energy is the problem worth solving

While reading Sapiens it occurred to me that we have grown as a modern global civilization so far on exploiting resources, and energy remains at the center point. My feeling is: given how we have progressed, the next breakthrough at civilization level would happen when we are able to harness more energy and manipulate it at our will. That directly affects our manufacturing, research, and AI capabilities. Every jump in what humanity could do was really a jump in how much energy we could harness and direct.

The recent US/Israel/Iran war makes it quite clear that we are still dependent on oil and natural resources that we deem will be over in the next couple of decades. We talk a lot about clean energy and sustainable sources of energy but it isn't deployed at scale, and I believe it would take us a good amount of time to figure out how we are going to switch the entire manufacturing industry. We are going to need a different set of skills and labour to manage that.

the number that reframes everything

173,000 terawatts of solar energy hit Earth continuously. Global human energy consumption is roughly 18 terawatts — about 0.01% of what arrives. The total solar energy hitting Earth in a single hour is roughly equal to what all of humanity consumed in the entire year 2017.

We are sitting inside an ocean of energy. The scarcity isn't the sun. The scarcity is our ability to capture, convert, and store what's already arriving.

Plants by all means convert it for their own use — which the whole ecosystem then uses — at somewhere between 0.1% and 2% efficiency under real field conditions. Solar panels commercially hit 15–26%. Advanced lab systems exceed 40%.

So on raw efficiency we've already beaten plants. And yet the world still runs on oil.

I wonder how much money is poured into research which is constantly trying to find new interdisciplinary ways to have better efficiency than plants. Where is the comparison between plants and artificial methods that we have? Because plants didn't win on efficiency — they won on something else entirely. They package energy into storable, portable, chemically dense fuel. Photosynthesis is inefficient at capturing light but extraordinarily good at making something you can carry around and use later. That's what we haven't figured out at scale.

what this means for AI right now

Data centers running large AI models are already one of the fastest growing energy consumers on the planet. Inference is expensive. Training is enormously expensive. The current trajectory of AI development assumes access to cheap, abundant, reliable electricity — and that assumption is going to get tested.

For cloud-based AI systems this is mostly a grid problem. Big, but not structurally new.

For physical AI it is a different problem entirely.

the physical AI problem specifically

Robots can't run on a cloud connection. A robot operating in the real world needs local energy — on-board, dense, reliable.

We got a LiPo battery for our LeKiwi mobile manipulator. It runs for an hour or two. Add more load — good cameras for running vision-language-action models, heavier actuators, more compute on the edge — and I genuinely wonder how long it would last. For research it's fine, you connect to a static power board. But the future is mobile robots doing autonomous meaningful work, and they are going to need better local energy supplies to do that.

Right now we are essentially importing the energy density assumptions of consumer electronics into robotics. That works in the lab. It breaks down when you want a robot to operate for hours, carry meaningful payloads, work in environments without charging infrastructure, or function in places like construction sites, disaster zones, or eventually other planets.

The tradeoff is always: either we spend our time optimizing for what we've got, or we ask and dig for more. I like the idea of maxing the usage of what we have — there is a lot of room to get better at energy efficiency in actuators, compute, and energy recovery. But if there is a lot of room for us to harness more energy, why not pursue both in parallel? Global charging stations are one answer. Better local energy sources are another. I don't think we've seriously asked what the right architecture looks like for physical AI systems operating at scale.

what are we actually doing about it

Are we exploiting our knowledge of biology, physics, chemistry and math to find novel ways? Are we too narrowly focused on getting a certain type of energy from some source in a certain way?

The sun is the largest source of it. How can we make best use of it? What is something we could do interplanetarily that would help us even more? What's stopping us from using the natural resources of other planets?

I'm not saying I have answers. I'm saying I don't see enough people asking them seriously and working across disciplines to do it.

what I'm trying to do about it

I'm trying to write a proper survey on this. The idea came from a conference call for papers — using AI as scientists, autonomous agentic workflows to do what no human team can realistically do: cover literature across multiple fields simultaneously at scale. A single agent gathered around 2,000 papers for a preliminary pass. Not all relevant, but with proper workflows — an overall plan, multiple agents spawned in parallel, each covering papers that fit in their context window, filtering and reporting back — you can actually produce a comprehensive survey across biology, chemistry, physics, and engineering simultaneously. Something as a human that's not possible. Agents can come up with gaps and ideas that you can then work on. That's the most compelling use case I can think of for agentic workflows right now.

The goal is to map the real bottlenecks — not by source (solar, wind, nuclear) but by constraint: storage, density, conversion, distribution — and see where the genuine gaps are between what biology figured out over three billion years and what engineering has built so far.

Energy is at the forefront of everything we do and we must think about it deeply. To move to the next phase of humanity we need better ways to harness energy — it directly influences the amount of production we're capable of and also opens doors for things we can only imagine theoretically.

I'll update this as I learn more.