Azeem Azhar

The bear case requires lots of things to align.

It requires enterprise adoption to not to show up by the mid half of this year, that Microsoft doesn't kind of turn around, that's loss of momentum, that OpenAI really, really struggles, that there's some kind of deep debt default shock, and that there becomes a visible data center construction overhang.

I think if all of those things happen together, and I mean together, there might be a difficulty in the second half of this year.

But you just have to go back and look at the demand signals.

One of my favorites, I read this on Onyx, was a single developer who consumed 100 billion tokens in 39 days.

Or if you look at what's happening in memory, I mean, memory, HBM memory, which is what's needed for these AI factories is sold out for the whole year for 2026 from Hynix and from Micron.

Prices won't drop for a couple of years and the spot prices on GPUs

are rising, not falling.

This is not over supply.

And what I think we're going to start to see this year is we're going to see early frontier companies, big companies, deploying agentic systems.

And once they get it working, and it might take a few months, they'll put it in front of tens of thousands of employees.

And that's an enormous step change increase in demand.

To give you a sense of what that increase looks like,

I'll regularly run processes through some of my multi-agent systems that will be half a million tokens for something that I'm trying to push forward on the frontier.

So this is the cataclysmic mood I think has receded a little bit.

So always stay level-headed, read boomorbubble.ai and read the newsletter because that's our job.

So this is where we are.

Let's go reprise that.

My three anchors for the year.

The act of making has been transformed.