Is AI a bubble? How Norwegian businesses can navigate the hype
Record-high investment, warnings about 'circular financing' and a recent report showing that 95 percent of corporate AI pilots deliver no measurable return have reignited the question: are we in an AI bubble? We look at the arguments for and against, and what a Norwegian business should do regardless of what happens on the markets.
'Are we in an AI bubble?' has become one of the most frequently asked questions in the technology world right now. Record-high investment in data centres, chips and models, combined with a growing number of warnings about 'circular financing', has led many to draw parallels to the dot-com bubble around the year 2000. At the same time, AI keeps being adopted in practice. For a Norwegian business, the question is less about whether the stock market is in a bubble, and more about how to get real value from the technology without buying into the hype uncritically.
What people mean by an 'AI bubble'
When someone talks about an AI bubble, they rarely mean that the technology does not work. They are referring to the possibility that valuations and investment at the top of the market have run ahead of actual earnings. The numbers are vast: according to an analysis reported by Investing.com, technology companies borrowed over 428 billion dollars in bonds in 2025, mainly for AI investment, and Goldman Sachs raised its estimate for the industry's 2026 capital spending from 465 to 527 billion dollars. The question is whether that money is ever earned back.
The arguments that it is a bubble
The most concrete concern is about how the investments are interconnected. Several commentators, among them CNBC's Jim Cramer, have warned that today's circular financing echoes the dot-com era: chip makers invest in model companies, which in turn buy chips and data-centre capacity, which is in turn financed by the same players. The system partly feeds on itself, which makes it vulnerable if one part fails.
The second concern is return on investment. A widely cited report from MIT found that around 95 percent of corporate generative AI pilots delivered no measurable return. If the big productivity gains fail to materialise, much of the justification for the enormous investment disappears, and it is precisely this gap between expectation and actual value that makes people uneasy.
The arguments against
On the other hand, there are good reasons to think this is not pure hot air. The capital pouring in meets real demand: the same analysis estimated that the AI sector received just over 200 billion dollars in capital in 2025, a 75 percent increase from the year before, and the use cases are expanding from text to image, audio and video. There is also an economic point that cheaper and more efficient AI does not necessarily reduce demand, but opens up new areas of use.
And even the dot-com comparison cuts both ways. When the bubble burst in 2000, enormous value was wiped out, but the internet did not disappear. On the contrary, it transformed all business in the decades that followed. A financial correction in the valuation of chips and data centres is not the same as the technology itself ceasing to be useful.
What Norwegian businesses should do regardless
The good thing about being an ordinary business rather than an investor is that you do not need to know whether the market is in a bubble to make good decisions. The point is to act in a way that leaves you well positioned whatever the outcome.
Start from a concrete problem, not from the technology. The 95 percent that see no return usually fail because they bolt a chatbot onto an unchanged process instead of solving a real bottleneck. Measure the value: if you cannot point to time saved or quality gained, it is a pilot, not an investment. Avoid tying yourself too tightly to a single vendor or a single model, so you can switch as prices and quality change. And keep control of your own data, which is the part of the value that endures no matter what happens to share prices.
Whether or not the top of the market is a bubble, the winners at the bottom will be those who used AI to solve actual problems, soberly, measurably, and without getting swept up in the hype.