Friday charts: Climbing the scale of civilizations

Nikolai Kardashev’s search for alien civilizations forced him to think big.

Instead of looking for radio signals or flying saucers, the Soviet cosmologist looked for markers of energy consumption. A sufficiently advanced civilization, he reasoned, would consume energy at a scale so massive we could spot it from Earth — like seeing the glow of New York City from space, but across interstellar distances.

He ranked these potential neighbors on the "Kardashev scale" — a measure of a civilization's technological advancement based on the amount of energy it’s capable of harnessing. 

Type I civilizations master all the energy available on their home planet: the sunlight hitting the surface, the kinetic energy of ocean waves, the geothermal heat from below.

Type II civilizations capture all the energy generated by their nearest star; most likely by surrounding it with a "Dyson swarm" — millions of solar collectors intercepting a star’s radiating energy before it dissipates uselessly into space. 

Type III civilizations do the same, but across entire galaxies, harvesting the energy of hundreds of billions of stars by means we can’t yet fathom.

Earth? We’re Type 0.

Humans are still busy burning fossilized plankton and spinning windmills — nothing to get us noticed from the other side of the galaxy, let alone the universe.

Now, however, we’re finally starting to think bigger: Demand for AI will soon push our energy infrastructure out into space.

This week, Elon Musk announced plans to merge xAI with SpaceX, in part to accelerate the launch of space-based data centers that will draw their energy directly from the Sun.

Solar-powered data centers in space are a "no-brainer," Musk says: "The lowest-cost place to put AI will be space."

Also this week, China — which is extremely good at producing energy on Earth — announced plans to launch "gigawatt-class space digital-intelligence infrastructure" — data centers in space — within five years.

This is Type I thinking: the first time in human history we're seriously planning energy infrastructure that operates at solar-system scale, rather than just what’s available to us at the surface.

All because of AI.

It may not feel like civilizational progress to you. So far, we’ve mostly been building data centers to make more cat videos and better Instagram ads.

But once you've got a swarm of solar farms in orbit…You’ve got a swarm of solar farms in orbit!!!

Kardashev gave us a scale for measuring civilizations, and AI is forcing us to climb it.

Let’s check the charts.

We’re gonna need a bigger power grid:

The "digital universe" is projected to use 4,000 TWh of power in the year 2050, with AI querying the fastest-growing factor. This is the equivalent of building 1,000 Hoover dams (AI tells me). Might be easier to build the data centers in space.

Mid-life crisis?

Goldman Sachs estimates that the US power grid will need $700 billion of investment between 2025 and 2030. US power grid infrastructure is, on average, 40 years old.

The energy race:

China appears to be much better than the US at adding power generation (although it’s not clear how hard the US has recently been trying). OpenAI calls this an "electron gap" that puts America’s lead in AI at risk. Either way, the energy race is likely to extend into space.

The startup gap:

A Stanford report finds that the US has a growing lead in the number of AI-related startups.

Power prices:

The price of electricity from solar has fallen 88% over the past 15 years. Now imagine how cheap it could get in orbit, where there are no clouds and no weather, it’s always daytime, and nothing is lost in transmission. 

We’re gonna need a bigger data center:

Nvidia estimates that an AI data center consumes as much as 100x more energy than a regular cloud data center.

Big Tech capex:

By 2030, Big Tech is expected to be spending $500 billion a year on capex. We will all have to click on a lot of ads to ensure they make a return on it.

Off the grid:

In the US, the wait time for a connection to the power grid can be years. Impatient hyperscalers have been building their power sources instead, with 48 GW of "behind-the-meter" (i.e., off-the-grid) data centers announced in 2025. That’s about 1/3rd of the total.

AI?

A steep decline in wages for new grads (adjusted for inflation) began before the release of ChatGPT in late 2022, but accelerated thereafter.

How productive will AI be?

A paper from Stanford shows that the share of GDP paid to "computers" has been in decline since the dotcom boom. If the same pattern holds with AI computers, that should (I think) be good news for productivity, employment, and living standards. 

Not so productive:

Americans spend more on OnlyFans than they do on OpenAI and The New York Times combined.

Let’s hope we find something more productive to do with our space-based data centers.

Our ranking among galactic civilizations depends on it.

Have a great weekend, Type III readers.

— Byron Gilliam

_{Brought to you by:}

Arkham is a crypto exchange and a blockchain analytics platform that lets you look inside the wallets of the best crypto traders — and then act on that information.

Arkham’s Intel Platform has a suite of features including real-time alerts, customisable dashboards, a transaction visualization tool, and advanced transaction filtering — all of which is accessible on all major blockchain networks, and completely free.

Arkham’s main product is the exchange, where users can express their trade ideas against the market.

Crypto’s premier institutional conference is back this March 24–26 in NYC.

Don’t miss SEC Chairman Paul S. Atkins’ keynote on Day 1.