In 2010, campuses were built on OM3. In 2015, on OM4. In 2020, OM5 was arriving. In 2026, for 100G+ speeds OS2 singlemode is most often recommended. The media choice is made once during construction and stays in the ground for 25+ years. This article walks through when each choice wins, and why "cheaper cable" on outdoor campus links usually doesn't mean a cheaper total solution.
Singlemode vs multimode — physics
OS2 singlemode (9/125 µm core/cladding)
**Principle**: a single optical mode propagates through the fibre. The small core (9 µm) eliminates modal dispersion. Usable at 1310 nm (O-band), 1490 nm, 1550 nm (C-band), 1625 nm wavelengths.
- **Bandwidth**: practically unlimited (over thousands of GHz·km, limited by chromatic dispersion + nonlinear effects)
- **Distances**: 10 km (10G), 40 km (10G-LR), 80 km (10G-ER), 120 km+ (DWDM long-haul)
- **Attenuation**: 0.35 dB/km @ 1310 nm, 0.25 dB/km @ 1550 nm (ITU-T G.652.D)
- **Fibre price**: 0.80–1.20 EUR/m (8-fibre OS2 outdoor LSZH cable)
OM3 multimode (50/125 µm core/cladding, laser-optimised)
**Principle**: multiple optical modes propagate through the fibre simultaneously. The larger core (50 µm) → easier connectorisation, but modal dispersion limits distance at high speeds.
- **Bandwidth**: 2,000 MHz·km @ 850 nm
- **Distances**: 300 m (10G), 100 m (40G OM3 SR4), 100 m (100G OM3 SR4)
- **Attenuation**: 3.5 dB/km @ 850 nm, 1.5 dB/km @ 1300 nm
- **Fibre price**: 1.20–1.80 EUR/m (8-fibre OM3)
OM4 multimode (50/125 µm, laser-optimised higher quality)
- **Bandwidth**: 4,700 MHz·km @ 850 nm
- **Distances**: 400 m (10G), 150 m (40G OM4 SR4), 150 m (100G OM4 SR4)
- **Attenuation**: 3.0 dB/km @ 850 nm
- **Fibre price**: 1.50–2.50 EUR/m
OM5 multimode (WBMMF — wide-band MMF)
- **Bandwidth**: 4,700 MHz·km @ 850 nm, extended further for SWDM at 850–950 nm
- **Distances**: 400 m+ at 100G SWDM4 (one fibre pair instead of 4 pairs)
- **Attenuation**: 3.0 dB/km @ 850 nm
- **Fibre price**: 2.20–3.20 EUR/m
- **Niche**: lane reuse in data centres where more fibres can't be added — SWDM across 4 wavelengths saves 75% of fibre count
Transceiver economics — where singlemode long looked expensive
At the same speed, singlemode optics have traditionally been more expensive than multimode, because DFB / EML lasers are more complex than VCSELs (vertical-cavity surface-emitting lasers in multimode).
| Speed | OM3/OM4 SR transceiver | OS2 LR transceiver | OS2 ZR transceiver (40 km+) | |-------|------------------------|--------------------|------------------------------| | 10G | 60–120 EUR (10G-SR) | 120–280 EUR (10G-LR) | 380–650 EUR (10G-ER) | | 25G | 110–180 EUR (25G-SR) | 220–380 EUR (25G-LR) | — | | 40G | 280–450 EUR (40G-SR4) | 480–820 EUR (40G-LR4) | — | | 100G | 380–650 EUR (100G-SR4) | 750–1,200 EUR (100G-LR4) | 1,600–2,800 EUR (100G-ER4) | | 400G | 1,200–1,800 EUR (400G-SR8) | 2,200–3,500 EUR (400G-LR4) | — |
**On short runs** (< 100 m, inside one rack-row) MM SR is markedly cheaper.
**On long runs** (> 300 m, between buildings) the SM transceiver cost spreads over more useful functionality: the same fibre pair (or via WDM, multiple wavelengths) carries you 10 km, avoiding splice-boxes and regenerators.
Real decision 2026 — 5-building industrial campus
Client: production campus, 5 buildings, distances between buildings:
- Building A (HQ + datacentre) ↔ Building B (production 1): 220 m
- Building A ↔ Building C (production 2): 280 m
- Building A ↔ Building D (warehouse + logistics): 340 m
- Building A ↔ Building E (R&D): 380 m
Requirements: - Today: 10G Ethernet between all buildings - Plan 3–5 years: 40G/100G between A and R&D (Building E) - Plan 10 years: possibly 400G to some buildings - 24/7 operation, < 4 hours RTO on link failure
Option 1 — OM4 to all buildings
- 4 routes × average 305 m × 12-fibre OM4 LSZH outdoor cable @ 2.10 EUR/m = **2,562 EUR fibre material**
- Connectors, splice-boxes, manholes, in-wall cabling = 14,000 EUR
- Transceivers: 10G-SR for OM4 @ 90 EUR × 8 (both ends, 4 links) = 720 EUR
- **Total CAPEX: 17,282 EUR**
**Problems in 3 years**: 40G OM4 SR4 reaches 150 m. Links A↔D (340 m) and A↔E (380 m) **can't carry 40G on OM4**. Solution: either OM5 SWDM (expensive and niche), or an additional SM cable.
**Problems in 5 years**: 100G OM4 SR4 also reaches 150 m. No building in the campus except A↔B (220 m) can carry 100G on OM4.
**Cost of "upgrade" in 5 years**: rip-and-replace 3 OM4 routes to OS2 = 35,000–55,000 EUR (manholes, splicing, reconfiguration).
Option 2 — OS2 to all buildings
- 4 routes × average 305 m × 12-fibre OS2 LSZH outdoor cable @ 0.95 EUR/m = **1,159 EUR fibre material**
- Connectors, splice-boxes, manholes, in-wall cabling = 14,000 EUR (same as for OM4)
- Transceivers today: 10G-LR for OS2 @ 180 EUR × 8 = 1,440 EUR
- **Total CAPEX: 16,599 EUR**
**OS2 is actually cheaper than OM4** in this scenario, because the fibre is cheaper and transceivers make a less marked difference at 8 units.
**In 3 years**: 40G LR4 (10 km reach). No problem for the 380m link. Transceiver price: 600 EUR × 2 = 1,200 EUR per link. Three links upgraded to 40G = 3,600 EUR.
**In 5 years**: 100G LR4. Price: ~900 EUR × 2 = 1,800 EUR per link. Three links = 5,400 EUR. **No physical infrastructure change.**
**In 10 years**: 400G LR4. Estimated price 1,500–2,200 EUR per transceiver. **Still the same OS2 route.**
Conclusion
The client chose OS2 for all 4 routes. The argument was simple: today's CAPEX 16,599 EUR vs. 17,282 EUR (difference < 5%), but expected TCO over 15 years:
- OM4 path: 17,282 + 45,000 (rip-and-replace on 40G/100G upgrade) + 10,800 (transceivers) = **73,082 EUR**
- OS2 path: 16,599 + 10,200 (transceiver upgrade to 100G) = **26,799 EUR**
**A 46,000 EUR saving over 15 years from choosing OS2 over OM4** — at a < 5% difference in today's CAPEX.
When multimode still wins
Inside a datacentre, short rack-to-rack runs (< 100 m)
A large datacentre floor with 200+ servers in tens of racks. Spine-leaf topology. Spine switch in one rack, leaf switches in every server rack. Distances < 50 m.
- OM4 SR transceiver @ 100 EUR × 100 pairs = 10,000 EUR
- OS2 LR transceiver @ 300 EUR × 100 pairs = 30,000 EUR
- Difference: 20,000 EUR
**On short runs OM4 is markedly cheaper.** Plus: MM VCSELs have lower power consumption (better at 100+ transceivers), lower heat.
Dedicated 40G/100G run inside one building (< 150 m)
Office floor with multiple server rooms (one main, one auxiliary). 40G or 100G backbone between them. OM4 SR4 reaches 150 m, OS2 LR4 would cost 4× more in transceivers.
Existing OM3/OM4 infrastructure, short-term use
The client buys a building with existing OM3/OM4 backbone. Plans to refurbish in 3 years. Logical solution: use the existing fibre, lay OS2 during the new refurbishment.
When singlemode wins
Inter-building runs > 300 m
The boundary moves down every year. In 2015 it was 500 m. In 2020 it was 400 m. In 2026 it's 300 m (because of 400G/800G requirements).
Future-proofing beyond 10 years
Nobody knows what speed we'll need in 15 years. OS2 fibre from 1995 is used today for 400G ZR. OM3 fibre from 1995 already can't handle 10G over 200 m. **The fibre itself is a 25–30 year investment, transceivers are consumables.**
Outdoor routes (between buildings), even when < 300 m
Pulling fibre through manholes, across a street, through a parking lot — expensive. Rip-and-replace in 5 years when speed requirements change = the same CAPEX again. An OS2 route dug once is a route that lasts.
WDM economics
Singlemode supports CWDM (Coarse WDM, 8 wavelengths per fibre) and DWDM (Dense WDM, 40–80+ wavelengths). When bandwidth grows, you just add a new wavelength, not a new cable.
Multimode WDM (SWDM on OM5) exists, but only 4 wavelengths and only short distances.
Engineering details
Reflections and back-reflection
Singlemode at 1550 nm has lower back-reflection than multimode at 850 nm. That means cleaner signal at higher speeds. At 100G and above back-reflection becomes a problem on MM; OS2 doesn't have it in practice.
Polarisation-mode dispersion (PMD)
On OS2 fibres from the 90s PMD becomes visible at 10G+ over 80 km+ distances. Modern OS2 fibres (G.652.D from 2005+) have PMD under 0.2 ps/√km. **On campuses (< 1 km) PMD is irrelevant**, but for long WAN links it's verified before 100G+ deployment.
Connectors — LC, SC, MPO
- **LC duplex**: standard for 1G–25G, ~3.50 EUR per connector
- **SC**: older, 5 EUR per connector — not for new installations
- **MPO-12 / MPO-24**: for 40G/100G/400G parallel SR4/SR8 — parallel fibres in a single connector. Price 25–60 EUR per connector. Requires careful orientation (polarity A/B/C)
On installation require **every single LC-LC cable tested with OTDR** or MPO tested with MPO loopback. Without testing you get a link with 3–6 dB extra loss that fails at higher speeds.
Splice-boxes and manholes
On outdoor routes, cabling is only one part. Splice-boxes (manholes for splicing) cost 800–2,500 EUR per unit + splice technician work 65–95 EUR/hour + 30 minutes per fusion splice.
**On a 12-fibre cable you have 12 splices in every manhole** = 6 hours of work = 400–600 EUR per manhole junction. This sometimes gets forgotten in the quote.
Our default recommendations
- **Inter-building runs > 200 m**: clearly **OS2** (12 or 24 fibres depending on redundancy). Future-proof 25+ years.
- **Intra-building backbone, datacentre spine, short server-rack runs (< 100 m)**: **OM4** remains economically advantageous for today's 40G/100G.
- **Datacentre with a 400G+ plan in 5 years**: mixed — OM4 for short rack-to-rack, OS2 for spine-aggregation.
- **Office floor backbone (one or two floors)**: **OM4 or OM5** depending on planned speeds. OS2 unnecessarily expensive in transceivers.
Avoid: - OM1 and OM2 (62.5/125 and 50/125 old) — no. Max 1G over 220–300 m. Not used in 2026. - Mixed OM3/OM4 (different sectors with different types) — diagnostics is a nightmare. - Single-fibre links without redundancy on critical routes. Always a redundant routing path through a second cable + a different physical corridor.
---
*We design and install campus fibre infrastructure for industrial campuses, datacentres and office complexes. The first call (45 minutes) walks through planned speeds, building lifespan and physical route availability — usually shows that OS2 is the right choice even on runs that multimode would still handle today.*