SENECA'S CLIFF, THE LIMITS OF DEVELOPMENT, POLISH AI AND POLISH DIGITAL SOVEREIGNTY
FROM THIS ARTICLE YOU WILL LEARN
ToggleWhy is energy a limitation?
I was inspired to write this post by two recent books I read: "In Poland, or Everywhere: A Tale of the Decline and Future of the World" by Edwin Bendyk and "Why Nations Fail" by Daron Acemoglu and Robinson James, both of which I highly recommend.
The discussion about AI and Polish digital sovereignty ignores the hard laws of physics – including the ability to harvest energy. With the July White House document, "America's AI Action Plan," which focuses on, among other things, accelerating innovation and expanding the US AI infrastructure, it's worth juxtaposing the question: where does Poland stand in this race and what really limits us? Contrary to appearances, the bottleneck isn't necessarily the number of GPUs, but primarily quantity and quality (availability) of energythat we are able to produce and deliver in accordance with the laws of thermodynamics.
Additional information -> Americas-AI-Action-Plan.pdf
The thesis I wish to substantiate is that the material growth of countries (including Poland)—including AI development—depends on the availability of high-quality net energy. The key indicator here is EROEI (Energy Return on Energy Invested), which is the return on energy invested in its acquisition and delivery.
What is Seneca Cliff?
The most general approach to the barrier to growth, which I believe we are currently encountering, is the concept of the Seneca cliff – introduced in 2017 by Italian scientist Ugo Bardi – which refers to the increasing complexity of a system (civilization) and marks the point at which further growth in structural and functional complexity is impossible. The inescapable laws of physics and biology become barriers to development. According to Ugo Bardi, the approach to the Seneca cliff is manifested by the climate crisis, the collapse of ecosystems, and the rising cost of energy and strategic raw materials.
It's worth noting that the distance from the cliff varies across countries. This is influenced by the complexity of their social systems and the availability of energy and other resources. The Seneca cliff is echoed in the less general concept of the energy cliff, a point at which no more high-quality energy can be produced, necessary to maintain and develop the complex structures of a state.
How does EROEI relate to the complexity of society?
As EROEI declines, we approach the so-called "net energy cliff" – a non-linear threshold where a slight deterioration in energy return dramatically reduces net energy remaining for the rest of the economy (transport, health, education, research – today also data centers and AI).
What does EROEI mean in practice (Hall, Murphy, Tainter et al.)?
Based on research on “net energy”, heuristic thresholds can be adopted:
- ~3:1 – areas of biological survival (without “surplus” for complex institutions).
- ~5–7:1 – basic institutions and local organization are possible.
- ≥~10:1 – safe maintenance of the industrial economy (transport, education, health, digital infrastructure).
- ~15–20:1+ – significant space for innovation, research and shock absorption.
These values are heuristics resulting from many Hall/Murphy/Tainter papers and analyses that show significant increases in social benefits above the ~20:1 threshold.
According to Edwin Bendyk, EROEI in the late Polish People's Republic was only ~5:1 and this was the reason for the collapse of the economy of the People's Republic of Poland.
More information -> EROI of different fuels and the implications for society
What is EROEI and how has it changed over time?
EROEI is the ratio of energy extracted to energy put in. Historically, this has been high for fossil fuels, but is decreasing with the depletion of the "easiest" deposits and the increasing complexity of energy chains.
- Coal - EROEI analyses for coal-based electricity generation technologies (on the "electric" side) show values of ~30–35:1 with narrow system boundaries (without full system buffering). Differences in values result from the adopted methodology (e.g. whether we count only extraction/transport and the block itself, or also the network, reserves, storage, decommissioning of energy sources due to their consumption). However, the actual EROEI (energy available for use) may be much lower and in extreme cases in some countries even drop to the value of ~3,5:1.
Additional information -> Energy intensities, EROIs, and energy payback times of electricity generating power plants
- Oil – long-term studies for the USA confirm a decline in the EROEI of oil/gas from tens of 1 to ~5–15:1 (depending on segment: discovery vs. production).
Additional information -> A New Long Term Assessment of Energy Return on Investment (EROI) for US Oil and Gas Discovery and Production
Attention: EROEI calculated "at source" (raw material) is higher than EROEI "at the use stage" (after all conversions). The latter is more adequate for assessing how much energy is actually left "for the economy."
Additional information –> Energy Return on Investment of Major Energy Carriers: Review and Harmonization
If we were to refer to Poland, there is no uniform, official table EROEI of the entire economy (EROI_SOC) for individual EU countries. However, we know that countries with a large share hydro/atomic they usually achieve higher Systemic EROEI than those with a mix dominated by solid fuels. In Poland, a significant portion of the energy mix has historically been based on coal, which – after taking into account unit efficiency and system costs – reduces the "net energy buffer" compared to countries with a high share of high-EROEI and dispatchable sources.
More information -> EROI of different fuels and the implications for society
Can EROEI be determined for economies as a whole?
Since countries use a mix of energy sources, a generalization is used Social EROEI (EROI_SOC). In the literature for developed countries, ranges of the order of ~10–30:1, but the result strongly depends on system boundaries and the method of accounting for energy import/export, losses, power reserves or storage. The greater the share of sources with high EROEI and availability (hydroelectric power plants, nuclear power plants), the greater the "cushion" for maintaining complex institutions.
More information -> EROI of different fuels and the implications for society
Is Poland on the verge of an “energy cliff”?
For an economy with a growing demand for 24/7 available power (industry, digitalization, data centers) energy quality (EROEI, availability, conversion losses) becomes a critical factor. When the rate of growth net energy fails to keep pace with the need to maintain complexity, there is a growing risk of “squeezing” the space for growth and innovation – faster than the total MWh would suggest.
More information -> Energy Return on Investment of Major Energy Carriers: Review and Harmonization
What do AI and data centers need? A hunger for high-quality energy.
IEA (International Energy Agency) estimates that global data center electricity consumption will double to approximately 945 TWh in 2030. (nearly 3% of world consumption) – a AI is the main driver of this growth. Goldman Sachs forecasts that the demand for data center power will increase by up to 2030 ~165% vs 2023, and to 2027 Power demand may increase by ~50%. This demand applies available energy (when we train/inference), not only "green MWh" on an annual basis.
It is worth mentioning that Microsoft has signed an agreement with Constellation Energy, the owner of the Three Mile Island nuclear power plant, under which, starting in 2028 and for 20 years, all energy produced will be used exclusively to power Microsoft data centers.
More information -> Energy demand from AI
More information -> Business Insider PL – “AI vs. electricity…”, August 2025. AI vs. Electricity: Energy Bills Are Taken Hostage in Big Tech's War on Energy
More information -> CRN, September 2024 -> Microsoft to Revive Nuclear Power Plant to Power AI
What drives growth, competitiveness and innovation?
The real "fuel" for AI research, scaling and implementation is cheap, massive and available energy with high EROEIOnly at levels ~15–20:1 a clear space is created for the dynamic development of high-tech fields and a buffer against shocks (energy/climate/supply).
More information -> EROI of different fuels and the implications for society
What are the conclusions for Poland? Priorities for the decade.
Direction: rapid increase in the share of sources of high EROEI and availability - above all atom and hydro – supplemented by wind (on/offshore) and renewable energy sources linked to storage/DSR and grid modernization. The official schedule for today is the launch first nuclear unit in 2036 (Lubiatowo-Kopalino). It's late, but still within the window to power the wave of data centers and Industry 4.0 after 2030, provided that onshore wind is simultaneously unlocked, grid expansion is implemented, and dispatchable capacity is accelerated for the transition years.
More information -> Poland and US Sign Bridge Agreement for First Nuclear Power Plant
Comments:
- Measurement stage: EROEI is sometimes reported "at source", "at final energy" or "at useful stage" (after all conversions/losses). The closer to "useful", the more lower EROEI – and all the more relevant for economic planning.
More information -> Poland and US Sign Bridge Agreement for First Nuclear Power Plant
- System boundaries: Do we include storage and power reserves (important for variable sources), the grid, construction/decommissioning? This can shift the result by multiples. That's why I use ranges and trends, not one number.
More information -> Energy intensities, EROIs, and energy payback times of electricity generating power plants
More information -> A new energy-intensive industry is growing rapidly in Poland
If you want to talk about your IT system, please contact us.
- Paulina Śrama, phone: +48 605 586 507, email: paulina.srama@upgreat.com.pl
- Michalina Puzanów, phone: +48 667 752 750, email: michalina.puzanow@upgreat.com.pl
or if you prefer, use the form on our website -> Contact | Upgreat
Co-founder and CEO of UpGreat Computer Systems Sp. z o. o. The company has been supporting large and small enterprises from all over Poland in the field of information technology for 30 years. His main areas of interest are IT security, complex network solutions and group work systems. After hours, he savors the nuances of Stanisław Lem's prose.