NVIDIA Didn’t Buy Groq for Speed

The headline makes it sound simple: NVIDIA is buying Groq for roughly $20B.

Most people stop there.

I don’t.

Because this deal isn’t about faster inference, cheaper chips, or “more AI.”
It’s about control across compute regimes — and that has real implications for how infrastructure gets priced.

The lazy takeaway is that Nvidia is doubling down on inference.

That misses the point.

The Groq acquisition rewards optionality, not raw power:

Nvidia isn’t betting on one future.
They’re making sure they’re relevant across all of them.

That’s a subtle but important shift.

Inference-first architectures like Groq work best when:

That is not the same problem being solved by 50–100MW+ hyperscale campuses.

This isn’t an either/or future.

It’s a fragmented compute future — and Nvidia just acknowledged it.

When the stack fragments:

In that world, the winners aren’t the operators with the most power.

They’re the ones with the most adaptable power.

That’s the lens I’m using.

Here’s the part that matters economically:

That flexibility has real value.
And right now, it’s underpriced.

Because optionality doesn’t show up cleanly in a spreadsheet — until the regime changes.

This is why I continue to prefer Cipher Mining (CIFR) & Iris Energy (IREN) over more rigid “AI data center” stories.

Not because others are bad businesses.

Because flexibility matters more than branding.

Mining-native operators already solved the hard problems:

That’s the foundation you need before you start swapping compute architectures.

The interconnection queue is a headline.

It is not the business.

What actually matters:

On that front, CIFR & IREN in particular have consistently executed on what they talk about — and tend to understate what they’re working toward.

That matters more than press releases.

“NVIDIA buying Groq” isn’t a victory lap.

It’s an admission that the future of compute won’t be clean, centralized, or uniform.

It will be fragmented.

And in a fragmented world:

That’s why I still like CIFR & IREN here.

Not because they’re chasing the next trend —
but because they’re positioned to survive all of them.

That’s the kind of asymmetry I care about.

Before it prints.

— Connor
Alpha Before It Prints

For anyone wondering how this framework translates into actual positioning:

I run two live portfolios that reflect two very different parts of my thinking.

The Black Sheep Base Case Portfolio is exactly what it sounds like — core positioning for how I expect the broader market to resolve when structure matters more than headlines.

The Alpha Framework Portfolio is different.

That’s where I take long-term swings on smaller companies I believe can materially outperform over full cycles — names that usually look wrong before they look obvious.

A few past examples from that framework:

No alerts.
No perfection.
A lot of patience.

That’s not a promise — it’s just context for how I think and how I size risk.

🔗 Connect with me:
✖️ Twitter | 💼 LinkedIn | 🚀 Subscribe

© 2025 Alpha Before It Prints
Unsubscribe