Breaking! OpenAI CEO Altman personally admits: the past year was all my fault—will the AI bubble be about to burst?

Buddy, hold on tight. Today we’re talking about something explosive—OpenAI’s Sam Altman has actually publicly admitted that he messed up.

Here’s what he said: The past 12 months weren’t our best year—it was mostly my responsibility. But don’t panic—he added that the next 12 months will be the best year in history.

He said this on July 17 on X, addressing everyone directly. His exact words were: “The past 12 months were not our best year ever; that is mainly on me, but we are about to enter the best year to date.” He also added some broader lesson: AI’s mission is to give people freedom and wealth, not to use fear to force users to choose you.

Coincidentally, right before he posted this, a long-time bearish commentator named Ed Zitron threw a bomb. He said the real AI bubble is actually the OpenAI bubble. This guy compared OpenAI to a “systemically important institution.” If it collapses, the impact would be as severe as Lehman Brothers—data centers, AI infrastructure, global tech stocks—all would be repriced accordingly.

Zitron’s sights were set on three targets: First, inference costs are still too high, and the more users you have, the more likely you are to lose money; Second, capital expenditures are expanding wildly, but cash flow can’t keep up—the data center projects can take many years to break even; Third, in the next few years, OpenAI will still need external financing—if money gets tight, the model won’t hold.

Of course, OpenAI ignored these claims. But the debate over AI investment returns in the market has already burned to the surface.

After Altman’s admission of fault, netizens split into two camps. Some people think he was honest, but more are more concerned about the product. Jeff said: “If it’s just talk, users won’t feel it; but if it’s better workflows and lower service costs, users will feel it.” Another user @onecloudtech强调: “Trust is built through product releases, one by one.”

The current situation is that U.S. tech companies are launching an unprecedented wave of data center construction. Microsoft, Google, Meta, Amazon—money is being poured in without blinking. Companies like Oracle and CoreWeave are relying on long-term leases and project financing to scramble for compute capacity. Zitron warns that once demand from core customers like OpenAI fails to meet expectations, or when capital markets re-price returns, companies like Oracle and CoreWeave—whose valuations are supported by AI demand expectations—will be the first to get cut.

He also pulled in Anthropic and SoftBank. Anthropic’s route differs from OpenAI’s, but the “burn money” logic is the same. SoftBank is betting on the entire AI infrastructure—during valuation adjustment cycles, the pile of assets it holds will be turned upside down.

But don’t jump to conclusions yet. Howard Marks, the head of Oak Capital, said he was initially skeptical that AI is a bubble, but the more he thinks about it, the more he believes it has long-term value. Modern AI’s inference and context understanding capabilities are not comparable to the speculative bubbles of the past. He likened AI to general-purpose technology platforms like the internet and electrification.

Academic views are more neutral: technological progress is real, but it’s also true that localized valuations are getting overheated and capital expenditures are getting ahead of themselves. It’s more like “a technological revolution layered with localized bubbles,” not pure speculation.

For us people who talk with data, the key metrics that need watching are changing: enterprise AI revenue growth, product conversion/payment rates, the speed at which inference costs decline, data center utilization rates, and the investment return cycle. Whether Altman’s commitment can deliver across these dimensions is the key to whether the valuation logic behind AI trading can really hold up.

So don’t just listen to what he says. Wait for the product, wait for the earnings reports, wait for the inference cost curve. Whether it’s a bubble or not—numbers will tell.

Follow me: Get more real-time analysis and insights from the crypto market!

#PreIPOs第二期OpenAI认购 #GateDEX fully integrated with RobinhoodChain #TSMC Q2 net profit surges 77.4% $BTC $ETH $SOL