EROI confusion and Humanity as a flea on a dog’s back

Artem Burachenok
6 min readSep 27, 2023

Recently I’ve been thinking about two aspects of EROI that I couldn’t wrap my head around:

  1. What’s the connection between the LCOE and EROI for different energy sources and how to reconcile the mediocre EROI numbers for some of the energy sources (solar, wind) with the great LCOE numbers?
  2. What’s the right way to think about the energy cliff — the situation where the EROI of the overall energy system becomes so low, that the majority of people have to get involved in energy generation activities (e.g. subsistence agriculture) as energy produced by the system (due to its overall low efficiency) is barely enough, or even insufficient, to support the people involved in its operations.

Below I’ll try to argue that LCOE and EROI are connected through the cost of human labor, and show how looking through this lens allows us to consider energy sources with subpar EROI and favorable LCOE; that the amount of human labor per one unit of net energy to humanity [1] is the critical variable defining the energy cliff; that there is a case for a low EROI system that nevertheless produces very affordable energy for humanity.

— — —

Let’s start with a thought experiment that I hope will show the difference between EROI and LCOE. Consider two identical oil pumps (equal EROI) with one difference — to operate the first one due to its higher level of automation you only need 2 people. To operate another you need 100. The energy consumed by automation running the first pump is similar to the energy consumption of human beings working at the second. So EROI will stay roughly equal for both. However, additional salaries paid to the extra 98 workers will necessarily increase the LCOE of the second energy generator (the electronic equipment of the first one isn’t free either, however it can’t match the monthly cost of human labor).

Generally, the lower the aggregate life-cycle human labor cost per unit of the generated net energy to humanity the lower the LCOE [2]. Think about the most remarkable energy generator — nature. For our hunter-gatherer ancestors picking up a fruit was collecting energy at the LCOE of zero. What’s the nature’s EROI? The hunter-gatherer can’t care less. What they are concerned with is the fact that no human labor was involved in producing the fruit, so they don’t have to pay for it.

Researchers of the EROI rightfully point out that historically, societies that relied on low EROI energy sources (subsistence farming) had quite low quality of life. Here, I believe it’s important to make a distinction between the EROI and the amount of labor per unit of net energy. The way I see it, the actual issue was that due to the state of technology, the amount of labor needed to produce the energy was high making energy costs quite high too. Back to the hunter-gatherer example above — the cost to them is zero because the fruit grows “by itself”. It would be quite high if they had to grow the tree from scratch, even if the EROI of the tree stayed the same.

To further demonstrate that the situation above is more about the labor than the EROI, consider the medieval society that by some miracle received a batch of smart, self-sustainable, self-replicating robots, that started to grow and collect food, produce tools, etc. While the underlying EROI of the system would actually decline (robots would use some of the grown food to power themselves) the rest of the food they produced would have been delivered to the population for LCOE = 0. (The total amount of food produced would have been limited — see below for the discussion of this issue).

Now let’s think through another thought experiment. Consider a mega-machine, generating energy. It has a very low EROI of 1.25 (against a decent EROI of 20). It generates 100TW of power and consumes 80TW for its lifecycle maintenance needs, leaving 20TW as net energy. Due to a high level of automation, it only employs 1,000 people who program and oversee its operations. The operations of the machine include its full lifecycle — from mining/recycling raw materials to producing spare parts, to performing maintenance, repairs, etc (save for a tiny number of external parts like sophisticated microprocessors the machine requires). For an external observer, the machine is a marvelous gigantic black box, that “hums”… and produces net energy of 20TW — roughly the current power consumption of humanity. The wages of the 1000 workers employed by the machine as well as the costs of a few external components are so tiny, that the cost of produced energy is virtually zero, despite the extremely low EROI of the mega-machine.

When imagining this machine I can’t help but think about a flee on the back of a dog. The flea can’t care less that the dog consumes 99.9% of the energy it generates for its own needs, as long as the flea has 0.1% of the energy available to it. Is it wasteful — it absolutely is. But so is the sun from the position of Earth’s nature, and so is the dog, from the position of the flea.

What are the potential limitations of such a machine? I see 3 types of limitations. First, the gross energy available should be enough to allow for the required net energy with the machine’s EROI. As an example, capturing 100TW of energy (per the example above) will require about ~2% of the earth’s surface with the current utility-scale hydro/wind/solar energy density, and way less if we use the energy density of fission. Second, since LCOE is defined by the lifecycle human labor costs [3], the lower the total amount of human labor the lower the LCOE. Here, it seems fair to expect continuing progress in AI to bring automation to the needed level. The third and the trickiest issue is the raw material requirements. This one will likely limit the total theoretical power of the machine however it should take us a while before we reach this ceiling.

What happens when the system runs into one of the limitations? Then, we are back to energy scarcity (and energy prices defined not by LCOE, but by demand elasticity) and increased involvement of the population in energy generation activities on one hand and to R&D efforts to solve the limitation on the other.

To summarize,

  • The key to affordable energy is the reduction of the amount of lifecycle labor per unit of net available energy.
  • While all things equal it is better to leverage high EROI energy sources, the lower EROI sources can also work as long as we can grow the total amount of net available energy while keeping (reducing) the amount of lifecycle labor per unit of net energy.
  • Automation replacing human labor in all processes of the energy generation lifecycle allows for lower LCOE.
  • R&D in higher energy-density energy sources and toward increasing EROI of the existing sources is still crucial to the increase of the total net available energy to humanity.

— — —

[1] Here, by net energy available to humanity I mean energy available after subtracting total energy needs connected with maintaining the energy source. E.g. for an energy source that produces a total of 100 W of energy, and uses 80 W to maintain itself (including maintaining physical structures, equipment, mining/recycling inputs, etc. — full lifecycle energy needs), will produce 20 W of net available energy.

[2] There are other factors, affecting LCOE, like lower (higher) input energy costs if we use legacy energy sources to build new types of energy generators and governmental subsidies — direct and indirect in the form of cheap credit. These factors while potentially material for specific generation technologies aren’t consequential to the nature of my general argument, so I put them aside.

[3] Lifecycle means all human labor involved in the operations of the energy machine, including labor applied to create the external parts it uses.