About the need for Orbital Super-Computing.

Ernesto Eduardo Dobarganes
5 min readJan 12, 2023


The Cloud… needs to be in outer space.

This should be a very short piece that makes the case for taking most of our Computing and Super-Computing power to outer space, as a method to:

  • Minimize the contribution to Global Warming that results from catastrophic exponential CPU-related thermal emissions & energy consumption.
  • Reduce computation costs.

Before anyone starts screaming at their screen about how insignificant is the amount of heat that our computers aggregate to our planet’s thermal load (when compared, for example, to heat emitted by ICEs)… let me share the terms of the Computing-emitted heat equation and then project them into the future.

Render of a SpaceX Starlink satellite by Mark Garlick/Getty Images

Total Heat generated by Computing

According to some estimates, the total computing capacity of the world in 2021 is on the order of several hundred exaflops (1 exaflop = 10¹⁸ floating-point operations per second).

As for energy consumption and heat generation, a rough estimate is that the global data centers alone consume around 200 GW and generate around 450 GW of heat. However, it is important to note that other devices like PC, laptops, servers and mobile phones are also consuming energy and generating heat.

450 GW of heat alone, is equivalent to the output of 450 Nuclear Power Pants… or 5,965 Hiroshima-like nuclear detonations.

Moore’s Law

The Moore’s Law estimates/predicts the doubling of transistor density/count per chip or unit of area every 18 months.

Transistor count is one of the most important factors for establishing the computing power of a given chip or device. It is mostly the doubling of this number what enables parallel computing and the corresponding increase in software complexity given that the other components (Frequency and Efficiency) have actual physical constrains to which we are very close already.

Image of a CPU Waffer by PadovaTech

Please notice, that the exponential implications of Moore’s law, do not apply to other energy-intensive goods or services, like cars or transportation that have non-exponential growth profiles.

PPW: Performance per watt

Performance per watt (PPW) is a metric that is used to measure the energy efficiency of a computing system. It is a ratio of the performance of the system (usually measured in FLOPS) to the power consumption of the system (usually measured in Watts). The unit of PPW is typically FLOPS/W.

Asymptotical CPU Efficiency

In general, a higher PPW value indicates that a system is more energy efficient, because it is able to perform more computation while consuming less energy. This metric is particularly important in the context of High-Performance Computing (HPC) and Data Centers, where energy consumption can be a significant cost and where there is an ever-growing demand for more energy-efficient and powerful computing systems.

Overall, the aim of PPW optimization is to get most computation work done with the minimal energy expenditure, with this technology, advancements and innovations are continuously made to make computing more energy-efficient and sustainable.


Artificial Intelligence and Machine Learning are one of the most marvelous technological products of the human mind. Their working principles are so basic that they could easily be part of the working circuitry of the universe itself. Their plasticity and pattern-finding capabilities promise to conquer every area of human sense & decision-making, but…

Their training and use is a very energy-intense process.

The adoption of AI/ML seem to have, more than exponential, a polynomial profile. See the LensaAI app worldwide download chart below:

In the first week of December 2022, mobile photo and editing app Lensa AI recorded over 5.8 million downloads from users worldwide. The popular photo app uses AI to generate artistic renderings of user-input selfies and pictures. Source: Statista

The Problem: Exponential Power Consumption and Heat Emission within our Atmospheric boundaries.

Now that we are clear in the 4 basic concepts, all we have to do is factor them into the equation of future Computation Power demands:

Exponential Graphics * Polynomial AI adoption / Asymptotic Power Efficiency = Runaway Exponential CPU Heat Emission and Energy Demand.

42 years of Microprocessor trend data sourced from Github

Planetary Energy Budget

In a previous piece where I present a System for Reverting Global Warming, I explain what the Planetary Energy Budget is, its dynamics, and why Solar Power does not contribute to the thermal load of our planet. Undestanding this concept is fundamental to comprehend this piece and the danger behind exponential computing heat.

Earth’s climate is largely determined by the planet’s energy budget, the balance of incoming and outgoing radiation. It is measured by satellites and shown in W/m2. The imbalance (or rate of global heating; shown in figure as the “net absorbed” amount) grew from +0.6 W/m2 (2009) to above +1.0 W/m2 in 2019.

In any case, no one needs to be a genius to understand that warming up the oceans with Data Centers is not good for global warming:

Putting Data Centers under the sea, is one of the worst things that anyone not wanting to contribute to global warming could do. Oceans are the thermal-buffer of our planet that regulates global temperature.

The Solution: Orbital Super-Computing

It is clear then, that unless we externalize this massive amount of heat to space (out of our atmosphere), we will quickly overheat the Planet.

So… there are only 2 solutions to our ever-growing demands for computation:

  • Take Computation out of the Atmosphere.
    (externalize heat and energy demand)
  • Power our Computing Systems exclusively with Solar or other renewables.
    (Use energy that is already part of the Earth Energy Budget)

Now, I know that there are certain constrains to the deployment of this strategy, the mains being Cost-to-Orbit and Space-Based Cooling systems.

On the former, Elon Musk and Space X, and the rest of the commercial launch companies are doing great work.

On the later, I will present (very soon) an Innovation that I think can help tackle the problem of cooling computing hardware in the vacumm of Space. Please also notice that all the heat emitted in the generation of the electricity used to power our computing was mostly left out of this analysis.

Muchas gracias !



Ernesto Eduardo Dobarganes

Self-taught Polymath. Trying to beat Einstein while staying humble. Invented fastest Engine & Vehicle ever (~299,972 km/s).