Europe’s digital sovereignty ambitions extend beyond cloud and data policy into the nuts and bolts of Industry 4.0 network orchestration. Through initiatives like IPCEI-CIS and 8RA, organisations including Reply Adeptic are building a federated cloud-edge continuum that can orchestrate multi-vendor, multi-network deployments at scale, embedding AI, automation, and interoperability to power next-generation industrial applications.
In sum – what to know:
Orchestration of heterogeneity – Reply Adeptic’s platform manages distributed compute and connectivity, from private 5G and Wi-Fi to LoRaWAN, using a three-tiered orchestration model and intent-based automation.
AI-driven lifecycle management – Artificial intelligence supports site selection, runtime network adaptation, predictive alerts, and optimal workload distribution, enabling real-time performance, resilience, and scalability.
Interoperability and collaboration – Reply Adeptic is helping to align vendors, integrators, operators, and cloud providers to standardise challenges to make a federated, sovereign cloud-edge ecosystem across Europe.
Europe’s push toward digital sovereignty is often framed around cloud infrastructure, data protection, and geopolitical autonomy. But beneath the headlines, a more technical – and maybe more transformative – shift is underway: the creation of a federated cloud-edge continuum capable of orchestrating complex multi-technology networks for the most demanding industrial environments.
This is the ambition of the IPCEI-CIS (Important Project of Common European Interest for next-generation Cloud Infrastructure & Services) and 8RA initiatives, two linked programmes involving more than 100 organisations reshaping how Europe deploys and operates distributed compute, connectivity and data services. Among the contributors is Italian outfit Adeptic Reply, part of Turin-headquartered consulting and integration firm Reply.
Adeptic Reply’s work focuses on a foundational challenge in this new ecosystem: how enterprises monitor and orchestrate multi-vendor, multi-cloud, and multi-network deployments at scale. At Industrial Wireless Forum last week, Gabriele Losposto, lead architect at Adeptic Reply, set out the architecture and rationale behind its contribution to the 8RA project – and provided a view from inside one of Europe’s most ambitious cloud-edge programmes, and a glimpse of how industrial connectivity will evolve over the next decade.
Why multi-network orchestration matters
But it is probably worth understanding why network orchestration is even part of a cloud-sovereignty project, first – before getting into the bones of the 8RA initiative. Modern industrial environments do not rely on a single connectivity domain, of course. They variously combine private 5G for deterministic latency and mobility, Wi-Fi for best-effort indoor coverage, LoRaWAN and other private IoT technologies for sensor comms, and non-terrestrial (NTN; satellite) systems for out-of-reach operations and fallback.
Throw in Ethernet and FTTx for LAN access and fixed backhaul (or midhaul where software control is possible) – plus a bustling stable of local compute nodes and infrastructure componentry, tightly coupled to the network to raise security, control, and performance, including for incoming AI workloads – and industrial enterprises have a veritable spaghetti-junction of networking technologies at the edge, connecting to the cloud, and back again into another networking mish-mash at another site, multiplied across their global footprints.
Of course, these networks are often owned or managed by different providers, as well. As enterprises seek control and visibility over their systems, in order to load them with performant AI tasks, and as Europe pushes a federated cloud-edge agenda, interoperability is the only game in town – the table stakes just to play; to make a sovereign cloud-edge continuum function transparently and efficiently at scale. Interoperability is a prerequisite, and not an optional extra; so are observability, automation, and orchestration.
This is where Reply Adeptic’s work fits into the broader European agenda. As Losposto explains, the goal of 8RA and IPCEI-CIS is to enable “a sovereign and competitive European cloud ecosystem… [that is] able to ensure interoperability and portability”. (Note, his quotes have been smoothed for clarity.) Network orchestration is part of that portability, to be able to move applications, workloads, or services from one environment to another with minimal friction, without having to rewrite, reconfigure, or otherwise heavily adapt them.
A three-tiered orchestration model
Reply Adeptic’s contribution is centred on an orchestration platform built to manage distributed resources across cloud and (network and on-prem) edge environments – with an equal focus on deployment and observability. At Industrial Wireless Forum, Losposto explained the platform in three layers, as below (see slide graphic also).
1 | End-to-end orchestrator – the federation layer
At the top sits the end-to-end orchestrator, responsible for interfacing with external organisations, enabling cross-domain federation and handling the first stages of application lifecycle management. “It’s based mainly on standard interfaces like GSMA APIs,” Losposto notes. At this layer, applications are defined, deployment intents created, and the system decides which sites or resources will be involved.
2 | Central orchestrator – the coordination layer
Below that sits a central orchestrator, an intermediate coordination tier responsible for translating high-level application requests into actions executed across multiple local domains. This layer is where Reply Adeptic manages service-level requests and ensures they are routed to the appropriate local orchestrators across private 5G setups, Wi-Fi domains, industrial gateways, and sundry edge clusters.
3 | Local orchestrator – the infrastructure layer
At the lowest level is the local orchestrator, responsible for communicating directly with the underlying network and compute infrastructure. This includes interacting with FTTx systems, O-RAN components, Kubernetes clusters, and virtualised network functions. “This is… where we directly interact with the infrastructure,” says Losposto. Each layer includes a dedicated lifecycle manager to enable the system “to be very optimal when deploying applications on different sites” – which is a key requirement when dealing with distributed industrial networks.
Dual-flow deployment and observability
The architecture uses intent-based orchestration, a concept borrowed from cloud automation and increasingly applied to networking. Enterprises express what they want (not how to do it) through a structured intent file (expressing their high-level goals), and the system reads the file and figures out which sites, networks, and compute resources are needed to satisfy the intent. It then automatically deploys the workloads onto Kubernetes-based edge clusters using GitOps practices, which track changes and ensure repeatable deployments.
Central to the architecture is a data model that defines a shared representation of applications, resources, and configuration. In a multi-vendor system, these shared models act like a Rosetta Stone, allowing different software components and network elements to interpret the same instructions consistently, even if they come from different vendors. This combination – intent-based deployment, shared data models, open-source orchestration – means the system can autonomously deploy and manage applications across diverse sites and networks.
In the end, a federated cloud-edge ecosystem is not possible if every industrial site or network vendor implements orchestration differently. Which is why IPCEI-CIS places so much emphasis on interoperability – and open interfaces and standardisation. And while application and network configuration flow downward through the orchestration stack, observability flows upward. Metrics and logs from edge clusters, network controllers, and radio units move from the local orchestrator up through central and end-to-end layers, enabling cross-domain visibility.
Losposto explains: “Once the application has been deployed… monitoring capabilities send relevant information back to the orchestration layer.” This allows each layer to make informed decisions — not just on performance, but also on SLA compliance and predictive maintenance. This dual flow – intent down, telemetry up – creates the conditions for closed-loop automation, a core aspiration of both telecom and cloud-native architectures.
Where AI fits into the orchestration loop
Artificial intelligence (AI) is embedded throughout the lifecycle of Reply Adeptic’s orchestration platform, acting as a continuous enabler across planning, deployment, runtime management, and monitoring. At the planning stage, AI supports site selection. “AI supports site prediction when choosing the right location,” Losposto says. The system evaluates the characteristics of potential sites, network conditions, and infrastructure capabilities to recommend optimal deployment locations for workloads, a task too complex for manual evaluation across multiple sites.
Once applications are deployed, AI continues to influence operations in real time. Runtime network adaptation, for example, allows the platform to dynamically reconfigure network parameters to meet service requirements or perform automated failovers in response to outages. Predictive analytics generate alerts to prevent service interruptions; AI-driven workload distribution ensures tasks are allocated optimally across clusters. So AI doesn’t just automate tasks, but actively maintains performance and resilience across a sprawling, heterogeneous network footprint.
In multi-network, multi-site environments – such as manufacturing plants, transport hubs, or autonomous systems – this intelligence becomes essential. Variables such as network conditions, mobility of devices, latency demands, and compute resource availability fluctuate constantly. Hands-on manual management does not keep up. AI allows orchestration to scale and adapt automatically, ensuring that applications have the network and compute resources to perform reliably. All of which is particularly valuable in critical industrial use cases.
Losposto highlights sectors where these architectures are most applicable: autonomous vehicles, manufacturing, and transportation. Here, applications cannot function in isolation – they must constantly interact with the network. A robot on a shop floor might require seamless handovers between private 5G and Wi-Fi networks, for example. A logistics hub may combine LoRaWAN sensors with satellite redundancy to maintain oversight in remote areas. Rail systems might deploy low-latency edge compute nodes across stations to support signaling and information services.
Each case presents heterogeneous networks that must operate in a coordinated manner – which is the whole orchestration challenge that Reply Adeptic is looking to address.
Cultural barriers and commercial roadmaps
And yet, the greatest barrier is not technical, observes Losposto. “The biggest challenge is to create a basis for interoperability,” he says. Diverse vendors, competing priorities, and inconsistent interpretations generate friction. The company’s role in the IPCEI-CIS / 8RA project is to align stakeholders – to “share needs, requirements, and goals [and] find a common way of working”. By bringing together vendors, integrators, operators, and cloud providers, the project is intended to enforce consensus around data models, interfaces, and lifecycle management – to ensure that AI-driven orchestration can operate seamlessly across Europe’s federated cloud-edge environment.
Reply’s work is currently in a research and development phase, expected to last around three years. After that comes industrialisation, and ultimately a deployment phase stretching “more than 10 years across Europe”. For a project spanning multiple countries, network domains and industry verticals, this timeline is unsurprising. The cloud-edge continuum is not merely a technology stack; it is a new operating model for European digital infrastructure.
