How to Choose Between Multimode and Singlemode for Your Project

The two-question framework

If you remember nothing else from this article, remember this. Before you order a single spool of fiber, answer two questions in this order:

1. How long is the link, today and twenty years from now?
2. What do the optics on each end cost?

Everything else, including jacket type, fiber count, connector style, and termination method, flows from those two answers. We have watched contractors agonize over OM4 versus OS2 cable pricing while ignoring the fact that the SFP+ optics on the switches will cost ten times more than the entire cable. Get the framework right and the rest is detail work.

Distance: the hard physics

Distance is not negotiable. Fiber type sets a hard ceiling on how far light can travel before the receiver gives up. Here are the working numbers a network designer should have memorized:

Multimode reach

• OM3 at 10G: 300 meters
• OM4 at 10G: 400 meters
• OM4 at 40G/100G SR4: 100 to 150 meters
• OM5 at 100G SWDM4: 150 meters

Singlemode reach

• OS2 at 10G LR: 10 kilometers
• OS2 at 100G LR4: 10 kilometers
• OS2 at 100G ER4: 40 kilometers
• OS2 at 100G ZR: 80 kilometers and beyond

Notice the gap. Multimode tops out around 400 meters at 10G and shrinks dramatically as data rate climbs. Singlemode comfortably covers ten kilometers at any speed under 400G. If your cable run is anywhere between 400 meters and the far horizon, singlemode is the only honest answer. We cover the underlying physics in our Single-Mode vs Multi-Mode Fiber article.

Optics economics: where the real money lives

Cable cost is a rounding error. Transceivers are the budget. Here are typical 2026 street prices for branded optics:

For a 48-port leaf switch carrying 10G uplinks to a spine, the difference between SR and LR optics across 96 transceivers (48 pairs) is roughly $5,000 in favor of multimode. That savings dwarfs any cable cost difference.

This is why hyperscale data centers still buy oceans of OM4. Reach inside a row is short, port density is enormous, and SR optics economics are unbeatable. Outside the data center, the economics flip. A campus with three 10G LR optics per building beats fighting the multimode distance ceiling forever.

Test those expensive optics with a calibrated handheld optical power meter and a matched stabilized light source before you certify the link. Optics fail; document them out of the path early.

The four common project types

Data center inside one room

Default to OM4 multimode. Reach is short, density is high, and SR optics dominate. Order pre-terminated MTP/MPO trunks for the highest fiber count runs and breakout cassettes for the LC patch panels. Stock LC duplex patch cords for tight rear-of-rack patching. If you anticipate moving to 400G in the next refresh, evaluate OM5 SWDM optics, but for most sites OM4 will carry you through 100G fine.

Campus backbone, building to building

Default to OS2 singlemode. Once you cross a parking lot, you are past 100 meters and probably past 200. Singlemode gives you 10 km of headroom at 10G, 40G, and 100G with the same cable. Pull a high count, terminate what you need now, and leave dark fibers for whatever comes in 2030. Use a fusion splicer at the building entrance to land the OSP cable into ISP riser cable.

FTTH or PON deployment

Always OS2 singlemode. PON systems use 1310, 1490, and 1577 nm wavelengths and rely on singlemode characteristics for the splitter loss budget. Multimode is not even an option in modern XGS-PON or 25G PON gear. Our FTTH deployment guide walks through the cable selection in detail.

Single building riser, four floors

This is the hardest decision. Run lengths are around 100 to 200 meters, both fiber types work, and the optics question becomes the tiebreaker.

• If the IT team uses standardized SR optics in their existing data center, mirror that and run OM4. Fewer SKUs, lower spares cost.
• If the building may be repurposed or expanded, run OS2 to future-proof. The cable will outlive three generations of switches.
• If you have budget, run a hybrid trunk with both fiber types in one jacket. Costs slightly more per foot but eliminates the decision permanently.

The future-proofing trap

Every contractor we know has heard the phrase "let's just put in singlemode for future-proofing" on a 50-meter switch-to-switch link. Sometimes it is the right call. Often it is overspending.

Future-proofing only matters if the future actually arrives. Inside a 50-meter run between two switches in the same MDF, the cable will probably get replaced in the next refresh anyway. The pathway is accessible, the cable is short, and the cost of pulling new fiber later is trivial compared to the optics premium you pay today on every port.

Future-proofing is most valuable when:

• The pathway is sealed (concrete, conduit poured, drywall finished)
• The run is long enough that re-pulling is expensive
• The bandwidth trajectory is uncertain
• The cost premium is small relative to the project budget

For a 12-strand riser between MDF and IDF on floor 4 of a 20-floor building, run OS2. The walls are not coming down again. For a 50-meter trunk inside an open-frame data center, run OM4 and budget for replacement at the next switch refresh.

The decision matrix

Connector and termination implications

Once you choose the fiber type, several downstream decisions become easy:

If you chose multimode

• Connectors are typically UPC, never APC. Multimode does not use angled polish.
• Termination is fast: pre-terminated MTP trunks, OptiTap or LC pigtails, mechanical splices for emergency repairs.
• Cleaning is critical because the larger 50 micron core is more forgiving but VCSELs are sensitive to angular errors. Always clean with a one-click cleaner and verify with a microscope.
• Bandwidth is sometimes labeled in EMB (effective modal bandwidth) MHz·km. OM4 is rated 4,700 MHz·km, OM5 is rated 4,700 MHz·km at 850 nm and 2,470 MHz·km at 953 nm.

If you chose singlemode

• Connectors should be APC for any path carrying analog video, RFoG, or PON wavelengths. UPC is acceptable for digital data center singlemode but APC is becoming the default. See our APC vs UPC guide.
• Fusion splicing is the gold standard. Mechanical splices are emergency only. Plan for a core-alignment fusion splicer on the truck for any project over a few terminations.
• Bend radius matters more on G.652 than on G.657, but always honor the manufacturer minimum.
• OTDR testing should use both 1310 and 1550 nm. Bend events show up at 1550 first.

The cost of being wrong

Picking the wrong fiber type is one of the few mistakes that requires ripping cable out of the wall. The penalty for over-specifying singlemode is paying $50 extra per optic. The penalty for under-specifying multimode is repulling cable through finished pathways and explaining to the customer why the brand new gigabit Ethernet circuit will not light at 10G.

When in doubt, lean toward singlemode for OSP and backbone, and lean toward multimode only when SR optics economics clearly justify it. The market is moving toward singlemode everywhere it is technically practical, and that trend will continue as silicon photonics drives optic prices down across all wavelengths.

Documentation requirements

Once installed, document everything. Every fiber pair gets:

• Fiber type (OM3/OM4/OM5 or OS2)
• Connector type and polish (LC/UPC, SC/APC, etc.)
• Bidirectional OTDR trace at appropriate wavelengths
• Insertion loss measured with power meter and light source
• Length measurement
• Pathway and termination panel labels

This protects you when someone six years later asks "can this run handle 100G?" and you can answer with measured data, not a guess.

FAQ

Related Reading

• Single-Mode vs Multi-Mode Fiber Explained
• OM1 to OM5: Multimode Grades Compared
• OS1 vs OS2 Singlemode Fiber
• SC/APC vs SC/UPC Connectors