At the heart of the AI discussion is one issue and issue alone. Are the hyperscalers spending too much? The Economist provides a nice guide, basically showing that we are twice the dot com boom in half the time!
Goldman Sachs also provides a comparison, which puts total amount of capex on a par with the dot com boom.
So basically we have huge amount of capex - more than any other boom on some measures. Adding to this bearish set up - the Capex is now debt financed
And you are seeing the CDS respond to debt issuance.
And then to cap it all off, the US AI models are getting undercut by Chinese models. If you are bearishly minded, then this is a slam dunk short case.
For me, I understood all this, but I saw the capex spending as defensive. With Elon Musk joining the AI party - Google, Microsoft and even Meta could feel threatened, and so would all need to spend aggressively to defend their turf. Or to put it another way, Volkswagon should have massively raised R&D on EVs when Tesla was launched, and now its too late. The big tech was not going to make the same mistake. From this line of thinking, the bear signal would be someone giving up. But I was reading the Microsoft results, which was excellent, and could not stunned by the future contractual obligations they reported. Well over USD 700bn.
The most recent quarter also saw their cloud business become their largest single business line - with over 30% year on year growth. If I saw this type of “contractual obligation” growth in my own businesses, I would also be spending big.
So the question is can Anthropic and OpenAI actually meet the the US 1 trillion backlog they have built up with the hyperscalers? These two companies alone make up the commitment to the hyperscalers.
And here is where it gets hard. Microsoft, Oracle, Google and Amazon must all know that they are extremely reliant on OpenAI and Anthropic. But they also must have their own visibility of the demand for AI related areas. And they should also have good visibility on the capability of Chinese AI, and yet they still spend. So to be bearish you need to take a view that the tech giants have missed something, or have gotten something wrong. And this does happen. The GFC was basically caused by a belief that AIG could not go bust (the investment banks bought CDS from AIG to hedge their MBS business risk. When AIG failed, they were no longer hedged.).
What could I see that the big tech companies not see? I don’t know. I think the biggest risk could potentially Chinese AI completely upends the AI system. But Chinese cloud pricing has been cheaper that the West for years without slowing cloud growth. I would also think this was more likely if Nvidia GPUs did not still trade at a premium in China. That China still wants and needs GPUs means that their AI is still built along similar lines, and not a completely different infrastructure.
Having lived through big busts in the 1990s and 2000s, I understand the fear that investors have. But I think to really commit to a bearish trade you need to find something that the market has wrong. My personal view is that markets have interest rates expectations wrong - and I have good reasons to think that. But on AI, what I see is that OpenAI and Anthropic have huge obligations to the hyperscalers, which they seem willing to invest on the back off. Given their greater transparency, I find that hard to call that bearish. However, if they are basing their investments on the idea that US 10 year treasury yield is going to stay below 5% - then they have a problem. But that is really not tech specific.
