In order to build a new town, far from existing towns, you need a major engineering project to build all the transport/communications infrastructure as well as all the buildings. That kind of thing would likely be done using systematic building technology like the machine that was demonstrated.
Consider Canada with a huge supply of petroleum that could be turned into natural gas and piped to a distant shore if only the land was unpopulated to avoid the NIMBY effect. Look at the land between Fort McMurray and the Arctic Ocean. Imagine building a gas pipeline parallel with a railway and a highway along with a port city and harbour on the Arctic Ocean. You would also need to build gravel pits, limestone mines, cement plants to burn the limestone using natural gas, and greenhous complexes using natural gas to generate heat and electricity for artificial lights.
It could be done, if Canada has the will. We already know that Russia has the will because they have done this, and they are running so many big engineering projects that they are likely building up to something bigger that would help ease the refugee problems of nearby countries by absorbing them into new Russian Arctic cities. Being a Muslim country gives Russia an advantage in this area, but Canada has other advantages so the game is not yet over.
Engineers generally like big problems like this, and a world where Global Warming will displace large populations needs to get really good at managing huge engineering projects.
Just look at Russia and their Crimea bridge, St. Petersburg bypass bridges, Vladivostok bridge, infrastructure projects associated with the Sochi Olympics, the Universiade games in Kazan and so on. And now Trump says that he plans on fixing the USA's decrepit infrastructure that has been neglected for the past two generations.
Engineering solves problems.
China building unoccupied cities has more to do with the fact that the average Chinese citizen is not allowed to speculate or invest in much other than real-estate.
They failed in that there was a huge investment to build them, but the expectations for these cities were never met. Additionally, there is tremendous upkeep or maintenance.
For example, those new roads are crumbling and require extensive repair work. Think of a road that is 5 lanes wide, but a 1 lane would suffice to serve all of the traffic. The unused portion will still degrade, even though it's not used.
It's kinda like buying a huge house before having children. You still need to fix, maintain and clean the extra square footage, even though it's not used.
And not all house can be built with an ideal orientation
Beyond thermal performance, optimizing windows for lighting potential is also interesting because electric light is a huge contributer to total energy consumption for offices, and outdoor light reflection and transmittance into interior areas depends a lot on geometry and material selection.
One example would be energy. Right now heat transfer and the transmission of daylight through walls/windows can be theoretically optimized to reduce energy costs simply by customizing the geometry of the windows and walls relative to their orientation and local context. We don't do so in architecture because the costs of customizing the geometry is too expensive. Architecture has a history of attempting mass-customization since the 1960s[1], but it's always been limited by the technology and realities of integrating multiple subcontractors and consultants. Those problems still exist, but the industry is changing pretty fast and I feel automated form generation, simulation, optimization, and production technologies are converging to a point when this could be achieved.
[1] i.e John Habraken.