ИИ-вычисления на орбите: прогнозы удешевления к 2030 году Translation: Headline: Orbital AI Computing: Cost Reduction Predictions by 2030

By 2030, the cost of artificial intelligence computations in orbit is expected to be lower than that on Earth. This prediction was made by experts from the 33FG research group.

Analysts calculated the expenses involved in sending equipment into space for solar energy harvest and compared them with similar costs on the planet.

Currently, with delivery costs to high orbit around $2000 per kilogram, satellites provide electricity at prices ranging from $18 to $26 per watt. This is roughly double the cost seen in terrestrial data centers ($12 per watt).

If the delivery costs are halved, the price of «space» energy will equal that of terrestrial sources. At a rate of $500 per kilogram, space energy would be about 30% cheaper, while at $100 per kilogram, it would be 50% less expensive.

Reusable Starship vehicles with in-orbit refueling could achieve such metrics by the end of this decade.

The authors modeled four types of architectures:

To transfer technology to highly elliptical orbits (HEO), refueling at low Earth orbit will be necessary, making delivery approximately 1.5 times more expensive than launching to low Earth orbit (LEO).

The first class incurs costs of about $2000 per kilogram. These orbital systems deliver electricity at $18 to $26 per watt, compared to ground systems at $12 per watt.

Starlink-class systems in HEO reach the same price point as terrestrial systems with launch costs around $500 per kilogram. Compute-optimized Starlink (HEO) achieves parity at $1000 per kilogram and starts to surpass terrestrial infrastructure when launch costs drop below $500 per kilogram.

At $100 per kilogram in HEO, orbital architectures can deliver power at $6 to $9 per watt, making them 25-50% cheaper than ground-based data centers.

Further reductions in launch costs have little impact on the economic landscape. The primary influence on system efficiency begins to shift from rocket costs to equipment prices.

In conclusion:

Analysts emphasized that orbital energy represents the future for humanity. In space, there is virtually unlimited solar influx and ample room for hardware placement. On Earth, both energy and space are becoming scarce.

However, the challenge of creating a well-thought-out architecture remains, ensuring that both equipment and launch costs are competitive with terrestrial solutions.

Currently, the most mass-optimized architecture is not cost-effective, and vice versa.

In November, Google announced the creation of a satellite system in low Earth orbit dedicated to harvesting solar energy and powering data centers.

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