From Hype to Reality Check

With GPT-5, OpenAI gives us a clear reality check, showing how big expectations often stumble against reality.

The launch of GPT-5

OpenAI introduced GPT-5. It is its most mature and useful model to date. However, it is not AGI, nor a massive leap-rather, an upgraded version of ChatGPT.

GPT-5 isn’t a single model, but a “team of experts” operating behind a smart router. When you ask a simple question, the fast “gpt-5-main” responds. By contrast, when the question is complex, the powerful “gpt-5-thinking” takes over.

In essence, the concept reflects NVIDIA’s vision for “heterogeneous agentic systems”. Instead of one titan doing everything, dozens of specialized models collaborate. Even so, the lack of transparency in routing creates new UX issues. Specifically, users don’t know which “sub-model” is answering each time. That explains why experiences vary so dramatically.

From hype to disappointment

Within hours of the launch, Reddit and Twitter lit up. Why? Because OpenAI quietly pulled GPT-4o, the favorite model of millions. Users felt they lost a friend, not a software update. In particular, GPT-4o felt companionable and encouraging. By comparison, GPT-5 responded like a “cold PhD expert.”

For that reason, Altman reacted quickly: in <24h he announced that Plus users ($20/month) would continue to have access to 4o. However, the 680M+ free users? No.

At the same time, many expected GPT-5 to be the first step toward AGI. They were disappointed when they realized that, in practice, it was simply a more robust GPT-4o. Essentially, the expectations-cultivated in part by Altman himself-snapped abruptly. They crashed into the reality of an upgrade without the “wow” effect that GPT-4 had two years ago.

The AI bubble question

It’s quite surreal to hear Sam Altman talk about an “overexcited market”. Especially since it comes right after the release of GPT-5-while, at the same time, he is preparing to spend trillions on data centers. Specifically, via the Stargate Project- a joint venture with SoftBank, Oracle, and MGX targeting $500 billion by 2029.

And he’s not alone. Microsoft is forecasting $120 billion in capex, Amazon is surpassing $100 billion, and Alphabet is reaching $85 billion. In total, more than $350 billion from Big Tech for AI infrastructure.

The paradox is blinding. The man warning of a bubble is, at the same time, the architect of the biggest AI race in history. Meanwhile, startups with a single PowerPoint raise millions. As a result, valuations reach “insane” levels, as he puts it. Déjà vu? The comparisons to the dot-com era practically write themselves.

The difference from the late 90s

Of course, there’s also the other side. Specifically, unlike the late 90s, today’s giants are funded by strong earnings and cash flow-not by debt. For example, OpenAI alone doubled its revenue in the first seven months of 2025. A $12 billion run rate, up from $6 billion at the start of the year.

But even so, when Alibaba’s Joe Tsai talks about “speculative data centre building without clear demand,” you understand something. Namely, that the fear of a bubble isn’t just theory.