Both put dedicated capacity on a fiber path between two points, but they are sold at different layers. Dark fiber is the strand itself: you supply the optics, you choose the modulation, and you own the operational responsibility. A wavelength is a dedicated channel on optics the provider already operates: you get a fixed rate handed off as a port, and the transmission equipment is somebody else’s problem. Choose dark fiber when you want to control capacity growth and have the engineering depth to run optics. Choose a wavelength when you want dedicated, non-contended transport without inheriting an optical network to operate.
These two are frequently shortlisted together and they are not really alternatives in the way that, say, fiber and cable are. They are the same physical path sold at two different layers of the stack. The strand between two buildings can carry dark fiber for one customer and a dozen wavelengths for others, and the light travels the same route either way. What differs is where the demarcation sits and therefore who owns the engineering.
That makes the decision less about performance than about capability and appetite. Ask what your organization wants to be responsible for. Dark fiber buys the maximum ceiling and the maximum obligation. A wavelength buys a defined rate with the optical layer abstracted away. Teams that answer this question honestly rarely regret the choice; teams that buy dark fiber because it sounds like the more serious option often end up paying for capacity they never light.
| Criterion | Dark Fiber | Wavelength Services |
|---|---|---|
| What you receive | An unlit strand or pair between defined endpoints | A dedicated channel at a fixed rate, handed off as a port |
| Who supplies the optics | You do — transponders, amplification, and spares | The provider does, as part of the service |
| Capacity ceiling | Set by the optics you install, not by the contract | Set by the rate you buy; upgrades are a commercial change |
| Adding capacity | Install more optics or additional channels on the same strand | Order another wavelength or a higher rate |
| Contention | None — the strand carries only your light | None — the channel is not shared, though the fiber is |
| Latency profile | Path length plus your own equipment | Path length plus the provider’s transponders — typically a small, stable addition |
| Protocol freedom | Anything your optics can carry, including non-Ethernet | What the service defines, normally Ethernet at a set rate |
| Encryption | Yours to implement, at whichever layer you choose | Yours to implement above the handoff, unless the service offers it |
| Operational responsibility | You run the optical layer and hold the spares | The provider runs the optical layer and monitors it |
| Fault isolation | You determine whether the problem is the strand or your equipment | The provider owns the transport up to the port |
| Cost model | Term or IRU on the strand, plus your capital for optics and spares | Recurring charge scaled to the rate and the route |
| Lead time | Route availability, then your optics procurement and commissioning | Provisioning on an existing platform where the route is already lit |
| Route diversity | A second strand on a genuinely separate route, if one exists | A second wavelength, which is only diverse if it is routed separately |
A single fiber pair can carry many wavelengths at once. Dense wavelength division multiplexing puts dozens of channels on the same strand, each on its own colour of light, each independent of the others. When you buy dark fiber you take the strand and decide how many channels to light and at what rate. When you buy a wavelength you take one of those channels on a strand somebody else has already equipped. The physical path can be identical. The difference is entirely about which side of the transponder the contract stops at.
This is the most consequential difference and it cuts both ways. Dark fiber has no contractual capacity at all — what it carries is a function of the optics you install, so an organization that expects to multiply its requirement can raise the ceiling by re-equipping rather than renegotiating. That is genuinely valuable at scale. It is also a capital commitment, and capacity you have not lit earns nothing. A wavelength inverts both properties: you order a rate, you pay for that rate, and increasing it is a commercial conversation rather than a procurement cycle. If your requirement is large but stable, the flexibility of dark fiber is a premium you may never collect on.
Under-estimating this is the most common way a dark fiber project disappoints. Lighting a strand means selecting and buying transponders that match the route’s loss budget, commissioning them, monitoring optical power, holding spares for the parts that fail, and having someone who can read an OTDR trace at two in the morning. On a short metro span between two of your own buildings that is a modest, well-understood job. Over longer distances, with amplification in the path, it is a specialist discipline. A wavelength service moves all of it to the provider, which is not a lesser choice — it is the correct choice for most organizations that do not employ optical engineers.
Both are dominated by physics. Light travels through glass at roughly two-thirds the speed it does through vacuum, so the route length sets the floor and no product choice moves it. Dark fiber adds only your own equipment above that floor. A wavelength adds the provider’s transponders, which is a small and stable contribution rather than a variable one. For applications where the requirement is consistency rather than an absolute minimum — replication, market data feeds, synchronous storage — both are appropriate and the difference between them is unlikely to be the deciding factor. Route length is worth more attention than product layer.
Neither service encrypts your traffic by default, and it is worth being blunt about that because both are sometimes described as though physical separation were the same as confidentiality. It is not. Dark fiber gives you the freedom to encrypt at whatever layer you choose, including optical-layer encryption on the transponders themselves, because you own them. A wavelength leaves you to encrypt above the handoff. Both remove the shared-infrastructure exposure of internet transport, which is a real and meaningful reduction in attack surface; neither is a substitute for an encryption decision that you make deliberately.
The structures differ more than the totals. Dark fiber is normally a term lease or an IRU on the strand, with your own capital on top for optics, spares, and the engineering time to run them. The recurring cost is comparatively insensitive to how much capacity you actually use, which is exactly why it rewards high and growing requirements. A wavelength is a recurring charge that scales with the rate and the route, with no capital exposure and no spares to hold. Distance, whether the route is already built, and how long you commit for move both models far more than the headline rate does. Any number quoted before an address is surveyed is a guess.
For a single pair of sites needing one large, stable, uncontended pipe, these two land closer together than their reputations suggest, and the deciding factors are usually organizational: whether you have optical engineering capability, whether your growth is genuinely exponential or merely healthy, and whether you would rather spend capital or operating budget. Two competent teams can reach opposite conclusions from the same requirement and both be right. What does not work is buying dark fiber for the ceiling and then never lighting more than one channel — that is a wavelength with extra steps and extra risk.
We sell both, so read this with that in mind. For most organizations moving traffic between two or three sites, a wavelength is the better fit: it delivers the dedicated capacity people actually want from dark fiber without the optical layer they usually have not budgeted to run. We recommend dark fiber where the engineering capability genuinely exists and the growth case is real — carriers, large campuses, data center interconnect, operators who intend to light multiple channels. The failure we see most often is a dark fiber lease bought for its ceiling and used at a fraction of it, which costs more and carries more risk than the wavelength it replaced.
Six questions that settle this choice faster than a feature table. The answer to each one tells you something the specifications do not.
If not, a wavelength is the honest answer. Dark fiber without someone to run the optics becomes an outage you cannot diagnose.
Multiples favour dark fiber, where more capacity is an optics decision. Increments favour a wavelength, where you pay for what you use.
Native Fiber Channel or unusual framing points at dark fiber. Standard Ethernet handoff at a defined rate is what a wavelength does well.
Dark fiber concentrates cost in optics you buy. A wavelength keeps it recurring with nothing on your balance sheet.
A lit route provisions a wavelength quickly. Dark fiber on an unbuilt route means construction, and permitting sets that timeline.
Ask about physical routing rather than product. Two wavelengths on one strand share every failure that strand has, and so do two strands in one conduit.
The channel is not. The strand it travels on may carry other customers’ channels, but they are separated in the optical domain — different colours of light, independent of each other. There is no contention for capacity in the way there is on a shared internet service. What is shared is the physical path, which matters for failure planning rather than for performance: a cut affects everything on that fiber regardless of how many channels are lit.
Frequently, and it is often the sensible sequence. A wavelength proves the route and the requirement without capital exposure. If the traffic grows to the point where lighting your own optics is cheaper than buying more rate, the economics will make that obvious, and by then you will have real measurements rather than a forecast to justify the change.
Neither, in any way that matters. Latency is dominated by the length of the path, and both products travel the same glass. Dark fiber lets you install optics with a marginally lower processing delay, but that difference is dwarfed by route length. If latency is the requirement, the question to ask about is the physical route, not the layer you buy it at.
No, and neither does a wavelength by default. Both are a path. Redundancy means a second path that is physically separate along its whole length, and that is a routing question you should ask explicitly — including whether the two routes share a conduit, a bridge crossing, or a single building entry, because those are where apparent diversity usually turns out not to be diverse.
Describe what the connection has to carry and where. An engineer will tell you which service actually fits — including when the cheaper option is the right one.
Service availability depends on location, network proximity, capacity, and engineering review. Share an address and we will confirm what can be delivered there.