Rory Sutherland

And they allocated more of their GPUs to internal product development and therefore were able to sell less of them in Azure.

And they basically explicitly said,

If we'd used more of our GPUs in Azure, we could have made that extra 1% of growth, so we'd have quote-unquote been fine.

If it wasn't for the 1%.

Very fundamentally, what you're seeing is, we said this, in the world of AI, they haven't, as we said, the corporate development team at Microsoft have executed brilliantly.

They own one third of open AI.

The product team at Microsoft have not exited brilliantly.

They don't have any compelling AI products that they own, either an LLM, which Google has, or even compelling apps.

They just don't have it.

So they're reduced to being... So a little bit of their quote unquote buzz was they were getting some perceived lift because they were a vendor to OpenAI and selling them Azure.

But now the market is kind of soured on that because they're saying, A, maybe it's a low margin business anyway.

It's not, I prefer to own the model than provide the compute.

And then B, if you sell it to OpenAI, is that really money good?

So I think it's kind of what happens in these markets when they turn against, I mean, there's an element of narrative and momentum that you think shouldn't be there in the efficient market hypothesis, but it is there because we're human.

And that's one of the things I've learned.

I used to be a total efficient market hypothesis guy.

And I think in the long run, I am.

But in the short one, narrative shapes everything.

And there's been two years of Microsoft narrative being really strong.

They all have open AI.