In 1942, Geoffrey Pyke proposed Project Habakkuk: an aircraft carrier made of pykrete — water frozen with 14% wood pulp, which is roughly as strong as concrete but floats. The project died because keeping it cold cost too much. But what if we revived the idea — not for a warship, but for a permanent floating city anchored to the Arctic seabed?

The material. Pure ice has a compressive strength around 5 MPa and is brittle. Add sawdust or cellulose fibers, and pykrete jumps to roughly 7 MPa compressive, 5 MPa tensile — comparable to weak concrete, but with one-ninth the density (≈930 kg/m³). It also creeps slowly under sustained load, like glacial ice, and resists shattering: a bullet that pulverizes ice merely embeds in pykrete.

Sizing the slab. Let's design for a 50,000-resident city — call it 5 km² of usable surface. We need freeboard (above-water height) of at least 5 m to handle storm surge and wave action. Archimedes sets the geometry: if the slab has density 930 kg/m³ floating in seawater (1025 kg/m³), the submerged fraction is 930/1025 = 0.907. So 5 m of freeboard implies a total slab thickness of 5 / (1 - 0.907) ≈ 54 m.

Volume: 5,000,000 m² × 54 m = 2.7 × 10⁸ m³. Mass: ≈250 million tonnes. For comparison, the largest concrete gravity oil platform (Troll A) is 1.2 million tonnes. We are building 200 Troll As — out of frozen slurry.

The refrigeration problem. Here is where Habakkuk died. Heat flows in from above (sun, air, residents) and from below (seawater near 271 K, only 2 K above pykrete's melting point). For a 5 km² top surface in summer Arctic sun, solar gain alone is roughly 5 × 10⁶ m² × 200 W/m² = 1 GW. Bottom surface heat flux into ice from seawater is around 20 W/m² with natural convection, so another 10⁸ W = 100 MW.

Total cooling load: ~1.1 GW thermal. A modern industrial ammonia chiller has a COP near 3 at these temperatures, so we need ~370 MW of electrical input — about the output of a small nuclear reactor. An SMR (small modular reactor) like the BWRX-300 delivers 300 MWe. Two SMRs would keep the city solid.

Insulation matters more than refrigeration. If we cover the top with 2 m of foam insulation (k ≈ 0.03 W/m·K) and a reflective deck, solar gain drops 20×. Suddenly the cooling load is ~150 MW thermal, well within one SMR. The bottom is harder — we can't insulate the water-facing surface, but currents bring fresh cold water and the temperature gradient is small.

Structural creep. Pykrete under sustained stress flows at roughly 10⁻⁹/s per MPa. A 54-m slab sees ~0.5 MPa at its base, so it spreads laterally at ~1.5 cm/year. Manageable if you re-freeze the edges seasonally — essentially the slab self-heals, since seawater splashing onto the cold rim instantly freezes into more pykrete.

The catch. Climate change. Mean Arctic air temperatures have risen ~3°C since 1980. Each degree warmer roughly doubles your cooling load during the melt season. By 2080, today's design might need four SMRs instead of one.