Adagio Orbital Compute Layer
Symphony Space · Orbital Compute Layer01 / 07
Symphony Space · Orbital compute layer

From sensor to answer, in orbit

Adagio edge nodes at 400 km capture what Earth- and space-watching satellites already transmit; Adagio-XL hubs at 800 km fuse it across operators. Answers go to the ground; tasking goes back to the satellites.

Partners & investors · November 2026

A shared processing layer in LEO for the satellites that watch Earth and space

Edge nodes at 400 km capture the downlinks of commercial sensor satellites; Adagio-XL hubs at 800 km add heavy compute and Ephemeris optical links to customers and SDA layers. Sophia Space TILEs and Satlyt software turn raw data into products in orbit: 32 edge nodes and 8 hubs, ~2.7 EOPS at full deployment, $639M base-case revenue in 2035.

−25–50%wait before data leave the satellite, as an add-on to the customer's ground network
MB, not GBraw data stay in orbit; only products come down, freeing ground passes
3 sourcesradar, RF and optical fused across operators in orbit, before downlink
~5 minto a hub for customers with an optical terminal
Symphony SpaceAdagio and Adagio-XL platforms, integration and operations; stores and swaps GPUs in orbit with a robotic arm.
Sophia Space · TILE4 Blades (Jetson AGX Orin) per TILE: 400–800 TOPS, 8 TB, self-powered, passive cooling.
SatlytEdge AI software: data reduction at the edge, edge-to-hub bandwidth optimization, logistics control.
EphemerisOptical terminals (SRT, LRT up to 100 Gbps) and the Ephemeris Orbital Network for backhaul.

Figures from Symphony Space mission and market models (draft). Use ← → or the buttons below to navigate.

Who needs it

Sensor fleets are growing faster than their downlinks

A SAR or hyperspectral scene can be several GB and waits for a few minutes of ground contact per pass. Defense customers buy time to insight, and cross-sensor fusion still happens on the ground. Our benchmark covers about 40 operators and programs, representing ~630 satellites in orbit today (~550 commercial EO and SSA, ~85 SDA Tranche 1).

Fit: High = large data volumes, defense customers, ISL-ready. Fleet sizes approximate (public sources, 2026).

Where it fits

A processing layer between sensors and transport

Ground & users12 US gateways (CONUS, AK, HI, GU, PR), operations centers, tactical users
▲ products, not raw data
Battle management & orchestrationSDA Battle Management layer, SDN Mission Operations Center, AI routing
Transport / SDN backbonepLEO mesh; SDA Transport T1/T2; T3 moved into the Space Data Network
Compute layer (Symphony)Edge nodes at 400 km capture existing X/Ka downlinks; XL hubs at 800 km take OCT optical crosslinks; detect, fuse and track in orbit
▲ raw data over crosslinks
Sensor layersSDA Tracking & Custody · commercial optical, SAR, RF, hyperspectral · SSA fleets

In the FY27 budget the Space Force folded SDA's Tranche 3 Transport Layer into a new Space Data Network backbone (~$1.5–1.6B requested). Its force design calls for orchestration, "Space Exchange Points" and more commercial vendors, accelerated under Golden Dome.

"I can't see it without it."U.S. Space Command SATCOM lead, asked whether Golden Dome needs space-based compute
  • Standards first: SDA OCT optical crosslink compatibility, RF crosslink for non-optical customers
  • Multi-tenant: partitioned government and commercial workloads
  • Closes the loop in orbit: tasking, cueing and manoeuvre products sent back to satellites for autonomous decisions
  • Refreshable: robotic arm swaps TILEs at each GPU generation
  • Contract path: SpaceWERX SBIR → STRATFI, AFRL / SDN demos, SDA HALO
Mission analysis

Edge nodes below the customers, XL hubs above them

Edge node, 400 kmAdagio-XL hub, 800 kmDownlink captured by an edge nodeOptical crosslink to a hub

24 h simulation (circular orbits with J2). Capture: edge node inside the customer's ≤65° antenna cone, ≤2,500 km. Optical: ≤5,000 km, line of sight ≥200 km above the surface. Animation is time-accelerated and illustrative.

Market & revenue

An $11B market by 2035, with a defensible slice for a neutral layer

TAM
$11.1B
SAM
$4.5B
SOM
$639M

TAM: space-based data centers in 2035 (Market.us, 21% CAGR). SAM: the LEO share of commercial, civil and defense demand. SOM: about 13% of SAM in the base case, excluding downlink capture.

  • New stream: downlink capture as a "space ground station", $250k per satellite-year
  • Capex: ~$1.95B to 2035 for 32 edge nodes and 8 Adagio-XL hubs; peak funding need ~$1.15B (2033)
  • EBITDA positive 2030, free cash flow positive 2034, payback 2036

Illustrative planning model, not a forecast. Hover or tap bars for detail.

Roadmap & next steps

From a two-node pathfinder to a 40-satellite compute layer in five years

2027Team & anchorSophia, Satlyt and Ephemeris agreements; SBIR DV513 with Satlyt; capture MoUs with 2–3 EO operators
2028–29Pathfinder2 edge nodes at 400 km capture a cooperating customer's downlink; first paid compute pilots
2030–31IOC · 16 edge + 4 hubs2 matched planes: capture ~165–240 min/day per matched customer; optical 84–97%; OCT test with SDA
2032–33FOC · 32 edge + 8 hubs4 matched planes; optical crosslinks 99.3–100%; ~2.7 EOPS in orbit
PartnersSophia Space (swap-compatible TILE, 400/800 km qualification), Satlyt (edge AI, logistics control plane), Ephemeris (SRT/LRT terminals, network).
Anchor customersSpace Force SDN / Golden Dome, SDA, AFRL; ICEYE, HawkEye 360, Planet, Umbra; SSA operators Vyoma, NorthStar, Digantara.
InvestorsFund the pathfinder (~$100M incl. ground segment, per model) and the bridge to IOC alongside government co-funding.

Reference documents: SYM-DDJ-ADG-ODC-001 White paper · -002 Market and business case · -003 Mission, architecture and industrial breakdown.