Будущее ИИ-вычислений: космическая энергия обгонит наземную к 2030 году Headline: The Future of AI Computing: Orbital Energy to Surpass Terrestrial by 2030

By the year 2030, orbit-based AI computing will be more affordable than that on Earth. This forecast comes from experts at the research team 33FG.

The analysts evaluated the costs of deploying equipment into space for solar energy capture and compared these expenses to similar resources available on the planet.

Currently, with the expense of transporting cargo to high orbit around $2000 per kilogram, satellites provide power at approximately $18-26 per watt. This is roughly twice as expensive as the rates found in terrestrial data centers ($12/watt).

If the delivery costs are halved, the price of «space» energy will become comparable to that of «earth» energy. At $500 per kilogram, «space» energy could be around 30% cheaper, and at $100/kg, it could drop by 50%.

Reusable Starship vessels with in-orbit refueling could make these figures attainable by the end of this decade.

The authors modeled four different architecture types:

Sending equipment to Highly Elliptical Orbit (HEO) requires in-orbit refueling at Low Earth Orbit (LEO), resulting in about 1.5 times higher delivery costs compared to LEO.

The first class costs around $2000 per kilogram. These orbital systems run at $18-26 per watt, compared to terrestrial systems at $12/watt.

Starlink-like systems in HEO can achieve prices comparable to terrestrial rates when launch costs reach about $500 per kilogram. Compute-optimized Starlink (HEO) reaches parity at $1000/kg and begins to undercut terrestrial infrastructure if launch prices drop below $500/kg.

At $100/kg in HEO, orbital architectures yield $6-9/watt, making them 25-50% cheaper than terrestrial data centers.

Further reductions in launch costs have minimal impact on the economics. The primary factor affecting system efficiency becomes the price of the equipment rather than the rocket.

In summary:

Analysts emphasized that orbital energy represents the future for humanity, as space offers an effectively limitless supply of solar energy and ample room for infrastructure. On Earth, both energy and space are becoming scarce commodities.

However, the challenge remains in designing a well-thought-out architecture that makes both equipment and launch costs competitive with terrestrial solutions.

Currently, mass-optimized architectures are not cost-optimized, and vice versa.

It is noteworthy that in November, Google announced plans for a satellite system in low Earth orbit to capture solar energy and power data centers.

In May, China launched 12 satellites as part of a project to establish a network of orbital supercomputers.