Single-Mode vs Multi-Mode in 2026 — The Long Argument, Decided

I worked through this debate three times in three different network organisations. The arguments were always the same. Multi-mode is cheaper. Multi-mode is good enough for short reach. Multi-mode is easier on connectors. Single-mode is overkill for in-rack. Single-mode is more expensive to terminate.

Most of those arguments were true at some point and stopped being true by about 2023. The fight is over. The economics tipped. Most procurement standards I read in 2026 haven't been updated. The result is networks being designed in 2026 to a standard that was current in 2018.

This is the version of the argument that actually reflects 2026 economics.

What changed

Three things shifted between roughly 2020 and 2023.

Transceiver economics. Single-mode SFPs and QSFPs dropped to within a few euros of equivalent multi-mode parts at the volume tiers most networks buy at. Above a hundred-piece order, the multi-mode discount is in the single digits. Below that, multi-mode is often the same price or more, because the production volume of multi-mode parts has dropped while the production volume of single-mode parts has risen.

Fiber cost. OS2 single-mode fiber dropped to roughly parity with OM4 multi-mode at the cable-plant level. OM5 (the wideband multi-mode pushed for short-reach in the late 2010s) is more expensive than OS2 in most markets. The "single-mode is more expensive" argument requires citing fiber prices from a decade ago.

Reach flexibility. A single-mode patch panel can support any current short-reach to long-reach optic. A multi-mode panel locks you to multi-mode optics. As link speeds rise, multi-mode reach falls — 100GBase-SR4 reach was 100m on OM4, 400GBase-SR4 reach is 70m on OM4, 800GBase-SR8 reach is even tighter. Single-mode reach didn't change. The flexibility argument moved decisively to single-mode.

What's left of the multi-mode case

Two places where multi-mode still wins on TCO.

Sub-50m runs at older speeds (sub-100G), at volume. If you're building a 10G or 25G access layer with patch lengths under 50 meters, multi-mode optics are still cheaper per port. The aggregate savings can be meaningful in dense access-layer deployments.

Existing multi-mode plants you don't want to rip out. If your data center has thousands of installed OM4 strands and you're not refreshing the cabling, you'll keep using multi-mode optics as you upgrade switches. That's a different question from "what should you install fresh in 2026".

Everywhere else, the argument has moved.

The procurement-standard problem

Most enterprise procurement standards I read in 2026 default to multi-mode for "anything under 100 meters". The standard was written between 2015 and 2018, when the economics still favoured multi-mode. The standard hasn't been refreshed because nobody in procurement felt a forcing pressure to refresh it.

The result is cable plants being installed today, in 2026, that lock the data center into multi-mode for the next ten years. The first generation of optics deployed on that plant might cost slightly less than the single-mode alternative. The second generation will struggle with reach. The third generation will require a cable refresh that the original procurement decision was supposed to avoid.

This is the kind of cost-amortisation error that doesn't show up until the third refresh, when nobody who made the original decision is still there to be held accountable.

What "default to single-mode" actually buys

Three concrete benefits in a 2026 fresh install.

Forward compatibility with any reach class. Today's plant supports anything from short-reach to 80km optics. The flexibility costs essentially nothing in 2026 terms.

No optic-class lock-in. Procurement choices for transceivers stay open. Vendors can be played against each other on equal footing instead of "vendor A doesn't offer this in multi-mode, so we have to take vendor B's higher price".

Simpler cable plant. One fiber class, one connector spec, one inspection profile. The operational complexity of multi-fiber-class plants is small but real, and it accumulates over time.

The cost premium for single-mode in a fresh install is genuinely zero to single-digit-percent in 2026 prices. The TCO calculation, properly done, almost always favours single-mode.

Where I'd still consider multi-mode in 2026

Two narrow cases.

Brownfield with significant existing OM4. Refresh the optics, keep the cable. The TCO of ripping out OM4 to replace with OS2 isn't justified for most networks. Plan the next-generation cable refresh to single-mode and live with mixed-class operations in the interim.

Specific high-density short-reach applications at older speeds. Dense 10G/25G access layers, large port counts, where the SR-optic-per-port saving multiplies into real money. Even here, the case is shrinking as speeds rise. By 50G access, the multi-mode advantage has largely evaporated.

For most other situations in 2026, the right answer is single-mode and the procurement standard should reflect that.

The recommended action

If you're updating a procurement standard or specifying a new cable plant in the next twelve months, the default should be OS2 single-mode unless you have a specific cost-justified reason to deviate. The exceptions are narrow. The default cases used to be multi-mode. They aren't anymore.

This is one of those engineering decisions where the right answer is widely known among practitioners but isn't reflected in procurement documents. The gap is the procurement update lag — which can easily be five years behind the technical reality. Closing the gap is the unglamorous work of opening the procurement standard and rewriting the cabling default.

The fight is over. The standard should reflect it.