Quick math says average American owns/uses ~60x more battery capacity than they did 20 years ago. I bet the same happens for AI compute - huge growth in use per capita over the coming years... cnbc.com/video/2026/07/02/we…

"We spent 15 years with the theory that software is eating the world," says Gigascale Capital Founder @schrep. "I think AI is eating software. And what's left is competition in the physical infrastructure." cnb.cx/3RftQUT

We don't need 4,000 companies. We need ~20-40 — the @FormEnergyInc, the @heronpower, @CFS_energy, @_panthalassa, @solcoatech — and the whole physical economy gets rebuilt around them. Full episode: podcasts.apple.com/us/podcas…

Fusion: the triple product has improved faster than Moore's Law for decades. It's incremental, incremental, incremental — then whoa, the thing's running. @CFS_energy's magnet could levitate an aircraft carrier.

I got a lot of feedback during the fundraise: "this is great, can you just say data centers and not say climate?" No. Cheaper, better, faster — that happens to also be cleaner. I'm not going to whisper it.

Solar is the cheapest electron. Batteries are the cheapest peak shave. My product keeps getting cheaper and yours doesn't — at some point I cross, and I win. I think we've crossed.

The grid one that breaks people's brains: we build networks to average load, with buffers. We built power to peak. The grid's buffer exists now. It's called batteries. Enormous capacity gets unlocked when we build to averages.

Vertical integration isn't ideology, it's information. You can't solve a supply chain on a spreadsheet — if you are just on the internet you are just like everyone else. In the supply chain, I'm finding the copper problem 6 months before the WSJ writes about it.

You don't have to invent everything. Heron Power's solid-state transformers use power electronics already shipping tens of millions of units in EVs. Borrow the part that ships at scale. Own the innovation that doesn't.

This is why we're talking about ~30-50% margin companies, not 3–5%. Concentrate billions of R&D into a product you can make a lot of and sell a lot of. That flywheel hires the world's best talent to build the next one. That's how Nvidia works.

Go spec a power transformer today. Hire an engineer, write specs, design doc, wait forever. A custom bespoke wedding cake every single time. @heronpower is the Costco sheet cake equivalent for transformers. One cake tons of customers.

Everything in your life that got cheaper — TVs, computers, phones — is mass manufactured. Everything that got more expensive is custom built. Every time.

The thesis is simple: back technologies t–2 years from commercial competitiveness. Meaning they win on features and economics — not on vibes and hope.

Did a long chat with @yayitsrob on Shift Key about the Gigascale thesis, hard tech, and why the "US can't build anything" take is wrong. A few things I believe: 🧵

This was very fun - worth a listen!

Well it ended up being ~32,000 people and 105 events, by far the biggest yet. Give me 10 years and I'll put on a worthy successor to the 1915 San Francisco Worlds Fair

6/ Most leaders describe their culture's strengths. The ones who actually understand it can tell you what it costs. Thanks to @molly_g for the conversation on the WorkLife podcast: youtube.com/watch?v=9cwt_rMX…

5/ That moment probably saved us hours. But more than that, it proved the culture was real. Psychological safety isn't a soft concept. It's how you win, because no one is sitting on the fix, too afraid to speak up.

4/ One story stays with me. Major site outage. Millions of dollars a minute, the whole team scrambling. Mid-chaos, an engineer linked to a diff and said: "I think it might be this." It was their own change. That only happens when you already know what comes next isn't blame. It's "thank you, you just got us back up faster."

3/ But here's what actually made it work: when something broke, we didn't ask "who did this?" We asked "why was this even possible, and how do we make it impossible going forward?" That's a learning culture, not a blame culture.