Cameron Gleeson

I hate that word, but it absolutely is.

And also if you look at the breadth of their customer base for that, that's also very reassuring.

But we can also sort of look at the hyperscalers.

If you look at Alphabet's results, so they obviously own Google and Gemini is their sort of core model.

The growth in what we call tokens, which you can think of a token as a unit of work for AI, grew 60% quarter on quarter just in their first quarter of this year.

And the Google Cloud business has revenue growth of in advance or slightly more than 60%.

So just huge growth to these core businesses.

But where we start to, you know, really, you know, get a bit more comfort is thinking about that monetization.

So not just growing revenue, but are we seeing profitability?

And the margins of that Google Cloud business went from below 10% to up to now 32% operating margin.

So that's evidence that people are willing to pay.

Like right now there is clearly some constrained supply of compute, but we're seeing that that

you know, massive growth in basically the consumption of tokens.

Yep.

Now, a lot of this goes back to a more sophisticated use case for AI.

Originally, you were in chat GPT, you were asking a prompt that might consume, say, in a conversation, maybe a thousand tokens.

Yep.

We then move to agentic AI, where you've got maybe a multi-step workflow, a number of variables, inputs, outputs.

And you might be, if you're calling an agent or if you're basically asking an agent to do work, you might be talking about 10,000 tokens.

Mm-hmm.