OdysseyAI

Odyssey (OdysseyAI) is an El Segundo, California-based company building compute infrastructure designed to operate in orbit rather than on the ground. Its stated approach is a space-native architecture intended to exploit the physics of orbit, positioning orbital compute as an alternative to terrestrial data centers, which the company frames as constrained by electricity grid capacity, interconnection queues, transformer lead times, freshwater use and multi-year construction timelines. The company describes its vehicles as orbital supercomputers equipped with high-density, energy-efficient compute engines and multiple inter-satellite links so that each unit functions as a node in a larger constellation-scale system.

The product plan centers on the Odysseus vehicle line, offered in three classes: S-Class ("Scale-Out Worker") for edge inference, scale-out sampling and Earth-observation processing at more than 4,000 PFLOPS and under 200 kg; M-Class ("Performance Core") for core training at more than 15,000 PFLOPS and under 400 kg; and L-Class ("Heavy Lifting") for mixture-of-experts routing, gradient aggregation and checkpoint storage at more than 60,000 PFLOPS and under 600 kg. The vehicles are intended to form a laser-linked mesh in low Earth orbit. The company states that its architecture decouples compute density from heat generation and treats thermal and radiation-induced noise as a computational resource, and claims resulting cost per workload well below traditional Earth- or space-based computing tiers.

A separate startup-directory listing describes the company as deploying full-stack computing in orbit that turns satellites into autonomous nodes able to operate independently of ground support or Earth-based compute, classifies it under Aerospace & Defense, and lists it at seed stage.

Business model

Odyssey plans to sell compute capacity on its orbital supercomputers to external customers, with early allocation on Odysseus-1 and subsequent vehicles offered through a reservation process for researchers, operators and builders. A startup directory characterizes the model as B2B SaaS for space applications.

Revenue is intended to come from customers purchasing or reserving compute capacity on orbital supercomputers, described as a per-workload compute service priced below comparable Earth- or space-based alternatives.

Traction

As of the sources reviewed, the company's first vehicle, Odysseus-1, is pre-launch with flight targeted for Q1 2027 on a Falcon 9, and early compute allocation is being offered via a reservation page. An external startup directory lists the company at seed stage with one disclosed investor, Vibe Capital.