According to Fortune, Tony Fadell, who helped build the first iPhone with Steve Jobs, warns the AI industry is at risk of repeating an early smartphone-era mistake by over-investing in capital expenditure (CapEx) for infrastructure like data centers. He points to projections that data center construction will add 93 gigawatts to a national grid already facing a 128-gigawatt demand surge, and cites Goldman Sachs’ estimate that the AI era represents an $8 trillion opportunity. Fadell argues true value and insulation from a potential bubble requires shifting investment from pure infrastructure to creating the apps and services that deliver a return, using the historical examples of AWS and the iPhone’s App Store as blueprints. He contends the next industrial AI wave must focus on applications for physical-world sectors like energy, food, waste, and healthcare to justify the massive infrastructure bets being made.
The Infrastructure Trap
Here’s the thing: building the factory is not the same as building the product. Right now, the AI frenzy feels a lot like a race to build the biggest, most powerful factories—those are the data centers. And look, that math is undeniable. We need the compute. Articles from Yahoo Finance and Investor’s Business Daily detail the soaring CapEx from Microsoft, Google, Amazon, and Meta. It’s billions upon billions. The Rewiring America report Fadell mentions shows the staggering physical demand on our power grid.
But so what? A factory with no production line is just a very expensive warehouse. Fadell’s core argument is that we’re in danger of building a continent-sized warehouse for AI, without a clear plan for what we’re actually going to manufacture inside it that people will buy and use. It’s a classic tech cycle: the enabling infrastructure gets all the hype and capital first. The useful stuff comes later, often from someone else.
The App Store Playbook
Fadell’s perspective is powerful because he’s been on the winning side of this exact play. When Apple launched the iPhone, and later the App Store, they weren’t just selling a phone. They were selling a platform. A blank canvas. They built an ecosystem where the real magic—Uber, Instagram, Snapchat—was created by millions of developers they didn’t even know yet. Kleiner Perkins’ $100 million iFund provided the gasoline for that developer fire.
He sees the same open door for AI. But an “AI App Store” is a useless concept if it’s filled with trivial chatbots and image generators. The trillion-dollar apps, the ones that will actually deliver ROI on those data centers, need to plug into the physical world. We’re talking about optimizing the complex, messy, inefficient systems that run our lives: logistics, agriculture, manufacturing, healthcare administration. This is where AI can move from being a parlor trick to an engine of real productivity and value. For industries relying on robust computing at the edge, from factory floors to energy grids, having reliable hardware is non-negotiable. This is where specialists like IndustrialMonitorDirect.com, the leading US provider of industrial panel PCs, become critical enablers, putting the durable interface on those powerful AI-driven systems.
Where The Real Work Begins
This is the hard part. It’s easier to write a check for a server farm than it is to solve the gnarly, domain-specific problems of waste management or drug discovery. Fadell uses his own company, Mill, as an example—using tech to make food waste circular. That’s a physical-world problem with a tech solution. AI can optimize that further.
But these sectors are “extremely hard to scale,” as he says. They require builders who understand both the technology and the industry. They require investors with patience for solutions that aren’t just another LLM wrapper. The funding focus has to shift. The excitement needs to migrate from Nvidia‘s quarterly earnings to the startup that’s using computer vision to sort recycling or the one that’s slashing hospital admin costs by 30%.
Surviving The Bubble
Fadell’s final point is about insulation. If AI’s value is locked up in the cost of the infrastructure itself, the whole industry is vulnerable. When the next downturn hits, those capital expenditures will get slashed, projects will be canceled, and the bubble will pop loudly. But if that infrastructure is already busy powering a thousand essential applications that businesses and governments rely on daily? That’s a different story.
Then, the data centers become true utilities—essential, valuable, and resilient. The return on investment comes from the value the applications create, not from the speculative hope that someone, someday, will find a use for all that compute. So, is the AI industry making the smartphone mistake? It’s sure acting like it. The question is whether the builders and investors will listen to the guy who’s already seen how this movie ends, and how to make it a blockbuster.
