Introduction
The Physical Map of the Computational Infrastructure
The global computational infrastructure has shifted from the terrestrial layer to the orbital one. According to estimates by Goldman Sachs, the monthly consumption of tokens for artificial intelligence will reach 120 quintillions by 2030, an increase of 24x compared to 2026. This exponential growth exceeds the energy and thermal capacities of terrestrial data centers, which have already reached density power limits. The process begins with training on instantiated models: 12,500 computational satellites will be deployed in low Earth orbit by Nayuta Space to form the Alaya constellation, each equipped with solid-state processors dedicated to parallel computing.
The physical route is defined by launch via the Xuanniao-R rocket, a two-stage vehicle 70 meters long and 3.8 meters in diameter. The second stage becomes directly the structure of the satellite after being expelled into orbit. The trajectory concludes with horizontal aerodynamic reentry, a technology developed to reduce launch costs to approximately 1,000 yuan per kilogram (approximately $148), according to what was announced by Nayuta Space. The net effect is an increase in computational availability at a operating cost halved compared to terrestrial models.
Dynamics of Bypassing Terrestrial Infrastructure
The transition from terrestrial to orbital computing represents a logistical bottleneck. Traditional data centers, concentrated in areas with access to low-cost electricity (e.g., Scandinavia, Texas), are constantly under pressure to expand thermal capacity. According to JP Morgan’s data, the volume of token calls for language models increased by 20x in June 2026 alone, exceeding weekly consumption projections.
Alaya solves this problem through a geographically distributed solution that is not dependent on ground infrastructure. Communication between satellites occurs via picosecond-speed lasers, with global transit delays of less than 50 milliseconds. Unlike terrestrial servers, which require 12 to 48 hours for passive cooling between intensive operations, Alaya satellites use thermal radiation in space as a heat dissipation mechanism. The tariff differential compared to terrestrial data centers is estimated at 40% for the unit cost of computing power, with activation time for capacity reduced to less than 72 hours from launch.
Strategic Leverage of Orbital Control
Acquiring and managing an orbital constellation is no longer just a technological project, but an act of logistical control on a global scale. Companies that possess computational capabilities in orbit can decide which models are available for training and what type of data has priority in processing. The value shifts from simple access to computing to structural influence on the flow of data.
The competition is no longer just about the price of GPUs, but about the ability to integrate physical infrastructure with security and authentication protocols. Services based on Alaya could become standards for encrypted processing, with an operational advantage that translates into margins 25% higher than terrestrial competitors. The strategic leverage lies in controlling data routes: those who manage the orbital network decide who has access and when.
Impact on Margin
The narrative states that AI is transforming markets. The data shows that the operational cost per unit of computation has decreased from $1.80 to $0.75 with the migration to Alaya. This net impact translates into a 32% increase in operating spread for companies using synthetic models on a large scale.
The gap is evident in the availability of capacity: while terrestrial data centers are hampered by environmental approval delays and power limits, Alaya has already successfully completed three experimental launches. The Impact KPI is an 18% improvement in the operational cycle for training multimodal models, reducing the time from 72 to 59 hours.