Silicon Valley, August 25, 2025: Only weeks after shelling out billions to poach top talent from competitors such as OpenAI, Meta CEO Mark Zuckerberg has stopped all AI hiring, according to reports. The unexpected action comes amid increasing speculation that the artificial intelligence craze is a bubble — and a new study released by MIT is adding to the concern.
MIT Study: 95% of AI Projects Fail
A recent MIT study showed that 95% of projects that use AI do not achieve business objectives. The research surveyed:
300 public deployments of AI,
150 executive interviews,
350 employee surveys.
In spite of $30–40 billion of enterprise investment in generative AI, the vast majority of projects generated little or no quantifiable impact on revenue.
Most telling was that firms that tried to create their own in-house AI tools had a still higher failure rate than those that used third-party solutions.
Investor Fret and Hyperbole
The research’s conclusions shook investors who have been wagering big on AI to drive expansion. Even OpenAI CEO Sam Altman conceded:
“Are we in a phase where investors are overexcited about AI? In my opinion, yes.”
That recognition has only fueled speculation that the “AI hype train” could be coming to an end.
Not All Failures — There Are Success Stories
There are exceptions, even with dismal statistics. In 2023, enterprise software provider Ignite hired AI to replace 80% of its developers. Two years on, CEO Eric Vaughn reports the transition has brought 75% profit margins, no regrets.
This implies that the issue is not with the AI models per se but with the way organizations implement them. Most of the failures, MIT reports, were due to:
Unstable workflows,
Insufficient context,
Inadequate alignment with day-to-day activities.
In other words, the models can be “smart enough,” but the people implementing them are not.
The “AI Vibe Coding” Problem
An analogy in the report likened AI coding adoption to addiction:
The initial experience is exhilarating, as if overnight a billion-dollar piece of software could be constructed.
But after repeated dependence, firms are left with bugs, wasted investment, and disillusionment.
The cycle illustrates why developers won’t perhaps be as readily replaced as some predicted — at least yet.
Looking Ahead
Until now, AI has been both a promise and a threat. The hype persists, but the more failures accrue, the more skepticism grows. Giants such as Meta have put hiring on ice and scientists are cautioning against overenthusiasm, leaving the million-dollar question evident: Is AI the next dot-com bubble — or simply undergoing growing pains?
