On Sunday, OpenAI chief executive Sam Altman sat down with podcaster David Senra and said something you almost never hear from the person running the most valuable startup in the world: the AI industry has been too ambitious, the economy has more inertia than he expected, and the technology has not yet had its iPhone moment. He even called the missing moment "mostly a product failure."

That is not the usual tone from a man whose company is valued in the hundreds of billions. It is also not a throwaway line. Altman's comments land at a strange moment for the industry, when the two loudest signals coming out of tech point in opposite directions. Adoption is moving slower than anyone predicted, while the demand for compute has gone, in Jensen Huang's word, parabolic. Both things can be true. Understanding why is the most useful thing you can do with your morning coffee.

What the Palm Pilot Era Actually Means

Altman's reference point is personal. He carried a Palm Treo around 2003, years before Apple shipped the iPhone. "Most of what the iPhone needed already existed," he said. "What was missing was the product idea."

That is the argument in miniature. The smartphone's components were all on the shelf: touchscreens, mobile processors, wireless data, a mail client. What Apple did in 2007 was not invent hardware. It reimagined how a normal person interfaces with a computer, then spent the next decade watching every other company in the world copy the interface.

Altman's claim is that AI is in the Palm Treo phase right now. The models can write code, reason through problems, and operate software. The interfaces people actually use to get work done have barely changed. Most of us still spend the day in a browser, a messaging app, and a spreadsheet. "We have all of the technological pieces, but we have not had the iPhone moment of completely changing how someone interfaces with technology," he said.

The Blockbuster Lesson Nobody Wants to Hear

Altman's second point is the one that should worry founders. He was responding to Shopify chief executive Tobi Lutke, who told the same podcast that 2026 would be the year every business is up for grabs, and that he intends to rebuild Shopify from scratch at night to find the AI-native version.

Altman pushed back: "I'm not sure if 2026 will be the year that every business feels up for grabs. I might disagree with him a little bit there." The reason is habit. "The economy just has so much inertia. People keep doing the same things they're doing. They keep buying from the same company."

He reached for Blockbuster, which he drove past on the way to school. People kept renting from Blockbuster long after Netflix started mailing DVDs. Changing behavior, he said, is much harder than the tech nerds realize. That single sentence explains more about enterprise software sales than any earnings call this year.

The Paradox: Slow Adoption, Parabolic Demand

Here is where the story gets interesting for anyone running infrastructure. Altman's caution does not look anything like a demand problem. Nvidia guided second-quarter revenue to 91 billion dollars, up from 81.6 billion in the first quarter, and said it has 119 billion dollars in supply-related commitments. Jensen Huang's explanation on the earnings call was blunt: "Demand has gone parabolic. The reason is simple. Agentic AI has arrived."

TSMC says demand is strong through 2029 and 2030, and raised its 2026 capital budget to between 60 and 64 billion dollars. Its packaging capacity is so tight it is limiting customer growth. Meanwhile Altman revealed that OpenAI shelved Sora, its flagship video generator, "because it was taking up too much compute." He called compute the biggest bottleneck and said it is where he spends most of his time.

Read those three facts together. Enterprises are slow to change their habits. Frontier labs are buying every GPU they can source anyway. The capacity gets consumed either way, because the model builders are competing for every available watt. That is why the picks-and-shovels trade has worked while application revenue has disappointed.

The Search for the iPhone Itself

Altman's answer to his own critique is a pivot, not a retreat. OpenAI, he said, wants to be a platform company rather than a product company, with one interface and one API. It already merged ChatGPT and Codex into a single product line. It will not try to compete with its own customers. He wants OpenAI to be the layer underneath, the way the App Store sat underneath the apps that made the iPhone a phenomenon.

And then there is the hardware. Reports have circled for months that Altman and former Apple design chief Jony Ive are working on a device meant to be the interface that finally changes how people touch AI. Nothing has shipped, and Altman said nothing about it on the podcast. But the logic of his own argument points straight at it: if the pieces exist and the product idea is missing, the company that finds the product idea owns the next decade.

What This Means: The Window Is Still Open

For people actually building businesses, Altman's honesty is better news than his previous optimism. If the economy really does move slower than the models, then the disruption is not finished on day one. It is spread over years, which means there is still time to build.

Altman predicted "the greatest boom in people starting smaller businesses that we have ever seen." The field has not built enough products to accelerate it, he said, and that is a gap you can still walk through. Take that with the usual grain of salt, but the direction is real.

The takeaway for operators is simple. The models keep getting cheaper to run and more capable, while the interfaces around them are still terrible. That gap is the opportunity. The people who figure out how to make AI fit into how humans already work will be the ones who build the App Store moment, not the ones who wait for it.

What Comes Next

Altman named two risks that sit in tension with each other: losing control as AI grows more powerful, and too much power concentrating in one company or one model. He mocked the safety pitch that offers a cure for cancer in exchange for autonomy. "Dear peasants," he said, "we will bequeath upon you these gifts." It is a terrible sales pitch, and he knows it.

He also said the smartest people on Earth are about as smart as the smartest models, and that is about to flip. OpenAI is already rewriting its safety rules and has slowed a model called Astra. Whatever comes next will be decided by whoever builds the product that finally changes how people interface with technology. The pieces are all on the shelf, the same way they were in 2003. What is missing, still, is the product idea.

— Allan Ali, Sylt.ing