Northeastern shows first open-source mMIMO AI-RAN

Northeastern University built the first open-source massive MIMO AI-RAN prototype.

Researchers at Northeastern University said they have demonstrated the first open-source prototype of a massive MIMO AI-RAN system at the university’s Institute for Intelligent Networked Systems Open6G OTIC. The testbed, shown on May 20, 2026, combines hardware from AmpliTech Group, software from OpenAirInterface, and NVIDIA GPU-accelerated RAN processing.

What changed

The prototype ties together an AmpliTech mMIMO O-RAN Category B radio unit, NVIDIA AI Aerial for layer 1 and layer 2 processing, and OAI’s L2+ stack into a single standards-compliant system. Northeastern said the setup uses exclusively open-source software at the higher layers and was built as a reproducible reference implementation.

The team also used a two-stage precoding design: a 4-layer CSI-feedback-based MIMO precoder accelerated on NVIDIA AI Aerial, plus a 64-antenna codebook-based beamformer inside the CAT-B O-RU. In testing, the system sustained throughput across multiple user equipment devices under mobility, with layer-2 beam management helping maintain performance.

• Open6G OTIC had already certified the AmpliTech CAT-B RU for O-RAN conformance.
• The demo covers the full stack from physical layer to RAN control plane.
• Northeastern says the setup can be reproduced without proprietary closed-stack components.
• AmpliTech’s radio is a 64T64R massive MIMO unit.

Why it matters

For developers, the main signal is that high-performance massive MIMO no longer has to depend on vendor-locked integrations. The demo shows that AI-RAN functions can run on GPU-accelerated infrastructure while keeping the software stack open and interoperable.

For the market, the proof point strengthens the case for Open RAN and AI-native wireless systems in 5G and 6G work. It also gives researchers and operators a reference platform for testing multi-vendor interoperability, beam management, and compute-heavy RAN workloads in realistic conditions.

AmpliTech said the integration validates its radio technology in a disaggregated environment, while OAI framed the project as evidence that open standards can support full mMIMO AI-RAN builds. The practical question now is whether more labs and vendors can reproduce the same mix of openness, throughput, and mobility handling outside Northeastern’s testbed.