The case for backing up civilization on Mars as an AI and nuclear deterrent strategy.
“This is what it's like when a species prepares to depart for the stars. You don't depart for the stars under calm and orderly conditions, it's a fire in a mad house— this is what it's like when a species prepares to move on…” Terence McKenna
Something incredible is happening at the intersection of information and energy. Even as a teenager, I knew intuitively that I was put on this earth to first, build data networks and second, build liquid-nuclear reactors; Information and energy. I have spent the past decade studying both, building mesh networks, wireless hardware, designing solar arrays for fun. I lived in India, met fascinating people building the technology that would define the next hundred years and debated with the philosophers trying to protect us against this future going wrong. I believe the foundations are in place for civilization to leave this planet much sooner than we think, all driven by an existential, civilizational battle for AI.
Let me explain.
At TED 2025 in Vancouver on April 11th, 2025, Eric Schmidt (the first CEO of Google) took to the stage claiming “AI is Underhyped” (summary). He stated that OpenAI is hiring 10,000 engineers to make a system that can output the work of a million. Tech leaders today are now debating how quickly this will become practical. Microsoft expects AI to replace most software engineers in five years, Facebook thinks it will happen within six months. In the startup world, we are already seeing this happen today.
Where things get interesting, highlighted by Schmidt, is that this process is non-linear. Even if your competitor is just a few weeks ahead of you, these growth curves advance so rapidly that you will never be able to catch up. Your competitor, be it a company, a military or a country, will be so good at writing code and shipping matter that you will never be able to catch up.
Think about this for a second: in an economy and civilization driven by data and automation, staying technologically relevant becomes existential. Whole civilizations will lose relevance because they cannot keep up. An entire new world order will be defined in just a few years, most certainly in our lifetimes, as “AI goes exponential”.
Some call this artificial general intelligence (AGI), but it will be less about “AGI” and more about massive means of production. This race will likely not be defined by ships and kinetic battles, but be defined by who can build and power the chips. At this point, when human progress is defined purely as the race for information and energy, reality begins to take a very different shape.
For example, the idea of a thousand nuclear reactor gigawatt arrays are already being proposed:
This existential race for AI leaves countries with only a few options: You can try to steal, attack or infiltrate your competitors, this is being done now. You can “copy”, optimize and open source models (e.g. DeepSeek). What if none of these work? Your security is too good, your people cannot be bought, and your competitor is inching closer to AI that cannot be competed with.
Schmidt proposes one final solution that he claims is being discussed in the highest levels of governments around the world today: Pre-emptive nuclear strikes on datacenters.
Well, this sounds insane.
But if we think about it, it’s actually quite logical. Act now, or you will never be able to compete, your civilization will become irrelevant. In a world where it’s “My country first” these types of geopolitically logical attacks are what is keeping Eric Schmidt up at night.
Many of Silicon valleys elite are coming to a similar conclusion, that there is another solution to our existential AI induced crisis- one that sounds so insane it is not really being spoken about: Protecting civilization using space-time as a barrier and leave for Mars. We are secretly seeing Silicon Valley organize itself for the fundamental building blocks of a new AI driven civilization.
AI companies are investing heavily in the building blocks for this exponential race to the stars. For example:
-
ChatGPT: Sam Altman investing and taking Oklo public and investing in Helion Fusion.
-
xAI (Twitter): Just merged X and xAI. Currently controls the largest GPU cluster on earth, and majority ownership in SpaceX (currently the most advanced Mars program)
-
Google AI: Investments in Kairos, and other extensive (rumored) energy projects.
- Schmidt, who still holds billions in Google Stock, just bought a controlling interest in relativity space, a company that is building 3D printed rockets. Their stated interest is to create “An industrial base on Mars”
-
The Chinese Government: Focusing on an “optimization” and open source strategy for AI dominance. Rapidly building Thorium reactors at a pace 10X of that of the U.S, and is planning a manned Mars orbiter mission for 2033
We can see where the incentives are aligning, and the investments are moving towards software defined energy and space architectures. Whoever is first to win the AI race, with their civilizational architectures being software defined, will be able to move incredibly quickly. We will see progress at a pace that makes the 20th century look like slow motion.
Personally, using a team I have assembled over the past decade, we are building 3D printed nuclear reactors with a new venture called CalThermo. Our aim is to build a new type of software defined nuclear reactor designed to make stuff (Medical isotopes, chemicals, oxygen, etc) with near zero waste, using an old architecture at a scale never seen before.
The key here is software defined- CalThermo is designed to be (like every company should be now) an AI First company. As AI advances, we can plug it into reactor architecture and geometry directly. This becomes key to advance reality in a world where AI is racing to the top. Linking the fundamental building blocks of civilization to a software-defined architecture (in CalThermo’s and Relativity Space’s respective cases, 3D printing) allows progress to match AI advances.
This is why I believe things are going to happen so fast. Reactor and spacecraft development are now matching the speed of AI.
This pace will make us question what it means to be human. I personally believe there is no other way to build these new fundamental “deep” technologies. AI first, then let the software define everything else. Add safeguards to make sure everything runs safely.
When it comes to the incentive structure for going to Mars, my colleagues and I have written about these drivers. I thought it would be tax evasion which drove us to mars. I now believe this thesis to be incorrect:
The drivers for civilization to go to Mars will be defined by this existential demand for AI and Energy, and as a fundamental deterrence against pre-emptive strikes for civilizational AI supremacy. Moving your AI models to Mars makes mutually assured destruction (nuclear, biological or otherwise) much, much harder. In a world where China is building nuclear weapons and missiles in the desert, deterrence is all about making an attack incredibly expensive. It looks a bit like mars:
Being firmly anti-nuclear war ourselves (and to borrow from Cold War terminology), we believe this so-called Mutually Assured Destruction (MAD) calculus should be focused less on making a counterattack destructively expensive (viz., killing everyone), and more logistically expensive (moving your datacenters to another planet). The financial benefits of automation, unchecked exponential growth and no zoning laws will quickly pay for themselves in AI scale. Additionally, I find this more elegant than classical Cold War style MAD. (As a side note: We can argue that with AI, classical weapons of mass destruction may no longer be relevant, but that is for another article. Either way, AI drastically shifts this calculus).
From a strategic deterrence perspective, it is essential that we go to Mars and other places and build massive AI clusters. This route is far more likely to create MORE peace, maximizing the probability of continued, Western civilizational existence.
Again, while the idea sounds quite insane, a Mars base is being privately funded by companies like SpaceX at their Starfactory facility in Boca Chica, TX. They are currently ramping up their factory to produce one starship per week:
The pace and advances in automation will be astounding. We at CalThermo were able to 3D print a thorium breeder reactor architecture in just a few months for <$100,000 (unfueled). This will only accelerate.
The good news is that Mars has a lot of energy and “space” for such operations. Here is a rough concentration of thorium (the preferred source material for nuclear reactor architectures). While at general surface concentrations about 10X less than earth, we should assume that with robotics and advances in extraction we will get at least 10X better at mining “Rare Earths” very quickly:
There are even a few places on Mars with such high concentrations of Thorium and Uranium that Natural Thorium breeder reactors were detected on the surface. We propose going to get those materials, and use that to power a new civilization. Even fusion reactors (which will surely be aided in construction and scaled with new SciML models) will likely be powered by tritium and helium-3 bred first in the core of nuclear fission reactors. All of this, by majority will be used to drive AI.
This age, in particular the next five years, will be the most consequential for human existence in the last millennia. All of the building blocks for human colonization of other planets exist in some form or another. AI, in its quest for scale and we believe deterrence, are aligning all vectors very rapidly on a course for the stars.
We can’t wait to be a part of it.
Garrett Kinsman
Santa Cruz, CA, Earth.