When to move to 400G wavelength services

Key takeaways

• A wave is a private channel: A wavelength service gives you a dedicated optical channel on a provider's fiber, and the provider runs the line system.
• Watch peaks and parallel links: Busy-hour utilization, yearly growth, and a stack of parallel 100G circuits show when a route needs 400G.
• 400G and 800G are standardized: IEEE approved 400 Gigabit Ethernet in 2017, and the first 800 Gigabit Ethernet standard in 2024, and OIF defines interoperable coherent optics for both.
• Data center spending is climbing: Dell'Oro Group estimates that revenue from direct purchases for data center interconnect grew 40% year over year in early 2026.
• Confirm the details first: Protection, handoff type, splitting into 100G services, and the route itself decide whether a 400G wave fits.

By the time backups start missing their window, any upgrade is still an order and a turn-up away. Teams caught at that point often add another 100G circuit because it fits the gear they already run, then pay for several links and manage each one.

Buyers weighing 400G wavelength services need answers that launch news written for carriers rarely gives, such as when 400G pays off, how it fits existing 100G gear, and what to confirm before signing.

Teams that time the move well watch each route's growth and plan the upgrade a contract cycle ahead. They treat 400G as a planned step for specific routes, with the signs and the right questions settled before the order goes in.

What a 400G wavelength service is

A 400G wavelength service is a dedicated optical channel on a provider's fiber that carries 400 Gbps between two points, and engineers often call it a wave or a lambda. Wavelength services at every rate work the same way. The provider runs the line system, including the amplifiers and multiplexers, and hands off a client port at each end, usually Ethernet. Your traffic rides its own wavelength, separate from other customers' channels on the same fiber.

A 400G wave typically carries 400 Gigabit Ethernet, which the IEEE standardized in 802.3bs and approved in December 2017. On the line side, providers carry 400G over coherent optics, which pack more data into each channel. The OIF's 400ZR Implementation Agreement defines an interoperable 400G coherent interface for amplified point-to-point DWDM links of 120 km or less, and the OIF developed it first for cloud-scale data center interconnect.

One fiber pair carries many of these channels. The current 400ZR specification defines 48 channels at 100 GHz spacing or 64 at 75 GHz, on the frequency grid that ITU-T Recommendation G.694.1 sets for dense wavelength division multiplexing.

Signs a route has outgrown 100G

No standard sets the point where 100G stops being enough. The signs show up first in the route's own data.

• Peaks hit the ceiling: Busy-hour traffic regularly crosses the threshold your team uses to plan upgrades, even when the daily average looks comfortable.
• Parallel links pile up: The route already runs two or more 100G circuits, and each new one adds cost, ports, and another link to manage.
• Jobs miss their windows: Replication, backup, or bulk transfers that used to finish overnight now spill into business hours.
• Growth outpaces the contract: Traffic forecasts show the route doubling before the current circuit term ends.
• New workloads are planned: A new data center, storage platform, or compute cluster is coming online at either end of the route.

When two or more of these signs show up on one route, price 400G against another 100G circuit.

100G vs 400G vs 800G at a glance

The 100G vs 400G decision, and the later step to 800G, comes down to capacity, standards maturity, and what each end of the route can accept.

The IEEE approved 802.3df for 800 Gigabit Ethernet in February 2024, and the OIF released its 800ZR specification in October 2024 for single-span, amplified DWDM links of 80 to 120 km aimed at data center interconnect. 800G wavelength availability depends on whether the provider's line system on that route supports it.

Data center interconnect and AI workloads

Data center interconnect (DCI) is driving much of the recent growth in optical transport. Dell'Oro Group estimated that revenue from direct DCI purchases grew 40% year over year in the first quarter of 2026. Separately, it forecast the overall optical transport market to grow 16% in 2026, driven by rising demand for network infrastructure from AI data centers.

AI workloads are pushing the trend further as the largest operators spread training clusters across multiple sites. Dell'Oro's Jimmy Yu said in December 2025 that interest in ZR+ optics had grown tenfold in a year, driven by the need to scale across data centers to build AI training centers. Cignal AI pointed to "the need for rapid, long-reach connectivity between AI clusters" as a driver of demand for 800ZR+ optics.

Enterprises feel the same pull at a smaller scale. Moving training data between sites, replicating models, and connecting compute in a second data center all add heavy, sustained bursts of traffic to a few routes. Those routes are where a high-capacity WAN earns its cost. The same data centers usually need private cloud links too, such as private connectivity into Azure.

When a 400G wave beats dark fiber

Dark fiber gives full control of a route, and on the heaviest routes it can carry many 400G or 800G channels on optics you own. A 400G wave fits better on routes that need one or a few channels. You skip buying and running the optical line system, the provider runs the amplifiers and line equipment, and the contract term is usually shorter than a typical dark fiber IRU.

A wave also fits routes where dark fiber isn't available, or where the traffic outlook is too uncertain for a long commitment. The full trade-off between owning the fiber and buying lit service comes down to cost, control, staffing, and contract length.

What to confirm before you order a 400G wave

A 400G order has more moving parts than a 100G one. Confirm these six details before signing.

• Protection: Ask whether the wave is protected, and whether the backup path runs on a physically separate route.
• Handoff type: Confirm the client interface at each end, such as one 400GbE port or several 100GbE ports, and whether the provider accepts 400ZR or ZR+ optics in your routers as an alien wavelength on its line system.
• Splitting into 100G: Ask whether the wave can be delivered as four 100G services, since the OpenZR+ specification supports up to four 100GBASE-R clients on one coherent line and support depends on the provider's equipment.
• The route itself: Get the path as a map, and check its distance, its separation from your other circuits, and whether its line system supports 800G later.
• Turn-up and project management: Confirm who manages the order through turn-up, what the provider needs from each site, and how the wave is tested before handover.
• Upgrade path: Ask how a later move to 800G would work on the same route, and whether it needs new equipment at your sites.

Put answers to these six points in the order they appear, so the provider designs the wave around them.

Build the route your traffic needs next

A 100G route that runs near its ceiling costs more each year in parallel links and missed windows. 400G wavelength services fix that when the route, the handoff, and the protection match what the traffic needs.

Momentum delivers point-to-point routes at 100G, 400G, and 800G, and the route decides the design. We recommend the right route even when it isn't ours. Where fiber is missing, we lay our own and splice into carrier networks, and one named project manager stays with each route through turn-up. The same team connects your data centers to the clouds they use, and one contract covers global connectivity and managed networks across the rest of your sites.

Pull the utilization history for your busiest routes and the growth you expect on each over the next three years. Then book a meeting, and we'll show which routes need 400G and how to get there.

FAQs

Can a 400G wave be split into 100G services?

It can, depending on the provider's equipment. The OpenZR+ specification supports up to four 100GBASE-R clients on one 400G coherent line, so a single wave can feed four separate 100G handoffs. That lets a team move to 400G on the line side and keep its existing 100G router ports. Confirm support with the provider before ordering, since not every transponder offers it.

Do you need new fiber for 400G?

Usually not, if the route already carries DWDM. One fiber pair carries many wavelengths, and the 400ZR specification defines 48 channels at 100 GHz spacing or 64 at 75 GHz. A 400G wave needs a free channel and a line system that supports coherent 400G over the route's distance. Older systems or very long spans may need new equipment, and the fiber itself can often stay.

What decides whether 800G is available on a route?

The line system and the distance decide it. 800G needs coherent optics and amplifiers that support it, and on the same fiber, a higher data rate reaches a shorter distance. The OIF's 800ZR specification, for example, targets single-span, amplified links of 80 to 120 km for data center interconnect. Longer routes and older line systems may not support 800G yet, so the answer varies route by route.

Should you skip 100G and go straight to 400G?

It depends on growth. If a route already needs more than one 100G circuit, or forecasts show its traffic doubling within the contract term, starting at 400G avoids paying for parallel links and a second migration. Routes with steady, moderate traffic can stay on 100G longer. Price both options over the full term, including ports and optics at each end.

How do you protect a 400G wave against a fiber cut?

Order a protected wave or a second wave on a physically separate route. A protected service moves traffic to a backup path when the primary fails, and the backup protects against a cut only where it shares no trench, conduit, or building entrance with the primary. Ask for both routes as maps and check where they run before signing.