Definition

Neocloud refers to specialized AI compute providers that rent GPU clusters and AI-optimized infrastructure outside traditional hyperscale clouds — enabling enterprises to run open-source models at scale. The term also encompasses novel deployment models including orbital/space-based data centers.

GPU Rental Neoclouds (Primary Usage, 2026)

AI-native clouds like together-ai and runpod provide Nvidia GPU rental, inference, and training optimized for open-weight models — often cheaper than proprietary API tokens.

Key Characteristics

  • Hybrid terrestrial-orbital architecture: Ground data centers paired with satellite-mounted compute
  • Vertical integration: Companies that build both launch infrastructure and AI compute (e.g., spacex)
  • Constellation-scale compute: Leveraging existing satellite networks (Starlink) for distributed AI inference
  • Energy and cooling advantages: Space environments offer passive cooling and solar power for orbital workloads

SpaceX Positioning

On June 12, 2026, gwynne-shotwell declared SpaceX “100% an emerging competitor in the artificial intelligence neocloud space” during the SPCX Nasdaq debut (2026-06-12-spacex-spcx-nasdaq-debut):

  • Building ground data centers alongside colossus (xAI supercomputer)
  • Deploying compute on Starlink broadband and mobile satellites before dedicated AI1 satellites (late 2027)
  • AI1 satellite program for dedicated orbital AI workloads

July 2026: Together AI Milestone

Competitive Landscape

PlayerNeocloud Angle
together-aiOpen-source model hosting + GPU rental; $8.3B valuation
runpodDeveloper-focused AI cloud ($100M Series A, June 2026)
spacexOrbital data centers + Starlink constellation
nvidiaAI factory / Spectrum-X data center stacks
amazon / AWSHyperscale GPU clusters for anthropic
googleTPU pods + Visakhapatnam 1GW AI hub
openaiStargate / Microsoft Azure compute partnerships

Sources

Recent Developments