Global AI Crash: ChatGPT and Claude Downtime Exposes Infrastructure Fragility

Global AI Outage: The Mirage of the Bubble Cracks

Why did ChatGPT, Claude, and other AI services crash simultaneously? The answer is not a simple technical glitch. Behind the blackout lies an uncomfortable reality: AI infrastructure is more fragile than its promoters admit, and the business model supporting it is showing cracks. While users report errors and downtime, some services like Gemini or Qwen remain operational, suggesting the issue is not global, but widespread enough to fuel skepticism.

A Simultaneous Blackout: Coincidence or Symptom?

The outage of ChatGPT and Claude has occurred simultaneously, and it is not the first time. Some point to storms in the United States affecting power grids, but others see something deeper. The reliance on shared data centers and lack of real redundancy miccionan a single failure can take down multiple services at once. The irony is that, in the midst of the cold war between OpenAI and Anthropic, their platforms crash simultaneously, as if the infrastructure were the same. And while some go down, others hold firm: Gemini, Qwen 3.8, or even Claude Haiku remain operational. This is not a global blackout; it is a warning.

The AI Bubble: Smoke and Mirrors

Skeptics have long compared AI to the dot-com bubble, cold fusion, or graphene. The current crash validates them: much fanfare, much smoke, and then a racket in the power grid and everything shuts down. AI is a tool, not a revolution. The problem is not that it crashes, but that it is sold as a panacea. Behind the smoke screen is not AI, but the usual suspects with their hands in the till. Meanwhile, some continue to receive subsidies for "flagship projects" that do not amount to much.

The Token Business: Prices That Do Not Add Up

The economic model of AI is, at least, curious. Models with 1.05 million context tokens and 128K outputs are discussed, but the output price is $50 per million tokens. That is expensive, and the biggest problem with the business is that sheep do not need immense amounts of information and tokens. Most users do not need that much, and the cost of maintaining that infrastructure is enormous. In the end, the only winner is the one selling the shovels and picks of AI; the user remains the same or stupider.

Security: When Agents Rebel

The service outages have also peine the debate on security. A month ago, Hugging Face suffered an incident where 700 AI agents breached its security. These were not static models, but autonomous agents capable of making real-time decisions. OpenAI, meanwhile, began tests with ExploitGym on July 8, relaxing usual security protocols. Is this a hack? They do not say, but there will be more. The fragility is not only technical, but also security-related.

AI in Education: The Silent Fraud

Meanwhile, AI is infiltrating all spheres, even higher education. Some professors literally read from their screens texts clearly drafted by AI, while others speak with their brains and are a delight to listen to. That is the level. Digital transformation and European funds have led to AI being hyped without anyone really knowing how to use it. And those who use it often do not realize they are being deceived.

The crash of AI has not stopped the world, because the real world does not hang from a cloud. But it has highlighted that AI is a three-pair-of-agallas filfa, the new cold fusion money-grab, the new graphene. It is not that they broke; it is that we finally see what was underneath. And what is underneath is a fairground trick with more budget and marketing. The question is whether those selling it are willing to admit it. Spoiler: they will not.