ECOC 2025 industry reflections - Part 2

Part 2: More industry figures share their thoughts after attending ECOC 2025 in Copenhagen: Lumentum's Rafik Ward, Scintil Photonics' CEO, Matt Crowley, Rob Shore of Nokia, and Jeff Hutchins of the OIF and Ranovus. 

Nokia's stand at ECOC.

Rafik Ward, Senior Vice President, Chief Strategy Officer and Chief Marketing Officer at Lumentum

At ECOC 2025, I saw a nice uptick in the discussion about co-packaged optics with a clearer view to commercialisation.

While co-packaged optics has been discussed for nearly a decade, we are moving from a world of pretty slides and architecture diagrams to one where increasingly large statistical data sets are shown. Meta showed co-packaged optics data over 15 million port hours with significant improvements with regard link failures. Similarly, Broadcom has been showing statistical performance across increasingly large datasets in the last year.

At Lumentum, we are doing our part to be ready for co-packaged optics with the announcement of our External Laser Small Form-Factor Pluggable (ELSFP) module, which supports eight lasers, each with 400mW class UHP (Ultra High Power) pump lasers. These 1310nm lasers stem from our heritage as a pump laser supplier to the telecommunications industry, where we have supported long-haul and submarine networks for decades. Thus, they are inherently reliable as they have been deployed into the more stringent part of the network, each carrying many channels.

There was a nice buzz at the show, which I have not felt in a while. Companies focused on optics outside the data centre are seeing a very healthy lift in their business and are proposing new ways of doing things, driven by trends like hollow-core fibre, next-generation in-line amplifiers and high-speed coherent optics.

Inside the datacentre, new architectures that have been discussed for many years, including elements like optical circuit switches (OCSes) and co-packaged optics, are gaining momentum. It is an exciting time to be in the optics industry.

Matt Crowley, CEO, Scintil Photonics

ECOC 2025 shifted two pieces of ‘conventional wisdom.’ First, regarding co-packaged optics reliability: datasets shared by Meta and Broadcom indicate that, for reliability, co-packaged optics is surpassing pluggable optics. Assuming those results hold across large-scale deployments, reliability concerns around the risks of deploying co-packaged optics should decline. Indeed, given the power and latency advantages, this should advance co-packaged optics adoption. After all, who doesn’t want a more reliable, lower-power network?

Second, dense wavelength-division multiplexing (DWDM) laser precision: multiple discussions made clear that fundamental process capabilities are the gating factor for DWDM deployments at scale. Traditional distributed feedback laser (DFB) arrays clustered around ±50GHz spacing tolerance are misaligned with ~±10 GHz DWDM requirements. You can work around these shortcomings in a small-scale demo by binning parts; you can’t bin your way into high-yield volume with tight channel uniformity and lifetime stability. New solutions for DWDM co-packaged optics must have the manufacturing precision to achieve high yields off the wafer if they are to scale in volume.

My overall ECOC takeaway? The winners in the next optical cycle will pair high reliability (including co-packaged optics where it fits) with precision manufacturing to deliver precise wavelength control of ±10 GHz at high yield. That combination is an absolute requirement for next-generation DWDM co-packaged optics network architectures.

What caught my attention at the show was the maturation signal: reliability curves trending the right way for co-packaged optics, and an industry-wide recognition that precision laser source manufacturing is now the central lever for scaling optics for AI.

Rob Shore, Head of Portfolio Marketing, Optical Networks, Nokia

One trend I’d note is the rise of 800-gigabit coherent pluggables. Eight hundred gigabit is the primary solution to address data centre interconnect applications. There is now also a massive demand for 800-gigabit coherent pluggables to address the new and rapidly growing ‘scale-across’ application, where workloads are shared across data centres.

Much of the discussion on next-generation optics is on higher modulation rates and smaller digital signal processor (DSP) node processes (2nm CMOS). New types of modulator technology were discussed at the conference and show as challenges with silicon photonics mount for higher baud rates. These modulator technologies include indium phosphide, thin-film lithium niobate, and thin-film lithium tantalate.

These is also increasing interest in optical fibre: hollow-core and multi-core fibre to reduce latency and fibre-count in campus data centre interconnect and intra-data centre applications.

Another area is multi-rail optical line systems to meet the power and space limitations of optical line station amplifier (OLA) huts. The target has shifted from some 100 fibres per OLA rack to 1,000 fibres per OLA rack.

Putting my marketing hat on, Nokia demonstrated two key solutions. We were part of the OIF multi-vendor interoperable, multi-haul 800 gigabit coherent pluggable demonstration. And we showed the ICE-D-based LPO intra-DC connectivity solution. ICE-D will help bring a deployable LPO solution to market, enabling power savings of up to 80 per cent.

ECOC 2025 focused on ideas to bring us into the next generation of optics for both inter-data centre and intra-data centre connectivity. And while there was a broad collection of innovative topics, the key takeaway is that the next generation of optics is at least 18 months to two years away.

Jeff Hutchins, the OIF’s PLL Work Group, Energy Efficient Interfaces (EEI) vice chair, and Ranovus

I spent most of the time at the OIF booth and so did not have much time to see other events.

I did note 400-gigabit/lane demonstrations such as the one by Keysight. More companies were promoting silicon photonics offerings for optical modules. And there were lots of thin-film lithium niobate foundries and other technologies for high-speed modulators that could be used for 400-gigabit optical lanes.

Meta’s presentation about the failure rate and link flaps was interesting with respect to co-packaged optics, suggesting that co-packaged optics would offer much better performance, especially for scale-up, where retries are expensive. Meta also gave a similar presentation at the more recently held OCP Global Summit event and suggested that they suspect the reason for the improvement was the reduction in human touch.

I also learned from Broadcom and Corning’s joint presentation regarding fibre connectivity and associated failure rates. Corning showed failure rates for fibre array units that were very good. Also, Terahop showed at the OCP Global Summit very good FIT (failures in time) rates for silicon photonics-based modules. This topic is key as the hyperscalers want excellent link and hardware reliability for scale-up networks.