprietary browser project. They began testing advertising placements inside ChatGPT in early 2026 (less of a user feature and more of a revenue play). Invested in a hardware device built with Jony Ive’s team, acquired for $6.4 billion, with a screenless form factor that remains unconfirmed and a launch window somewhere in the second half of this year. They also added shopping capabilities directly into ChatGPT, letting users browse and buy products from partners like Shopify and Walmart mid-conversation. The Wall Street Journal reported that current and former employees said the sprawl “made it difficult to identify the company’s core strategy”.
The intention, I think, was to tell an ecosystem story. Not just “we have a very good chatbot, powered by one of the world's leading LLMs”. But “we own multiple surfaces, we distribute AI across all of them, we’re the next Google or Apple”. That’s a very specific narrative, designed to read well in a pitch deck because investors understand ecosystems.
The problem is an ecosystem is only as good as the sum of its parts. And the parts need a collective purpose. Which, at least for now, has seemed to be missing in OpenAI’s strategy.
Apple’s ecosystem works because every product makes the others more valuable. The iPhone sells Macs. AirPods only work properly with an iPhone. The Apple watch connects seamlessly with multiple devices and iCloud ties the whole thing together. You can trace the logic from any one product back to the centre. What did Sora do for ChatGPT’s core value? What did the Atlas browser add that you couldn't already do by opening ChatGPT in the browser you already have? What does a screenless AI pin do that ChatGPT on a phone doesn’t, for a product that already has 900 million weekly active users on the device already in your pocket?
None of those questions had obvious answers. And I’m not sure they were asked.
Google can afford to experiment. OpenAI can’t.
Google can kill things because Google has a fundamental, long-term business model that isn’t going anywhere. Search and advertising has been printing money for twenty years. The experiments are funded by a machine that never stops working.
OpenAI doesn’t have that machine. It doesn’t have a settled answer yet to what its long-term business model actually is. And here’s the part that gets overlooked. Even ChatGPT, the product that's supposed to carry everything, is losing money. Analyst and leaked internal estimates suggest OpenAI lost on the order of $8 billion in 2025, with projections of roughly $14 billion in 2026 and cumulative cash burn that could exceed $100 billion by 2029, before the company expects to reach profitability.
So when you’re spending an estimated $15 million a day keeping a product alive, per Forbes reporting from late 2025, you’re not running a side experiment funded by a profitable core. You’re adding to losses that the core product isn’t covering either. When everything is losing money, nothing is low stakes.
Sam Altman previously described the 2025 approach internally as “betting on a series of start-ups” inside OpenAI, per the Wall Street Journal. That framing might work when you have limitless runway and a separate revenue engine funding the bets. When your runway is finite and your core product is also your funding justification, every bet needs a clearer answer to the question: what does this make better? And who will pay for it?
Building to be seen building
Here's where it gets more complicated than a product strategy failure. And I want to be clear that what follows is my read of the situation, not reported fact.
OpenAI started in 2015 as a non-profit with a stated mission to ensure AI benefits all of humanity. By 2019 it had created a “capped profit” structure to attract investment while technically maintaining the mission framing. By 2025 it had completed a full conversion to a for-profit public benefit corporation. An IPO is widely reported as a possibility as soon as late 2026, with recent private funding rounds implying valuations somewhere in the range of $700 to $800 billion.
And when the mission changes, everything that follows changes with it.
In a competitive market where Google is shipping AI features into everything, Meta is doing hardware, and every investor wants to know what your “platform play” is, there’s real pressure to match the field. Not because matching the field necessarily creates value. But because being seen to fall behind is its own kind of risk when you're heading toward a public listing.
So you ship. Not because you have a clear ecosystem play. Because, in my view, you can't afford to be seen as the one that didn't.
Fidji Simo, OpenAI’s CEO of Applications, told staff in a March all-hands meeting that the company “cannot miss this moment because we are distracted by side quests”. The Wall Street Journal reported she explicitly referenced a competitor as a “wake-up call”. The company that kept its focus narrow, built one thing really well for enterprise users, and didn’t chase the platform narrative has been steadily taking market share.
The pivot back to focus seems to be an acknowledged correction. And they only admitted it was a problem internally. There was no public statement made.
What comes next
To be fair, not everything from 2025 failed. OpenAI's enterprise segment was growing, with tens of millions of paying subscribers and a rapidly expanding business user base. That's a real number and a real signal. But it took a string of expensive failures alongside that signal to prompt the pivot. Nobody planned for it. It took both the wins and the failures to force the change.
OpenAI will probably survive this. 900 million weekly active users is not a small asset. The core model is still world class. The IPO, if it lands, buys them time.
But survival isn't the same as having a strategy. And from the outside at least, the pivot back to focus happening right now looks less like a strategic insight and more like a forced correction ahead of a public listing. The question is whether the decisions being made right now are being made for users or for the story that needs to land with investors before Q4.
Those aren’t the same thing. And the gap between them is where product strategies quietly fall apart. Especially if other players continue to deliver ecosystem plays that make sense.
Without a settled answer to what OpenAI actually is and what purpose it serves, the risk isn’t just extinction. It can be something more gradual. A company that finds a viable corner of the market and stays in it. And without the investment that comes from a coherent, profitable model, the core model doesn’t keep improving. It just stays as is. In a market moving this fast, staying as is is its own kind of slow decline.
I’m genuinely curious where others think this lands. Is the pivot to enterprise and coding the strategy OpenAI should have had from the start? Or is it just the least wrong answer available right now while the IPO clock runs? Thoughts?
~ Pete G