- 400G wave adoption poised to surge by 2029, per Ciena and Vertical Systems Group
- Most U.S. enterprises now are using 100G circuits, which will see a “steady rise” as well
- MOFNs are all the rage as hyperscalers ramp their data center buildouts
The AI wave is imminent, and wavelength services must accordingly keep up. While enterprises today are mainly using 100G wave circuits for transferring large amounts of data, it won’t be long before 400G usage ramps, according to a new Ciena report.
“Bandwidth growth in 400G circuits is set to soar” by 2029, said Ciena, using data from Vertical Systems Group (VSG). Given multiple 10G circuits are often replaced by fewer 100G circuits, the same pattern will likely follow as enterprises transition from 100G to 400G connectivity.
Unsurprisingly, hyperscalers and data centers are fueling the demand for 400G waves. FiberLight, for instance, got 64 orders in March alone for 400G wavelength “specific to AI,” CEO Bill Major previously told Fierce. Other long-haul operators like Windstream and Zayo have also seen an uptick in fiber strand counts to accommodate AI workloads.
But Ciena noted multiple sectors, like finance, healthcare and education, are still using 100G waves and will likely keep doing so along with 10G.
100G circuits, which currently make up three-quarters of the U.S. wave market, “will see a steady rise, while 10G circuits will experience modest growth,” the report said.
Of course, Ciena isn’t the only major supplier of wave tech, as Infinera, Nokia and Cisco were also the top vendors in the space in 2024, per VSG. Given Nokia just acquired Infinera’s optical biz, the Finnish company could have room to grow its market share.
Much ado about MOFN
Ciena’s report also highlighted the increased prevalence of managed optical fiber networks (MOFNs), where operators build dedicated private optical networks for cloud providers.
MOFN itself is nothing new, but Lumen is probably the biggest example of the technology entering the limelight. Lumen in the last year inked network transport deals with cloud titans Amazon Web Services (AWS), Microsoft and Meta. Plus, the operator is working with Google Cloud to build new 400G on-ramps.
Hyperscalers can have MOFNs across different regions and from different service providers, Ciena said. But that also introduces network monitoring challenges, as more MOFNs make it tricky for hyperscalers to “track performance with end-to-end network visibility.”
Why does MOFN matter so much? It’s merely another tool in the box for hyperscalers to aid AI workloads beyond the data center building, said Dell’Oro analyst Jimmy Yu.
Hyperscalers historically “leaned more towards buying their own equipment,” he said. But collaborating with a service provider gives them more flexibility in creating a network that’s tailored to their needs.
“This is especially important since the number of data centers is increasing,” Yu added.
We can expect to see 39 new hyperscaler data centers built in 2025, said Ciena's report, “the largest single-year data center deployment plan ever.”