There was a 110 year increase of 1000 times in world electricity from 1915 WW1 levels to today. The Industrial revolution increased the world economy by 100 times and increased electricity from nothing but after the first 25 years of establishing electricity increased it by 1000X.
Only about 1% of global electricity is used for AI right now. ~200 TWh. 500 TWh for all data centers.
The path is clear to use SpaceX Starship to have 3000-5000 launches per year and the mass production of 1 million satellites to achieve 1000 times the current level of electricity for AI within fifteen years.
100 GW per year for ten years would add 8000 TWH over a decade (Adding all current annual world energy additions every year)
300 GW per year for ten years would add 24000 TWH over a decade (World electricity every 10 years)
The 300 GW per year level would be over 10 times the 200 TWh used for AI this year. This will be part of trend to scale artificial general intelligence and superintelligence to what will feel like a century of progress and transformation.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
The is no rich low energy civilization.
As a corollary, there is no poor space resource exploiting one.
In truly astounding terms. A future where everyone has a wealth of equivalent billions, even if everything is so cheap you don’t see the currency go to those numbers
Using 1000 times more energy on the Earth’s surface would fry all biological life from the waste heat alone. Using it in space to power data centers in space and then beaming the results to earth could be a great use case for space based solar once we get energy demands high enough.
A dusk-dawn sun-synchronous orbit has close to 0% darkness. Rather than 40%.
It’s set up so the satellite is orbiting the earth in a plane perpendicular to the line from the earth to the sun. So no shadows. And it’s put at an altitude such that the equatorial bulge of the earth has gravity that is constantly nudging it as it passes by, causing that plane to rotate (precess) at a rate of 360 degrees per year. So it’s always perpendicular to the line to the sun. So it avoids shadows all year.
This was meant as a reply to the post below.
The catch is that there are two types of sun synchronous orbits, the sort that it always facing the sun, but unsynchronized with the Earth, and the sort that arranges to always have the sun in the same orientation relative to the Earth as you pass over a particular location. (That latter is great for photography… Not so great for solar power.)
I put in the numbers into GROK to analyze if space based power for AI was realistic.
I fully expected it not to be but …. maybe.
I assumed high LEO for minimal DeltaV maintenance and about 100 – 150 tons payload. Some battery power to smooth out the eclipses (40% at this altitude).
Each launch adds about 22 MW continuous power with the overhead I put into the equation.
Then, I analyzed the energy spent for production and launch to see ROI in terms of energy.
It turns out to be ~80 days! I would have guessed a lot more.
This was just the energy budget.
AI hardware was not included.
A dusk-dawn sun-synchronous orbit has close to 0% darkness. Rather than 40%.
It’s set up so the satellite is orbiting the earth in a plane perpendicular to the line from the earth to the sun. So no shadows. And it’s put at an altitude such that the equatorial bulge of the earth has gravity that is constantly nudging it as it passes by, causing that plane to rotate (precess) at a rate of 360 degrees per year. So it’s always perpendicular to the line to the sun. So it avoids shadows all year.
I adjusted the calculation to use a dusk-dawn sun-synchronous orbit.
I also changed the equation from using space grade solar cells to mass-produced “normal” cells to get a better financial ROI. With space grade cells, payback time was 172 years.
I also assumed 120 tons payload (this could be better or worse when Starship actually starts operation).
With lesser and cheaper panels but more efficient orbit, energy payback drops to ~45 days.
Financial payback is harder to calculate because it’s not isolated to the energy infrastructure. If the panels cost $1/Watt, financial payback is about a year looking at energy alone. However, there is no revenue without a local datacenter that consumes the power. Launching the hardware costs money but so does a ground based datacenter.
The financial TCO changes up and down a lot with variations to the energy price, possible effect of launch failures, radiation damage, orbital debris etc.
Interestingly when GROK compares to ground based AI datacenters, they have negative ROI unless energy is free (solar).
Space based wins with about ~15 year ROI but this is unrealistic because no compute hardware is viable for so long.
The revelation is that no AI datacenter is profitable with traditional revenue models. The industry must calculate with something other than just operational income.