Supply chain AI deployments require both edge and cloud infrastructure working together to achieve enterprise-scale success. While edge AI delivers speed and cost benefits, a cloud-based network layer provides essential updates, connectivity backup, security, and coordination that enable reliable, scalable logistics automation.
- Only 5% of supply chain AI initiatives have progressed beyond experimentation to enterprise deployment, often due to missing cloud infrastructure layers.
- Edge AI alone cannot deliver reliable scale—cloud networks provide critical updates, backup connectivity, and security when devices go offline.
- Purpose-built lightweight AI models on edge devices require cloud support for continuous optimization and cross-device coordination.
- Composable cloud solutions enable cost-efficient, flexible infrastructure proportional to edge deployments without vendor lock-in.
- Emerging AI agents in warehouses depend on cloud-based networks for multi-agent coordination, compliance monitoring, and real-time communication across distributed systems.
Over the past few years, the supply chain has served as a powerful proving ground for artificial intelligence on the edge. From warehouse robotics to transit tools for shipping fleets to workers’ wearable safety devices, new innovations in on-device AI promise transformational efficiency gains.
But the experimentation phase is coming to an end. A tight labor market and rising operational costs are increasing the need for logistics automation, pushing accelerated expansion of edge AI deployments to fill the gaps.
Purpose-built edge AI delivers real performance outcomes that measurably improve quality and reduce risk. But how many organizations are actually ready for large-scale deployment?
According to a recent study from UVA and GEP, only 5% of supply chain AI initiatives have moved beyond experimentation to the enterprise level. The study offers management lapses and broken processes as potential explanations. However, there is a technical problem many enterprises may be overlooking.
They’re missing a crucial layer of computing infrastructure.
A coordinating network layer, situated within a secure cloud environment, forms the vital connective tissue among a fleet of distributed devices. Supporting infrastructure keeps the full IoT ecosystem functioning, even when individual nodes go offline, enabling more reliable deployments at scale.
As AI models become more advanced and industry pressure intensifies, now is the time for organizations to reevaluate their IT infrastructure for next-generation resilience.
How the edge and the cloud work together
There are clear advantages to an edge-first AI strategy in logistics automation. Running AI on-device enables speed and precision in ways that cloud-only AI infrastructure cannot.
For example, a quality control device on the warehouse floor must be able to recognize defects before an item passes to the next phase of distribution–a cloud round-trip could delay detection and flagging, potentially allowing a defective item to slip through. Keeping the full lifecycle on-device speeds time to reaction, catching the error before any downstream impacts can occur.
Amid a global computing crunch, enterprises are also drawn to the financial advantages of the edge. Deploying AI applications closer to the data source, whether an IoT device or on local hardware, can reduce overall AI costs. That said, it has to be the right kind of AI.
Edge AI is fundamentally different from cloud AI, and the same general-purpose strategies applied to cloud AI typically fall apart in an edge environment. On-device deployment works best with purpose-built, specialized models: lightweight small-language models (SLMs) that are more precise and accurate than general vision models and sweeping, do-it-all LLMs.
But these models still need some cloud support to deliver the best results. On-device AI requires continuous, on-time updates to ensure models are optimized to the devices they run on. With a cloud-based network layer, updates are delivered and installed on time, across the full ecosystem of devices, which can number in the hundreds of nodes at large enterprises.
Edge AI also requires uninterrupted connectivity to function reliably. Without a cloud backup, when the device or immediate network – such as a private 5G network inside the warehouse – goes down, so does the device. Downtime and heavy latency can have serious consequences: missing items, costly delays, lost productivity, and even risks to workers’ safety. Network support keeps critical functions online until the malfunctioning device can be manually reset or repaired, while delivering regular updates to prevent such issues in the first place.
A supporting network layer also bolsters device security. While edge computing is known for being more secure than the cloud, distributed devices expand the potential attack surface, giving bad actors more on ramps to compromise operations. Secure networking across devices, combined with greater data governance and controls, minimizes edge vulnerability.
Finally, the network layer provides the necessary scaffolding for cross-border AI expansion. Global, low-latency networks can scale operations without compromising security or digital sovereignty. When supply chains are expanding their international sprawl to navigate economic and geopolitical challenges, global cloud support to back their distributed devices is critical to remaining compliant and while maintaining consistent performance.
It’s no longer edge vs. cloud, but rather the edge and cloud working together, functioning as two parts of a full stack infrastructure strategy that supports high-ROI edge deployments.
The composable cloud advantage
IoT devices regularly collect data for use in other AI-enabled business processes, such as demand projection for fulfillment, labor planning, and procurement. A supporting network layer promotes integration with other supply chain AI and software tools, so enterprises get the advantages of a robust data stream in real time.
Supporting such a high volume of data requires always-on networks, expandable storage options and flexible orchestration. A single-cloud hyperscaler or edge-only strategy won’t offer these features at the scale that complex modern supply chains demand; conversely, enterprises may be locked into features they don’t need, driving up costs at a time when margins are already tight.
As a result, there is a growing trend of enterprises adopting composable clouds for more sustainable, cost-efficient deployment. Consider the SLMs that run on edge devices. In a cloud environment, running and training these lightweight models on the same infrastructure designed for sprawling LLMs and entire business computing backbones is woefully inefficient. Both at the edge and in the cloud, infrastructure should be proportional to what it supports, with the flexibility for scale as models advance and organizations become more AI-mature.
With a composable cloud, enterprises can build bespoke infrastructure for their edge deployments. Instead of adopting a service and hoping that it fits with their vast network of distributed devices, they can implement infrastructure as a direct solution to real problems. This is where tangible business outcomes emerge, transforming edge AI from a promising innovation into a lucrative investment.
Infrastructure for the next-generation warehouse
AI technology is advancing faster than organizations can operationalize it. As edge AI deployments struggle to transcend the experimental stage, AI agents, which are already transforming back-office operations, are now coming to the warehouse floor.
Bringing agents to the edge could vastly uplevel automation. Deployed on source devices, agents can analyze data, make decisions and execute tasks far more quickly than if they were running via the cloud. Even so, just 37% of supply chain leaders trust agents to execute end-to-end processes in logistics operations. While they can act autonomously, agents still need to be monitored, and with access to sensitive data, are subject to the same compliance constraints as any other AI system or software program. A global, low-latency network layer protects compliance and allows for scale.
In supply chain operations, agents likely won’t operate in isolation, but as actors within multi-agent systems. Agent-to-agent communication requires coordination across devices within and outside the warehouse, generating the need for a cloud-based network to ensure information is conveyed securely, in real-time, across the full ecosystem.
Edge AI readiness, especially in the agentic era, starts with the right infrastructure. To build a foundation for continuous performance, enterprises must understand the technology in play, how it behaves, and what limitations it has and reconfigure their infrastructure accordingly. This means looking beyond the longstanding dichotomy of edge vs. cloud to compose a solution that will maximize efficiency, guarantee compliance, and accelerate scale.
