MaiaEdge to present on AI connectivity at DENOG18 in Essen
MaiaEdge, a Burlington, Massachusetts-based provider of cloud-managed hardware for private Layer 2 Ethernet connectivity, will participate in DENOG18, the 18th annual meeting of the German Network Operators Group, taking place 15–17 November 2026 in Essen, Germany.
Markus Hendrich, General Manager Europe at MaiaEdge, is scheduled to present a 30-minute session on 16 November titled "Ethernet Won Inside the AI Data Center. Can It Cross an Operator Boundary?" The session will examine the technical and governance complications that arise when AI inference workloads, which have been designed around low-latency Ethernet fabrics within a single data centre, are distributed across networks operated by independent parties.
The session agenda
Hendrich's presentation is expected to address how performance parameters such as latency and congestion control degrade or behave unpredictably once traffic traverses an operator boundary rather than a homogeneous internal fabric. The session will also cover data sovereignty considerations: specifically, how geography, jurisdiction, and legal ownership of network infrastructure shape the connectivity choices available to AI infrastructure operators.
DENOG18 brings together network engineers and operators from across Germany and the wider European internet community to discuss routing, peering, and emerging infrastructure challenges. It is a practitioner-focused event rather than a commercial showcase, making it a credible venue for a technically substantive discussion.
MaiaEdge was founded by engineers with prior experience in session border control and smart routing, including alumni from Acme Packet, which Oracle acquired, and 128 Technology, which Juniper Networks acquired. The company positions its platform as enabling automatic private Layer 2 path establishment across any transport and any network.
Market context
The question Hendrich's session title poses is becoming commercially urgent. AI training and inference architectures have grown increasingly distributed, with workloads spanning multiple data centres, co-location facilities, and cloud regions. Within a single data centre or campus, high-speed Ethernet fabrics built around RoCE (RDMA over Converged Ethernet) or similar protocols perform reliably. The challenge of maintaining comparable performance guarantees once traffic leaves a controlled environment and crosses third-party operator infrastructure remains largely unsolved at scale.
Carriers and network equipment vendors including Nokia, Ciena, and a range of SD-WAN and SASE providers are approaching this problem from the transport and security layers. Pure-play connectivity startups focused on Layer 2 abstraction across heterogeneous networks occupy a narrower, more specialised niche. For enterprise AI buyers, the practical implication is that distributed inference architectures depend on network SLAs that most operator agreements do not currently guarantee.
Regulatory dimension is also sharpening. The EU AI Act, combined with existing GDPR data-residency obligations, creates compliance complexity for AI operators whose inference traffic crosses national or jurisdictional boundaries within Europe. Hendrich's focus on sovereignty considerations is timely given that many enterprises are only beginning to map the jurisdictional exposure of their inference pipelines.
This is a media alert announcing a speaking engagement rather than a product launch or funding announcement, and MaiaEdge has not disclosed new technical specifications, commercial partnerships, or financial results alongside it.