If pykrete is such a cheap strong material, why don't we use it for regular buildings?

Probably because we have the option of tunnelling into regular ground instead for subways. But tunnelling in general is underutilized and overpriced, likely due to lack of innovation, risk aversion, high time preference, property rights problems, NIMBYism etc etc etc. Elon is absolutely right that Tunnels would solve traffic, and tunnelling into rock is better than making a pykrete megastructure.

But you certainly could make a pykrete structure on rock foundations (which are not everywhere!) and have a pretty economically valuable and affordable megastructure in the form of a multilayer city. I tentatively suspect that tunnelling the lower layers and then building skyscrapers (like Manhattan Island) is the superior option though.

In general, megastructures are underbuilt for political economics reasons. Things like The Hoover Dam and the US Highway System would probably be illegal to build today because of NIMBYism. Most of Manhattan Island would be illegal today too.

Also we have aging and shrinking populations in The West so there's immense social pressure to run things as a form of managed decline rather than innovate and build. Old people have lots of capital and limited life left, so they want to make everything quiet and static and peaceful, and they throw their monetary and political capital around to keep things that way.

This is a very silly and inefficient way to lower the sea level though.

It's much easier to just reduce the incoming solar flux.

Well setting aside the legal/sovereignty issue, Antarctica is freezing cold, permanently dark for 6 months and gets covered in thick layers of snow every year. And it is not accessible to sea transportation due to a lack of rivers and railways.

It's just much worse land.

You could try to settle some of the outlying parts but the cold and dark and remoteness would still make it really bad land.

You can move your iceberg to the middle of The Atlantic Ocean and be close to Europe and America but also sovereign. That's worth a lot more.

Are there any places that have notoriously sleepy governments and you could just make a compound there and do your experimental society? This ice structure thing seems pretty capital intensive, which makes me think buying land is still a better bang-for-the-buck.

You can't really make a great new nation, startup city or network state in secret. You have to advertise it.

And if you are doing it against the wishes of the local government, that will discourage people from coming. This kind of effect even works if the locals are militarily weaker than you. Look at Israel!

Now there are some places where startup cities/communities are happening. Prospera is an example. But that requires some quite specific circumstances and isn't replicable. Legacy governments just don't want to give up control of their land. Another example is Liberland. Liberland could potentially be pretty good but Croatia hates it and they send police there to trash it even though they don't have a legal claim on the land.

If you want a nation-type org without the land you can start a cult or a religion. Mormonism sort of did this but they are ultimately still ruled by the US central government and it is gradually destroying them. If Mormonism was happening on an island it would be much more robust. Scientology is another example. They have their own compounds and own parts of various cities in the USA and the UK. Religious groups like The Amish are a further example - they use language and religion as barriers to keep foreign ideas and people out, but I don't think that is replicable at scale today because there's a decent chance that legacy governments would shut you down. So you can avoid the need for land to some extent but you will pay a large price.

I am still not convinced the ice structures require any less work to maintain.

It's not just about the maintenance requirements, it's also about the durability of the land and cost per unit area.

Seawater will destroy normal concrete over decades, so is not suitable. But there is a substitute, geopolymer concrete.

You can make floating platforms at sea using geopolymer concrete which is costed at about $200/m^3 for a compressive strength of 50MPa and about a 5MPa tensile strength (though you can improve that with reinforcement bars).

The problem with this is that you are paying $200 per cubic meter. If you want a really solid floating island that sits many meters above the ocean, each square meter of land will require multiple meters of concrete underneath it even with a mostly hollow structure. So you are paying over $1000 per square meter for your land assuming that is is 50-100 meters deep and has a 90% void fraction.

The Troll A Platform cost $1.2bn (corrected for inflation) for about a 250 x 250 meter useable area which works out at about $20bn per square kilometer of land area, or $2000 per square meter.

It has a dry mass of 683,600 tons of mostly concrete, which is about 10 metric tons of concrete per square meter of useable land area. For a 100km x 100km floating island that would be $20 trillion. It would use 100 gigatons of concrete, which is 1/5th of the total amount of concrete ever made by humans (550 gigatons).

Basically you are paying premium prices for this and I don't think it is economically feasible.

As for maintenance I think both ice and geopolymer concrete are going to be relatively cheap to maintain per unit area for very large islands. With ice the main worry is localized failure of your cheap bottom insulation job. But a small fleet of automated robotic submarines and a grid of temperature sensors can probably keep that in check. Remember you are already paying $ hundreds of billions for this so maintenance is going to be a rounding error.