James is live for our Tech Turning Point event. 

I'm writing this from backstage.

Tomorrow, we’re going to talk more about Tech’s Turning Point…

And where our gurus believe the puck is headed.

Today, however, let’s dive deeper into what James and I learned yesterday in San Jose at the GTC event.

First, the GTC conference has a track record.

In 2022, the H100—80 billion transistors, a generational leap from the A100.

In 2023, the H100 NVL chip, two GPUs joined, mandatory infrastructure for every serious data center on the planet.

In 2024, Blackwell: 208 billion transistors, AWS, Google Cloud, Azure, Oracle all falling in line.

In 2025, the pivot to physical AI, agentic AI, the Rubin Architecture—projected at 900x the original Hopper H100.

Each year the story compounds. Each year, the skeptics look smaller.

This year didn't break the pattern.

What Nvidia revealed makes most of today's AI investment story look like a rough draft.

Here's what most investors are still missing:

Jensen’s Law

Moore’s Law said that the number of transistors on a chip doubles roughly every two years, driving exponential gains in computing power at falling cost.

That’s dead.

“Jensen's Law” isn't about doubling chip speeds—it's about the entire AI factory.

The compute stack. The cooling stack. The power stack. The software stack sitting on top of all of it. Individual chips are table stakes.

The real money is in who controls the full infrastructure layer.

That reframe matters enormously for where the next 1,000% winners come from.

It's not CPUs. It's not even GPUs in isolation.

It's the systems that make the whole thing run—and the picks that aren't on anyone's radar yet.

(James and I identified one called Astera Labs—ALAB—for the memory bottleneck.)

Meanwhile, the AI agent story is arriving faster than the market is pricing it.

Agents that automate entire workflows.

Agents that analyze market data and execute trades. Agents that build software, run businesses, handle customer service—while you sleep.

The infrastructure enabling it—payment rails, orchestration layers, the picks-and-shovels of the agent economy—is still dramatically undervalued.

The Great Rotation

The software selloff everyone's panicking about? It's a rotation. The market is clumsily repricing what software is worth in an agent world.

Some of those names deserve to get hit.

Others are getting thrown out with the bathwater—and that's the opportunity.

There's also a space angle here that most people are ignoring entirely.

Not the moon, not Mars. Data centers.

The next frontier of compute infrastructure isn't in Arizona or Texas—it's orbital, solar-powered, and closer to reality than the skeptics want to admit. The companies building toward that future are early and small.

That combination rarely lasts.

GTC 2026 was a reminder of something that keeps proving itself true: the AI story isn't slowing down.

It's accelerating.

And every year, the gap widens between the investors who understand what's actually being built—and the ones still arguing about whether the whole thing is a bubble.

It's not a bubble. It's a buildout.

Get positioned accordingly.