The money is not spreading out. It is piling up. Venture investors put $159 billion into startups worldwide in the third quarter, according to figures Crunchbase released this week, and nearly two-thirds of it, around $102 billion, went to companies building artificial intelligence. Close to 6,000 startups were funded in all. A handful of them took most of the cash.
The quarter's two largest rounds were $5 billion each, raised by Databricks and by Safe Superintelligence, Ilya Sutskever's lab that has still shipped no product. Behind them sat a cluster of companies that raised more than $3 billion apiece: the cloud provider Crusoe, China's Moonshot AI, Europe's Mistral, the infrastructure firm Nscale, Elon Musk's Boring Company and the video-generation outfit Kling. Crunchbase counted a record number of billion-dollar rounds in a single quarter.
Concentration, not breadth
What stands out is not that AI is attracting money. That has been true for three years. It is how narrow the funnel has become. A quarter in which 64 percent of all venture capital flows to one category, and where the biggest cheques go to a short list of already-enormous companies, is not a broad-based boom. It is a small number of very large bets on firms that need staggering amounts of capital to buy chips and electricity.
The individual deals we have covered this autumn fit the pattern. DeepSeek's raise toward $15 billion, ElevenLabs doubling to $22 billion and AMD's $8.2 billion purchase of World Labs are not outliers. They are what the top of the market now looks like.
The question underneath
Numbers this large invite the obvious worry, and plenty of investors are voicing it: that a lot of this capital is chasing revenue that does not yet exist. The counter-argument is that the companies raising it are the ones with real usage and real contracts. Both can be true at once. What the third-quarter figures make clear is that the industry's fortunes now rest on a handful of names, and if any of the biggest stumbles, the fall will not be contained to them.
Commentarii · 0