EnglishViews: 9999 Author: Site Editor Publish Time: 2026-06-16 Origin: Site
Why Expanding AI Compute Capacity Relies More Heavily on High-Speed Interconnect Fabric
In recent years, a clear trend has emerged in the AI industry: GPU performance continues to improve, AI cluster scales continue to expand, and network switching capacity grows rapidly. However, one often overlooked reality: an AI deployment’s overall throughput is not dictated solely by GPU compute performance.
As training clusters scale from hundreds to tens of thousands of GPUs, massive volumes of data need to flow at high speed between servers, switches, and storage nodes. In many cases, what restricts system efficiency is not raw computing power, but data transmission capability. Currently, 400G has become the mainstream deployment rate, 800G is rapidly gaining adoption, and 1.6 Tb/s (referred to as 1.6T hereinafter) has begun to enter the planning horizon for next-generation networks. As switch chip capacities continue to evolve, higher-rate interfaces such as 800G and 1.6T are gradually incorporated into next-generation data center network roadmaps.
However, increased bandwidth does not mean that all connectivity solutions fit every scenario. In actual deployments, enterprises need to consider not only speed indicators but also maximum transmission reach, power draw, TCO and operational overhead. Therefore, in the 1.6T era, how to select the right physical layer interconnect solution has become a core concern for data center and AI infrastructure builders.
The 1.6T Era: How to Choose Between DAC, ACC, AEC, and AOC?
Four major technical architectures are available for 1.6T interconnects: Passive Direct Attach Copper Cable (DAC), Active Copper Cable (ACC), Active Electrical Cable (AEC), and Active Optical Cable (AOC). Their core characteristics are compared below:
Type | Power Draw | TCO | Max Reach | Primary Deployment Scenario |
Passive DAC | Lowest | Lowest | Short | Intra-rack GPU/switch links |
Active Copper | Low | Medium | Medium | Medium Short-to-medium inter-rack |
AEC Active Electrical | Low | Medium | Medium | Multi-GPU server interconnections |
AOC Active Optical | Higher | High | Long | Cross-rack / cross-data hall |
Why is DAC Still the Mainstream Choice for AI Clusters?
In short-distance intra-rack interconnection scenarios, DAC delivers distinct advantages: drastically lower power draw than all active alternatives, minimal link latency, and superior cost competitiveness thanks to its passive design. For massive AI clusters with tens of thousands of links, full deployment of active cabling would drastically inflate power draw and overall TCO. Therefore, passive DAC has long served as the primary solution for short-reach high-speed interconnections between servers and switches.
When Should You Consider ACC or AEC?
When link length extends, or when elevated line rates introduce channel loss exceeding DAC’s operating limits, ACC and AEC deliver superior signal integrity performance. Equipped with built-in signal conditioning circuits, they boost signal margin and extend transmission reach while retaining relatively low power draw. For medium-reach inter-rack connections, ACC and AEC represent pragmatic options balancing performance and total cost of ownership.
When Does Optical Interconnection Offer More Value?
The strengths of optical cabling architectures grow more pronounced with extended transmission distances and expanded network footprints. Compared with copper cabling, optical cables feature lighter weight, lower signal attenuation and greater routing flexibility, making them better suited for long-reach and high-density deployments. For cross-rack, cross-zone or campus-scale data center interconnections, AOC and optical transceivers deliver far broader deployment flexibility. Nevertheless, given their relatively higher TCO and power draw, final selection still requires holistic evaluation of performance, reach and long-term ownership costs.
How Does Zhaolong Interconnect’s 1.6T Product Lineup Address Diverse Deployment Requirements?
All Zhaolong Interconnect 1.6T DAC, ACC and AEC assemblies leverage our in-house developed HSC-60 series high-speed twinax bare wire. We have built a comprehensive high-speed interconnect ecosystem spanning multiple technical architectures and form factors, catering to varying max reach limits, strict signal integrity thresholds and cross-vendor platform interoperability demands.
1.6T OSFP DAC: Purpose-built for 1.6 Tb/s high-speed interconnection demands, featuring ultra-low power draw, minimal latency and robust reliability. Ideal for intra-rack and other short-reach high-speed interconnection use cases. In high-density AI infrastructure deployments, it strikes an optimal balance between throughput performance and long-term TCO.
400G~1.6T ACC: For deployments requiring extended transmission reach or stringent signal integrity benchmarks, Zhaolong Interconnect supplies ACC and AEC product solutions covering line rates from 400G up to 1.6T. Integrated signal conditioning enhances signal margin while maintaining low power consumption, enabling stable high-speed transmission and expanding options for medium-reach high-speed interconnection.
Multi-form-factor interoperability: Zhaolong Interconnect products support mainstream packages including OSFP and OSFP-XD, enabling seamless compatibility with switches, NICs and server hardware from all major vendors. This significantly reduces operational overhead during system deployment and hardware upgrades.
Quick Selection Logic
First map the matching relationship between target link rate and required maximum transmission reach.
Where DAC’s operating range covers the deployment demand, prioritize DAC to secure the lowest power draw and optimal TCO.
When extended reach or excessive channel loss surpasses DAC’s effective operating range, evaluate ACC or AEC as alternatives.
For long-reach deployment scenarios, prioritize AOC or discrete optical module solutions.
In summary, high-speed interconnect selection in the 1.6T era does not follow a simple “faster is better” rule. It calls for holistic decision-making based on actual deployment reach, power budget constraints and cost targets. A clear grasp of each solution’s performance traits and effective operating range empowers pragmatic, cost-efficient infrastructure planning for AI builds.
Successful implementation of the above selection framework hinges on consistent product reliability and stable supply chain delivery. Zhaolong Interconnect’s decades of technical expertise and large-scale manufacturing capacity within high-speed interconnects provide the foundational support for mass-volume on-time delivery of our full suite of 1.6T DAC/ACC/AEC assemblies.
Company Strength: The Foundation of High-Speed Interconnect Capability
Competition in high-speed interconnect products extends far beyond raw bandwidth specifications. It hinges on core capabilities including proprietary cable design, precise signal integrity control, rigorous testing & validation, and mass-volume manufacturing & on-time delivery.
A pioneering domestic developer of high-speed interconnect cabling solutions in China, Zhaolong Interconnect has consistently tracked the evolution of Ethernet standards, boasting a full-spectrum product lineup spanning 10G through to 1.6T. We possess end-to-end in-house R&D and manufacturing covering high-speed twinax bare wire through finished cable assemblies.
Within the high-speed interconnect ecosystem, Zhaolong Interconnect actively advances industry collaboration: we partner with value-chain stakeholders to push the technical advancement of high-speed copper cabling and interconnect technology, while participating in the drafting of relevant MSA specifications and IEEE industry standards.
Catering to high-speed networking verticals including large-scale AI clusters, cloud computing and hyperscale data centers, the company continuously expands and optimizes its high-speed interconnect portfolio to deliver stable, high-performance connectivity solutions for global customers.