LNG Liquefaction Fused with Superconducting Datacenters

Key Idea:

Superconducting digital logic (Superconducting computing - Wikipedia) is a promising technique to overcome the limits of traditional CMOS. In principle, it can be much lower power (no Boltzmann limit).

Unfortunately, total cost of ownership (as well as power efficiency) is dominated by the cost of the cryogenic cooling.

An interesting idea is to use natural gas liquefaction infrastructure to achieve this cryogenic cooling. In essence, a nat gas liquefaction plant can be co-located with the superconducting datacenter, and the plant can be used for cryo-cooling.

This has the following advantages:

1. Extremely cheap due to massive economies of scale

   1. LNG liquefaction is big business, and is made at massive scale

2. Take advantage of all the innovation that’s gone into LNG liquefaction

3. Capital downside risk mitigation

   1. If the superconducting datacenter fails for whatever reason (technical or market), the LNG liquefaction plant can still be used, as the LNG export market size is expected to grow dramatically over the next 1-2 decades. By contrast, dedicated cryocoolers only for superconducting logic would be stranded if the project failed for whatever reason.

4. Economies of scale-up (large sizes) mean low capex and maximum efficiency

   1. In addition to more LNG cryocoolers being made, LNG liquefaction plants are also huge- and most of the components involved (eg pressure vessels, pumps, etc) get cheaper per unit of throughput the bigger they get. Many mechanical systems exhibit this phenomenon, which is why e.g. rockets are cheaper per kg the larger they get. This is also generally true for cooling efficiency (you generally get closer to Carnot the bigger you make the cooling system).

The key challenge is that LNG is at around ~110K, while most superconducting logic is at ~4K. In principle high-T superconductors could be used, although this does require changing the architecture of superconducting logic itself (though not impossible, and likely much easier than doing so for superconducting quantum computing).

Some early efforts at this: [2305.14184] High-Temperature Superconductor Quantum Flux Parametron for Energy-Efficient Logic, ‘Sail’ high temperature superconductor digital logic: improvements and analysis | IEEE Journals & Magazine, On the Path to High-temperature Josephson Multi-junction Devices, High-Transition Temperature Superconducting Digital Electronics With Helium Ion Irradiated Junctions.