800G for 200 Kilometers Without a Traditional Optical Shelf: Is the DCI Box Disappearing?
A data-center link that once required routers, gray optics, transponders, muxponders and a dedicated optical transport chassis can increasingly be built with 800G coherent optics plugged directly into the router plus a compact optical line system. New hardware can carry as many as 32 × 800G wavelengths — 25.6 Tbps — across a single fiber pair over approximately 2–200 km. So which boxes really disappear, and which ones are merely getting smaller?
The traditional optical shelf is not vanishing everywhere. But for a growing class of straightforward data-center interconnects, it is losing one of its oldest jobs: converting Ethernet traffic from the router into a long-distance coherent DWDM signal. That conversion can now happen inside an 800G QSFP-DD or OSFP module installed directly in the router or switch. Meanwhile, amplification and other line-system functions are beginning to move into pluggable modules too.
Is the DCI Box Really Disappearing?
The standalone transponder shelf increasingly can disappear. The optical network cannot.
An 800G ZR or ZR+ optic can generate a tunable coherent DWDM wavelength directly from an 800G router or switch port. That removes the classic gray-optic-to-transponder conversion stage.
The fiber pair, wavelength multiplexing, optical power budget, amplification where needed, monitoring and route engineering still exist. The difference is that those functions can increasingly be implemented with passive components and small pluggable modules instead of a large dedicated transport chassis.
Traditional DCI vs the New Pluggable Architecture
Traditional Transport Architecture
The router hands a short-reach client signal to a dedicated transport platform. The optical platform creates and manages the long-distance wavelength.
Router-Hosted 800G Architecture
The coherent wavelength is created in the router. The dedicated electrical-to-optical transport conversion layer is eliminated.
The Most Important Box That Disappears Is the Transponder
In a conventional design, an Ethernet router port typically connects through a short-reach “gray” optic to a transponder or muxponder inside an optical transport shelf.
The transponder receives that Ethernet signal and converts it into a high-performance coherent wavelength suitable for DWDM transmission.
800G ZR/ZR+ changes the sequence because the coherent DSP, modulator, receiver, FEC and tunable laser are integrated into a module small enough to plug directly into the router.
No separate short-reach optical hop is required between router and transponder.
The coherent conversion happens directly inside the router-hosted optic.
A simple point-to-point design may no longer need a chassis just to host transponders.
DWDM filtering, amplification and fiber engineering remain part of the system.
Now the Line System Is Shrinking Too
Coherent’s September 2026 Pluggable Optical Line System expansion pushes another traditionally chassis-based function into the pluggable world.
The enhanced POLS uses a full-C-band high-power variable-gain amplifier packaged in a QSFP form factor. It is designed around modern 800G ZR/ZR+ coherent optics and targets links from roughly 2 km to 200 km.
In practical deployments, passive wavelength multiplexing and demultiplexing are still required. The breakthrough is that the active amplification and management functions no longer necessarily require their own large shelf.
Why 32 Wavelengths Matters
DWDM allows each coherent transceiver to operate on a different wavelength. Those wavelengths are combined onto the same physical fiber pair.
An enterprise therefore does not need 32 long-distance fiber pairs to carry 32 × 800G. Under the Coherent POLS architecture, those channels can share one pair.
But 800ZR and 800G ZR+ Solve Different Problems
| Characteristic | 800ZR | 800G ZR+ |
|---|---|---|
| Core target | Standardized single-span DCI | Higher-performance metro / regional transport |
| 800G wavelength | Yes | Yes |
| Direct router insertion | Yes | Yes |
| OIF standardized DCI target | 80–120 km amplified | Beyond basic ZR envelope |
| Programmable line rates | More constrained | Often 400G / 600G / 800G modes |
| Longer-reach operation | Limited | Major advantage |
| ROADM environments | Primarily single-span DCI | Much better fit |
200 Kilometers Is Not the Limit of ZR+
This is an important distinction because the two numbers are easy to confuse.
The approximately 2–200 km figure describes the target application envelope of Coherent’s compact POLS architecture.
It is not the physical limit of an 800G ZR+ transceiver.
Cisco currently specifies amplified 800ZR links up to 120 km and 800G ZR+ links beyond 1,000 km under suitable optical-line-system conditions. Juniper’s current 800G coherent platform likewise supports metro through regional and longer-reach operation, including lower-rate modes extending beyond 2,000 km.
Where Each Architecture Starts to Make Sense
What the Traditional Shelf Still Does Better
| Function | Pluggable-First Architecture | Traditional Optical Platform |
|---|---|---|
| Point-to-point Ethernet DCI | Excellent | May be excessive |
| Native 800G transport | Excellent | Excellent |
| Client grooming | Router dependent | Muxponder / OTN strength |
| OTN switching | Not inherent | Strong advantage |
| Multi-degree ROADM mesh | Requires external photonics | Purpose-built |
| Raman amplification | Usually external | Integrated options |
| Integrated OTDR | May require separate tools | Available on large systems |
| Legacy 10G / 100G service aggregation | Router dependent | Strong muxponder use case |
| Maximum spectral efficiency | Very strong | Embedded engines can push farther |
| Operational simplicity for 2 sites | Major advantage | Potential overkill |
The Best Use Case Is Almost Boring: Two Big Data Centers
The cleanest architecture is a pair of large facilities connected by one or two known fiber routes.
Excellent Candidate
- Two fixed data-center endpoints.
- Dark fiber already available.
- 10–200 km route.
- Routers support 800G coherent optics.
- Primarily Ethernet traffic.
- No intermediate ROADM sites.
- No heavy OTN grooming requirement.
- Capacity measured in multiple 800G wavelengths.
Traditional Platform More Likely
- Many intermediate network sites.
- Complex optical mesh.
- Multi-degree ROADMs.
- Large mix of 10G/100G/400G clients.
- OTN switching and grooming.
- Complex protection schemes.
- Long-haul backbone functions.
- Integrated optical test and monitoring requirements.
The Economics Are About More Than Removing One Box
Fewer Optical Conversions
Removing gray client optics and separate transponders means fewer interfaces, fewer modules and fewer points of failure.
Less Rack Space
A dedicated transport shelf can consume valuable rack units in both data centers. Pluggable architectures move more functionality into equipment already installed.
Lower Power
Coherent pluggables are specifically designed to operate in router/switch thermal envelopes rather than requiring full transport-card power budgets.
Fewer Management Systems
Optical telemetry can increasingly be exposed directly through the router, reducing the separation between IP and transport operations.
Faster Turn-Up
A tunable coherent pluggable can be provisioned like a router interface instead of requiring a separate transport-chassis deployment.
Simpler Spares
Pluggable modules are easier to stock and replace than proprietary chassis cards, although exact vendor qualification still matters.
The Management Layer Changes Too
Traditional architectures divide troubleshooting between the router team and the transport team.
Router-hosted coherent optics blur that boundary.
Juniper, Cisco, Nokia and other vendors expose coherent parameters through the host router, including optical transmit and receive power, chromatic dispersion, FEC performance and other line statistics.
But Concentrating 25.6 Tbps Creates a Different Risk
Thirty-two 800G wavelengths on one fiber pair is an impressive demonstration of fiber efficiency.
It is also 25.6 Tbps riding through one physical failure domain.
A transponder failure affects a wavelength. A severed conduit can affect all 32.
At 6.4 Tbps, the Architecture Becomes Easy to Visualize
Consider a company needing 6.4 Tbps between two data centers.
At 800G per wavelength, the requirement is eight wavelengths.
A conventional design could use eight coherent transport channels in a dedicated optical platform. A routed-optical design can instead use eight 800G coherent pluggables installed directly in router ports, combined onto a fiber pair through the optical line system.
Add a physically diverse route and the design becomes 16 wavelengths total: eight on Route A and eight on Route B.
The Standards Roadmap Suggests This Is Only Beginning
Standardized coherent DCI begins moving into small router-hosted pluggables.
OIF defines interoperable single-wavelength 800G DCI for amplified 80–120 km links.
Active optical-line functions move into QSFP-sized modules for compact 2–200 km DCI.
OIF has already standardized 1.6 Tbps per wavelength for amplified DCI up to 120 km.
The Next Box to Shrink May Be the 1.6T Box
OIF released its 1600ZR Implementation Agreement in September 2026.
The specification doubles the capacity of 800ZR and defines an interoperable 1.6 Tbps coherent wavelength for amplified point-to-point DCI links up to 120 km.
That does not mean every data center will install 1.6T coherent optics immediately. It does show that the architectural direction is continuing.
Ten Questions Before You Remove the Optical Chassis
| Question | Why It Matters |
|---|---|
| Does our router support the exact 800G ZR/ZR+ optic? | Verify power class, thermal budget, firmware, licensing and vendor qualification. |
| What is the actual fiber distance? | Route mileage and straight-line distance can differ substantially. |
| What is the measured end-to-end loss? | Optical feasibility depends on loss, not simply kilometers. |
| How many wavelengths will we eventually need? | Design the mux/demux and amplification plan around future capacity, not only day one. |
| Do we need OTN grooming? | If yes, removing the transport shelf may remove a function the network actually needs. |
| Are there intermediate ROADMs? | Multiple photonic nodes increase optical loss and OSNR complexity. |
| Do we need legacy 10G/100G aggregation? | Traditional muxponders can remain efficient where many lower-rate clients must be packed together. |
| Who owns optical troubleshooting? | Eliminating the shelf does not eliminate the need to understand OSNR, FEC and optical power. |
| Is Route B truly diverse? | Verify conduit, bridge, manhole and building-entry diversity—not just different circuit IDs. |
| What happens at 1.6T? | Ensure today’s line system and fiber plan do not unnecessarily block the next wavelength generation. |
Reach claims in coherent networking are design-dependent. Actual transmission distance depends on fiber attenuation, connector and splice loss, channel count, optical launch power, receiver sensitivity, OSNR, amplifier noise, ROADMs, dispersion, FEC and operating margin. The approximately 2–200 km figure in this report refers to Coherent’s stated POLS application range, not the maximum reach of 800G ZR+ technology.
DCI Shelf-or-Pluggable Architecture Checker
Enter your DCI requirement. The model estimates wavelength count and determines whether the design resembles direct 800ZR, compact 800G ZR+ + pluggable line system, a more engineered open-line-system deployment, or a project where a full optical transport platform still makes sense.
| Architecture | Endpoint Equipment | Line / Shelf Equipment | Router Port Cost | Modeled Initial Cost |
|---|
Planning heuristic only. This does not calculate a real optical link budget. Actual deployment requires validation of fiber loss, connector/splice loss, OSNR, channel plan, launch power, amplifier gain, dispersion, FEC, ROADM penalties, transceiver interoperability and route diversity. Equipment cost fields intentionally default to zero so the calculator does not invent market pricing.
