Finding the ideal interconnect to bridge Edge to Core for AI

Which type of interconnect is most suitable in the Edge-to-Core network fabric?

Nov 5th, 2025

Dr. Carlos Berto, Director of Network Engineering at Axiom

As Edge computing continues to emerge, data is no longer confined to a centralized location. Edge computing infrastructure brings computation closer to users, devices, and sensors, processing data more efficiently in real time.

In order to fully leverage the benefits of edge computing, however, we will need to figure out how these workloads can be distributed more evenly. This will be heavily contingent on a defining aspect of the network fabric: the interconnect between Edge and Core. 

What to consider when choosing an interconnect for Edge-to-Core

Choosing the appropriate interconnect is an important decision with long term strategic implications because it affects how quickly data is transmitted, aggregated, and analyzed at the core. 

Before we can establish the data center link in the network fabric, we must take into consideration the following aspects: 

Applications

Different applications have different data requirements. AI model training is particularly data volume intensive as it peruses vast data sets to generate content in real-time. On the flip side, applications that have more efficient local processing require less data transmission overall.

Latency

Applications such as real-time analytics or guidance for autonomous vehicles can be extremely latency-sensitive. The compute intensive workloads can generate enough server-side latency to disrupt operations, so latency must be minimized to ensure fluid data transmission between points.

Power consumption

The level of power consumption is also another important factor to consider. Generally speaking, short to mid-range interconnects consume less power than long reach networks. Thus, achieving greater power efficiency is much more of a challenge in long reach networks.

Distance

The distance between edge and core varies by type of network and geolocation. No two networks and their bandwidth requirements are the same. More latency-sensitive applications require closer proximity from the edge to the user, while other applications aren’t quite as demanding.

Costs

Keeping costs in check is essential as the costs can stack up as network infrastructures scale. Cost of ownership is an obvious concern. As infrastructure expands, so does the hardware footprint. Combined with rising power consumption, costs are a growing concern with edge to core links.

Evaluating the pros and cons of the interconnect options

By taking the aforementioned concerns into consideration, we will have a more clear-cut idea on the type of interconnect that is needed.

Let's examine the pros and cons for each type of data center link: 

DACs

In the context of edge-to-core, the biggest argument in favor of DACs is that it offers the lowest latency out of all the interconnect options. DACs are also generally lower in cost, while boasting lower power consumption than its counterparts due to its passive design (non-power consuming).

While these are major benefits of DACs, the main drawback with DACs for edge-to-core is that DACs are much more susceptible to EMI (Electromagnetic Interference). This can degrade signal over longer distances. With limited reach, DACs are more ideal for short range edge to core links.

ACCs

As a variation of a traditional DAC, Active Copper Cables (ACCs) utilize a built-in, active signal driver or equalizer. This mitigates some of the signal loss that is caused by transmission via copper, which extends the maximum reach for an ACC by 2 or 3 meters further than a traditional DAC.

With longer reach, ACCs can be deployed at medium distances as well. However, as it uses active components (even with minimal power), they consume more power than fully passive DACs. They are also generally more expensive.

AOCs

Active Optical Cables (AOCs) simplify network configuration and cut down on costs as it integrates optical converters directly onto the fiber itself, instead of having to attach and detach separate optical transceivers on both ends of the fiber.

AOCs offer greater immunity to EMI, allowing data signals to be transferred over longer distances without signal degradation. This makes them a great fit in longer distance networks, even if they are more expensive and require higher power consumption with its active components.

AECs

Active Electrical Cables (AECs) are somewhat of a middle ground option between DACs and AOCs. It improves overall signal integrity and range for medium distance connections, all while having lower power consumption and being cheaper in cost compared to AOCs.

As a copper solution, however, its reach is limited compared to that of fiber counterparts such as AOCs. It is also an active solution that requires power, unlike the other passive interconnect solutions.

Transceivers/fiber:

Integrated assembly solutions such as DACs, ACCs, AOCs and AECs are not as effective in long reach networks spanning 40km or more. In these cases, transceiver and fiber integration is the most viable strategy.

The advantage of transceivers paired with fiber is that they can carry the signal over extremely long distances with low latency. The costs and power consumption are much higher compared to the short and medium range solutions, however, which is an obvious drawback for edge to core links.

Here's a visual representation of how these interconnect solutions stack up against each other in the edge to core network fabric:

Roles of interconnect solution in various use-case scenarios

Because of their unique characteristics, each interconnection solution has an ideal deployment to maximize its value in an edge-to-core network fabric: These use-cases offer the best balance between the attributes to consider and the practical deployment of these solutions: