Every year the World Economic Forum asks thousands of leaders and experts what worries them most, and turns the answers into the Global Risks Report. The 2026 edition puts geoeconomic confrontation, meaning trade wars and sanctions between rival blocs, at the top of the list for this year and the next few years. But the report's own numbers point to a quieter and arguably bigger story: artificial intelligence is climbing the risk rankings faster than almost anything else ever tracked.
Ten years out, the risk that AI causes serious harm now sits in fifth place. Two years ago it barely registered, sitting around thirtieth. That is the steepest rise of any risk in the report's history, and it lines up with what is actually happening in markets right now.
Money pouring into AI has reached a scale where even the companies involved admit it looks strange. Nvidia has discussed backing OpenAI's data center buildout with a financing deal worth hundreds of billions of dollars, and analysts have started calling these arrangements circular, meaning the same small group of companies fund each other's spending and each other's revenue. It works until confidence dips, and then it can unwind fast.
AI is also colliding with everyday business risk in a more physical way. Data centers running AI already burn through as much electricity each year as some entire countries, and that demand is set to more than double by the end of the decade. On a hot day, a data center, a factory and a neighborhood's air conditioners are now drawing from the same stretched power grid. A heatwave that used to be a nuisance can become an outage that stops payments, halts a factory line and cuts AI tools off all at once.
Meanwhile, the systems that used to absorb these shocks are weaker than they used to be. Insurance companies estimate that the gap between what natural disasters cost and what insurance actually pays out passed 400 billion dollars last year. Government debt worldwide is close to matching the size of the entire global economy, a level not seen since right after World War Two. Both of those facts mean the same thing: when the next crisis hits, governments and insurers have less room to soften the blow, so more of the bill lands directly on the businesses affected.
This matters most for companies that quietly depend on the same handful of providers for critical work. Most businesses now run core operations through one of a small number of cloud computing companies, and increasingly through one of a small number of AI providers too. That concentration is efficient in normal times. It becomes dangerous the moment two problems hit at once, because there is no backup plan for a supplier that everyone else is also depending on.
The practical takeaway is not to fear AI or climate change individually. It is to ask a harder question: which two or three things, if they went wrong on the same day, would actually shut the business down. Few companies have mapped that out. Fewer still have decided in advance who gets to make the call to shut a system down, or spend emergency money, when that day comes.