Towards a future space-based, highly scalable AI infrastructure system design
read the original abstract
If AI is a foundational general-purpose technology, we should anticipate that demand for AI compute -- and energy -- will continue to grow. The Sun is by far the largest energy source in our solar system, and thus it warrants consideration how future AI infrastructure could most efficiently tap into that power. This work explores a scalable compute system for machine learning in space, using fleets of satellites equipped with solar arrays, inter-satellite links using free-space optics, and Google tensor processing unit (TPU) accelerator chips. To facilitate high-bandwidth, low-latency inter-satellite communication, the satellites would be flown in close proximity. We illustrate the basic approach to formation flight via an 81-satellite cluster of 1 km radius, and describe an approach for using high-precision ML-based models to control large-scale constellations. Trillium TPUs are radiation tested. They survive a total ionizing dose equivalent to a 5 year mission life without permanent failures, and are characterized for bit-flip errors. Launch costs are a critical part of overall system cost; a learning curve analysis suggests launch to low-Earth orbit (LEO) may reach $\lesssim$\$200/kg by the mid-2030s.
This paper has not been read by Pith yet.
Forward citations
Cited by 12 Pith papers
-
Designing Dense Satellite Clusters for Distributed Space-based Datacenters
Proposes planar and 3D satellite cluster architectures for LEO datacenters that meet inter-satellite spacing, unobstructed sun-vector, and ISL constraints and support VL2-like Clos network fabrics.
-
SpaceMoE: Realizing Distributed Mixture-of-Experts Inference over Space Networks
Space-XNet partitions satellite constellations into ring subnets for MoE layers and maps high-activation experts to low-latency satellites, yielding at least 3x lower inference latency than random or ablation placemen...
-
SpaceMoE: Realizing Distributed Mixture-of-Experts Inference over Space Networks
SpaceMoE partitions MoE layers across orbiting satellite subnets in a ring and optimizes expert placement by activation probability and path latency, yielding at least 3x lower inference latency in thousand-satellite ...
-
Generating quantum entanglement from sunlight
Sunlight produces polarization-entangled photons through SPDC, achieving concurrence 0.905, fidelity 0.939, and Bell violation S=2.54 exceeding the classical limit.
-
Toward Communication-Efficient Space Data Centers: Bottlenecks, Architectures, and New Paradigms
Semantic communication in a multi-layer heterogeneous space data center framework can substantially reduce uplink pressure for orbital AI by sending compact representations rather than raw data.
-
Orbital Data Centers: Spacecraft Constraints and Economic Viability
Orbital data centers require 34-59 kg/kW total mass and launch-plus-build costs 3.4-13.5 times below current Falcon 9 prices to compete, viable only for low-communication edge compute with high utilization and long lifetimes.
-
Equinox: Decentralized Scheduling for Hardware-Aware Orbital Intelligence
Equinox uses a barrier-function-derived marginal cost to enable value-based adaptive scheduling and neighbor offloading in energy-constrained satellite constellations, yielding 20-31% throughput gains and higher batte...
-
Hot AI in Cold Space: Thermal-Crosstalk-Aware Scheduling for Sustainable Orbital AI Clusters
Position paper introduces Thermal-Aware Heterogeneity Thesis and TLB framework to mitigate thermal-fluid and thermal-radiative crosstalk in high-density orbital AI clusters by dynamic workload migration.
-
Hot AI in Cold Space: Thermal-Crosstalk-Aware Scheduling for Sustainable Orbital AI Clusters
Proposes Thermal-Load Balancing to migrate workloads to cooler orbital units and thereby restore MFU while lowering thermal stress.
-
Space-CIM: Enabling Compute-In-Memory Accelerators for Thermally-Constrained Space Platforms
CIM accelerators outperform GPUs in TOPS/W under space thermal constraints due to uniform heat distribution and radiator-linked co-design.
-
The Many-Body Problem of the Data Centre
Data centers embody AI as part of capital's body, creating a many-body problem from their non-unique form and enabling market-based comparison of intelligence value across organism-mechanism boundaries.
-
Revolutionizing Wireless Communications with Space Data Centers: Applications and Open Challenges
The paper outlines a hierarchical SDC network architecture with access, relay, computing, and control layers, explores applications in space communications, presents latency reduction simulations, and lists open resea...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.