Keysight Technologies is bringing an end-to-end testing strategy for next-generation AI data centers to ECOC 2026, as optical networks move toward 1.6T and eventually 3.2T links. At the September 21–23 event in Málaga, Spain, Keysight will demonstrate tools spanning photonic design, 220 GHz characterization, AI infrastructure validation, optical-link research and automated 1.6T production testing. The broader push reflects a growing industry challenge: as AI clusters demand more bandwidth, validating the optical and electrical infrastructure connecting those systems is becoming almost as important as designing it.
The next bottleneck in AI infrastructure may not be the GPU. It may be the ability to prove that the network connecting thousands of accelerators actually works at scale.
That is the problem Keysight Technologies plans to address at the European Conference on Optical Communication (ECOC) 2026, where the company will showcase technologies for designing, characterizing, validating, benchmarking and manufacturing the optical infrastructure required by next-generation data centers.
The company will exhibit at booth #1154 at FYCMA in Málaga from September 21 to 23.
The timing is significant. AI data centers are rapidly moving from 400G and 800G connectivity toward 1.6T, while research and development is already targeting 3.2T optical links. LightCounting expects 1.6T optical-transceiver shipments to reach tens of millions of ports in 2026, with 1.6T chipset sales exceeding $2 billion this year.
Keysight’s strategy is therefore aimed at an emerging infrastructure layer: the measurement and validation stack required to make those speeds deployable.
AI Is Pushing Optical Networks Into a New Speed Class
The bandwidth requirements of AI clusters are fundamentally different from those of conventional enterprise computing.
Large training and inference systems connect huge numbers of GPUs and accelerators, creating traffic between processors, memory and networking infrastructure. As accelerator performance increases, the network can become a constraint unless interconnect bandwidth scales alongside compute.
The industry is consequently moving toward higher-speed electrical and optical interfaces.
The Ethernet Alliance says 1.6T Ethernet is gaining momentum as AI workloads increase network requirements, with IEEE P802.3dj advancing specifications for 1.6 Tb/s Ethernet.
Keysight is positioning its test portfolio around that transition.
Its existing 1.6T infrastructure covers everything from simulation and pathfinding through validation, compliance and manufacturing test. The company’s platform supports 224 Gb/s signaling for current 1.6T implementations and extends toward 448 Gb/s-per-lane technology for future 3.2T systems.
That progression illustrates why testing is becoming harder.
Higher data rates increase sensitivity to signal integrity, optical characteristics, electrical losses, interoperability and manufacturing variation. Engineers cannot simply validate one component in isolation and assume the complete link will behave the same way.
Bridging Photonic Design and Manufacturing
One of Keysight’s ECOC demonstrations will focus on connecting photonic simulation with automated photonic integrated circuit testing.
The goal is to reduce the gap between design and physical validation.
Photonic integrated circuits increasingly combine optical and electrical functions on compact components. Engineers need to understand how those components behave under real operating conditions, while manufacturers need repeatable automated tests that can be applied at scale.
Keysight says its ECOC demonstrations will include automated PIC testing and characterization reaching 220 GHz.
That matters because the test environment itself has to keep pace with the components being developed. A measurement system that cannot capture the relevant bandwidth or automate enough of the workflow can become a bottleneck in development and production.
Keysight has already been moving toward unified electrical-optical-electrical workflows. In May, the company introduced an EOE simulation capability intended to allow engineers to analyze electrical-to-optical-to-electrical signal chains in a single environment.
The objective is to catch problems earlier, before designs reach silicon or physical integration.
3.2T Research Starts Before 3.2T Deployment
The other side of the equation is preparing for what comes after 1.6T.
Keysight will demonstrate technologies aimed at 3.2T optical research alongside 1.6T transmitter and receiver validation.
At its OFC 2026 demonstrations earlier this year, Keysight showed 448 Gb/s optical research capabilities using 224 GBaud PAM4 signals and up to 120 GHz bandwidth, alongside 1.6T optical transmitter validation.
The distinction between research and production is important.
A laboratory demonstration can establish that a signaling technology is possible. Commercial deployment requires engineers to prove repeatability, interoperability, error performance, thermal behavior and manufacturing yield.
That is why companies building AI infrastructure increasingly need testing platforms that cover the entire path from device development to system deployment.
Testing the AI Fabric, Not Just the Optical Module
Keysight is also moving its validation work higher up the infrastructure stack.
At ECOC, the company plans to demonstrate methods for validating high-speed AI and data-center interconnects, benchmarking AI fabric performance and emulating real-world workloads.
This reflects an important change in how networking hardware is evaluated.
Traditional optical testing can establish whether a transceiver meets its electrical and optical specifications. AI infrastructure operators ultimately need to know something broader: does the complete network deliver the performance required by the workload?
A link can pass a component-level test and still create problems when deployed inside a complex AI cluster.
Keysight’s approach is therefore increasingly system-oriented, connecting physical-layer measurements with traffic generation, workload emulation and network-level performance analysis.
Its AI scale-out portfolio already includes a 1,600GE traffic emulator, functional interconnect testing and 1.6T optical transmitter compliance capabilities.
1.6T Changes the Economics of Production Testing
Moving to higher optical speeds creates another problem: testing can become expensive and time-consuming if every device requires lengthy manual measurement.
That is particularly important as 1.6T optical components move from engineering samples toward high-volume manufacturing.
Keysight plans to demonstrate automated interconnect validation designed to reduce test time, improve yield and accelerate production ramps.
Automation can matter as much as raw measurement performance at this stage.
If manufacturers need significantly longer to test each optical module, faster component production does not necessarily translate into faster data-center deployment. Testing therefore becomes part of the manufacturing economics of the AI infrastructure supply chain.
AI Is Also Entering the Test Workflow
Keysight will also highlight AI-enabled test automation, including tools designed to simplify test development, troubleshooting and data analysis.
This creates an interesting feedback loop.
AI is driving demand for faster networks, while AI itself is increasingly being used to help engineers manage the complexity of testing those networks.
For engineering teams, the potential benefit is less about replacing measurement equipment and more about reducing the human effort required to configure tests, identify anomalies and interpret large quantities of measurement data.
That becomes increasingly valuable as the number of test parameters grows with each generation of optical technology.
The Bigger Industry Shift
Keysight’s ECOC agenda mirrors a broader change across the optical communications ecosystem.
The industry is moving from a world in which individual components could be optimized relatively independently toward one in which electrical interfaces, optical components, photonic devices, networking protocols and workloads have to be validated as interconnected systems.
OIF’s 2026 interoperability work similarly emphasizes multi-vendor validation for AI-era networks, including 448G electrical signaling, co-packaged optics and other technologies intended to reduce integration risk.
Other optical suppliers are also framing ECOC 2026 around AI infrastructure and post-1.6T connectivity. Coherent, for example, is highlighting next-generation AI interconnects, co-packaged optics and technologies beyond 1.6T at the event.
The implication is clear: 1.6T is becoming less of a future research target and more of an engineering and interoperability problem.
For hyperscalers, networking vendors and optical manufacturers, the ability to validate these systems quickly could determine how rapidly the next generation of AI infrastructure reaches production.
Keysight’s pitch at ECOC is ultimately about shortening that path—from photonic design and laboratory research to compliance, manufacturing and full-system AI workload validation.
As AI clusters become larger and more bandwidth-intensive, the companies that can measure, validate and troubleshoot every layer of the network may become as important to AI infrastructure scaling as the companies building the accelerators themselves.
Market Landscape
The optical networking market is entering a major transition as AI infrastructure pushes bandwidth beyond 800G.
- 1.6T is moving toward deployment: The Ethernet Alliance says the 1.6T Ethernet ecosystem is gaining momentum, while IEEE P802.3dj continues work on the associated specifications.
- 3.2T is already in research: Keysight and other photonics vendors are developing 448G-per-lane technologies aimed at the next generation of AI networking.
- Interoperability is becoming critical: OIF’s multi-vendor demonstrations show the industry’s focus shifting toward validating complete interoperable building blocks rather than individual components.
- Testing is becoming an infrastructure discipline: Engineers increasingly need unified electrical, optical and system-level validation instead of separate measurements that are manually correlated.
- Competition is expanding beyond transceivers: Vendors such as Coherent, Broadcom, Marvell and others are competing across optical components, interconnects, photonic integration and AI networking architectures.
Top Insights
- Keysight is targeting the testing bottleneck emerging as AI data-center networks move from 800G toward 1.6T and future 3.2T connectivity.
- Its ECOC demonstrations span photonic design, optical measurement, AI fabric validation, workload emulation and automated manufacturing test.
- 1.6T deployment increases the importance of interoperability, signal integrity and system-level validation across electrical and optical domains.
- Automated testing could become a major factor in optical manufacturing economics as hyperscalers demand high volumes of next-generation transceivers.
- AI is beginning to reshape the testing process itself, helping engineers automate test development, troubleshooting and analysis.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
