There's some nasty circular financing going on between chipmakers, cloud providers and AI labs, and it makes revenue look bigger than it really is. Building a data center is much harder and slower than a lot of builders have promised, so there's probably a huge amount of GPUs sitting in storage, waiting to be installed. And there's a huge amount of debt in the market, some of it off the books, that can blow up if those data centers aren't built on time.
But that doesn't mean AI is hype. It's extremely useful, even with some hallucinations, and we are just getting started.
If the financing blows up, some neocloud companies will probably go bankrupt. That's great for Google: they have a lot of cash, they can buy them cheap, and there's less competition. Same if OpenAI and Anthropic collapse under their commitments, especially with free Chinese open-source models making it hard to charge much. Google wins from that too. Less competition.
Of course, if this blows up it will hurt Google too, and the whole market will take a hit. **But if we are looking at the long term, Google is in a much better position than most.**
Google doesn't need the best model to win. As long as they don't go bankrupt, and nobody builds a real AGI that makes search obsolete, they can make money from AI through their huge distribution platforms and by selling compute, which people need no matter which model they use.
The only question left is whether $4.1 trillion is a good price. Apparently Buffett thinks so.