What if Moore's Law Never Ends?
From Microchips to Megastructures: A Journey Through 150 Years of Moore’s Law
"The number of transistors on a microchip doubles approximately every two years, while the cost per transistor halves." – Gordon Moore, co-founder of Intel. This quote is from 1965. That year, a chip had about 64 transistors on it; today we have a GPU with 150 billion transistors on it.
New innovations have been needed to keep up with this prediction, and it has slowed down recently. Instead of doubling every two years, it’s doubling every three to four years. Instead of the size of the chips getting bigger to accommodate the increase in transistors, they have remained about the same size; however, the size of the transistors has steadily been shrinking. As we get closer to the size of the atom, fundamental forces of nature are making shrinking the technology more and more difficult, and eventually, any further reduction in size will be impossible.
Modern transistors are around 2–3 nanometers in size, just 10–15 silicon atoms wide. We have only one or two more doublings before we run into the limit. The only way to increase transistors at this point is to increase the size. Some companies are already starting to build chips in three dimensions.
Now it may very well be too difficult to continue to grow processors much more at this point, but I don’t think Moore thought we would even get this far. Let's see what the future holds if we keep doubling:
In 10 years
– 📏 Size: Fingernail
– ⚡ Power: Toaster
In 25 years
– 📏 Size: Smartphone
– ⚡ Power: Apartment Building
In 50 years
– 📏 Size: Basketball Arena
– ⚡ Power: Large City
In 100 years
– 📏 Size: Texas
– ⚡ Power: Planetary Power Grid
In 150 years
– 📏 Size: Earth's Surface
– ⚡ Power: Nothing currently exists to compare
Obviously, at some point we’d need to start creating satellite CPUs that orbit the Earth, with solar panels to power them. In 150 years, the CPU would require a level of power hard to imagine. The Earth’s total energy consumption today is around 20 terawatts; the projected chip at 150 years would require around 7.56×10²³ watts—tens of billions of times more than Earth’s entire output.
Most modern solar panels are around 20% efficient. Assuming we keep increasing our technological abilities, the theoretical limit would be 85% efficiency using multi-junction cells. If we could get this efficient, it would still take 1,700 Earths' worth of surface area to power this CPU.
A very interesting part of the law is that the price halves each iteration of Moore’s Law, which means the cost of producing a planetary-sized CPU would be virtually nothing. This seems like it couldn’t possibly be right—just the cost of raw materials alone would be enormous—but with the technology we have in 150 years, perhaps this could be possible. Imagine some nanobots that are able to reproduce themselves, sprinkled onto some asteroids in the asteroid belt.
Of course, then you run into the problem of there simply not being enough raw materials in the solar system. The asteroid belt only contains 4% the mass of the Moon. The Oort Cloud contains somewhere between the mass of 1 and 100 Earths; however, it’s very far away—about 11 light-days—which means communicating with CPUs built this far away would be impractical. Unless, of course, the CPUs could eventually be brought closer to Earth’s orbit.
Beyond this point, the only choice would be cannibalizing other planets for CPU materials, and I’m sure some solar system environmental protection agency would be in place to prevent this kind of thing.
If we are all living in the Matrix at this point and no one cares about planets, and we are only looking to increase compute without any consideration to anything else, and we developed some kind of faster-than-light wormhole technology, in 208 years we would have turned the entire solar system into CPUs. In 306 years, we would have consumed the mass of the entire Milky Way galaxy. In 378 years, we would have consumed the mass of the observable universe.
We can assume this isn’t possible, as any civilization that was just a few centuries ahead of us would have done it already—unless our solar system is in some kind of universal wildlife preserve. Because Faster Than Light (FTL) technology is, well… faster than light, the whole universe could already have been converted to CPUs, but the light we’re seeing right now would be from all the old stars before the conversion began.
What could all this compute possibly be used for? Well, if all life in the future are digital consciousnesses, living in a virtual reality, the universe could be whatever anybody wants it to be. A recreation of what we had before, or a world with completely new laws of physics we couldn’t possibly imagine. Regular matter would only have a finite number of things that could be done with it, so a super-advanced civilization could use new mathematics to create new universes. Perhaps we’re already in one.