Every few weeks the same warning goes around: the AI boom is a bubble, and it is about to burst. It surges whenever a startup raises at a valuation that defies arithmetic, or when a gloomy study makes the rounds, and it has settled into the kind of claim people now repeat as settled fact. It deserves a closer look than that, because the honest answer is that nobody knows, and the evidence pulls in both directions.
Start with what feeds the fear, because some of it is real. Valuations are extraordinary. In a single recent stretch the coding company Cognition raised at 47 billion dollars, the enterprise startup Wonderful doubled to 5 billion dollars in six months, and global venture funding hit a record in the first half of the year, with AI taking the lion's share. When money moves that fast, some of it is always buying hype rather than a business.
The claim that lit the fuse
The single most quoted piece of evidence is a viral figure that 95 percent of corporate AI pilots are failing, drawn from an MIT study and picked up during a jittery stretch in the markets. It sounds devastating, and it is worth being precise about what it does and does not say. The study looked at pilots, the earliest and most disposable stage of any technology rollout, where a high failure rate is normal rather than alarming. It measured projects that had not yet produced a clear return, not projects that had collapsed. Read carefully, it is a story about how hard AI is to deploy well, which is not the same as a story about AI having no value.
That distinction matters, because the deployment problem is genuinely real. A Gartner survey this week found that only 22 percent of organizations have managed to scale AI beyond a single part of the business. If you want a bearish case, that is a stronger one than the viral statistic: not that AI does not work, but that most companies have not figured out how to make it pay yet.
What the bubble talk leaves out
The other side of the ledger rarely goes viral. Revenue at the largest AI labs is not hypothetical, it is growing quickly and being paid by real customers. Enterprise adoption is rising even where it is uneven. And the earlier fear that AI would instantly gut the workforce has, so far, not shown up cleanly in the employment data, which suggests a technology diffusing at the messy speed of real adoption rather than a fad about to vanish.
So is it a bubble? The most defensible reading is that both things can be true at once. There is almost certainly a layer of froth, particularly in valuations for companies with more narrative than revenue, and a correction in that layer would surprise no one. Underneath it sits a technology that people are genuinely using and paying for, on a release schedule that has not slowed down. Bubbles and durable technologies are not mutually exclusive. The railways were both. The internet was both. The claim to be skeptical of is not that some AI money will be lost, which is close to certain, but the confident prediction that the whole thing is about to pop on a schedule. That part is a mood, not a forecast, and it is worth keeping the two apart.
Commentarii · 0