Summary

  • Nepal's internet exchange says it crossed 100 Gbps of traffic exchanged inside the country in April 2026, and that earlier spikes have settled into something consistent. The sentence is the exchange's own, captured as provider-reported page text rather than a measurement anyone outside Nepal can independently re-run.
  • Which leaves the obvious question unanswered: 100 Gbps against what? With no figure for Nepal's international capacity, the milestone is a floor on domestic traffic, not a share of it — and the port count that carried it is not published either.
  • For anyone deciding whether to connect, peer or regulate, the practical content is cost, not pride: a megabit swapped between two Nepali networks rides a port and domestic backhaul instead of international transit, so every gigabit that moves onto the exchange changes what it costs to serve Nepali users to each other.

What the exchange actually said

The announcement is short and unusually cheerful for a traffic graph. "A Landmark of crossing 100G local NPIX traffic is achieved from this month (April 2026)," it reads, according to the provider-reported text captured from the exchange's homepage news feed. "We have seen spikes earlier too but seems its much consistent now." It thanks members "in helping to keep the LOCAL TRAFFIC LOCAL and improving the service quality with BETTER LATENCY."

Read it closely and note what it does not say. There is no claim of new capacity purchased, no facility opened, no cable landed, no router added. The stated cause is member behaviour: networks choosing to hand traffic to each other inside Nepal rather than hauling it abroad and back. That is a commercial decision made by dozens of separate companies, not a construction milestone, and it is the more interesting kind of number because it can reverse.

Provenance matters here, so it should be stated plainly rather than buried. The passage above is a verbatim excerpt reported by the research provider as page text from the exchange's own site, captured on 26 October 2026; the raw page was not independently fetched and re-parsed by BTW. The same sentence also appears in a second capture from the non-www host (rerun of the site capture), which is reassuring, and one earlier capture taken the same day recorded that no such confirmation could be located at the time, which is not. Retrieval through a provider is a sampling instrument, not a seal. The sentence is almost certainly on the page; the honest formulation is that the exchange says it, and BTW holds a dated provider-reported copy of it saying so.

A floor, not a share

One hundred gigabits per second of local traffic is a real quantity and a respectable one for a country of about thirty million people. It also cannot be interpreted on its own, and the exchange's own pages do not supply the missing half of the ratio.

To know whether 100 Gbps is impressive or modest you need Nepal's total internet traffic, or the capacity of its international links, or both. Neither appears in any of the captures examined for this article — not the homepage feed, not the members page, not the registry row. The published number is therefore a floor: at least this much traffic is being exchanged domestically, as of April 2026, by the exchange's own account. It says nothing about the traffic that still leaves the country.

That gap is not a technicality. A domestic exchange's whole economic purpose is substitution — keeping a unit of traffic home that would otherwise cross a border twice. Whether NPIX has captured five per cent of Nepali traffic or forty is the difference between a niche interconnection point and national infrastructure. Nobody in the retained record says which.

The members page names names and counts nothing

The members page is where a reader would expect the arithmetic. What a provider-written summary of it records is a list: World Link Communications, Vianet, Nepal Telecom and NCell among the participants, with Autonomica and Packet Clearing House listed as special members. The summary also notes that the roster reflects a broad cross-section of the Nepali ecosystem — access providers, a national operator, mobile networks, and the kind of root-server and exchange-support organisations that show up where a community is serious about local resolution.

What it does not give is a number. No member total, no port count, no per-member traffic, no capacity per port. This matters for a specific reason: the two named international special members are infrastructural rather than commercial — a root-server operator and a clearing house that supports exchange development — which tells you the exchange has institutional friends without telling you how much traffic any of them push across it.

The difference between a provider-written summary and quoted page text is also worth flagging once. The member names above come from a summary of the page rather than from a verbatim excerpt, so they are reported as "these organisations appear on the members page, per the captured summary," which is weaker than a quotation. The natural reading is that Nepal Telecom, the incumbent, and the largest private access networks share a switching fabric. That reading is plausible and remains unconfirmed at the level of detail a procurement decision would want.

The registry row says almost nothing about switching

There is a second public trace of this entity, and it is easy to over-read. The APNIC public member directory lists Internet Exchange Nepal in country NP at the VERY SMALL membership tier, and links it to its member website (APNIC member directory row).

VERY SMALL is a fee-and-resource classification inside a regional registry's administrative structure. It describes the size of the relationship with APNIC — how much address space and service the member holds, and what it pays — not how many ports it operates, how much traffic crosses its fabric, or how important it is to the country's connectivity. A large exchange can sit in a small registry band; that is a statement about registry accounting, not about interconnection.

There is a useful negative in this row, though. It shows the exchange is a distinct registered holder in the region's number-resource system, and it ties the entity to the same web address that carries the traffic announcement. Whatever else is unclear about NPIX — and plenty is — it is not a name without an institutional footprint.

Where a hundred gigabits has to come from

A single member port at a national exchange is typically a fraction of 100 Gbps; ports at that size would be unusual in any market, let alone one where the exchange's own tier band is the smallest the registry offers. The published figure is an aggregate across members.

That makes it an arithmetic claim about a port estate nobody has published. If a typical member connected at one gigabit, 100 Gbps of aggregate traffic implies a substantial number of connected members and a switching fabric with headroom well beyond what a small handful of ports could carry — and implies peak-hour contention, because an aggregate that equals port capacity on a shared fabric is not a working network. If instead a few members connect at ten or forty gigabits, the same total describes a much more concentrated exchange where a single withdrawal would matter enormously.

Both shapes are consistent with the announcement. They have opposite implications for resilience, for pricing and for what a new member should expect. The exchange, which knows the answer, has not published it.

What a reader should take away

The defensible finding is narrow and worth stating without ornament. As of April 2026 the Nepal Internet Exchange reports a stable 100 Gbps of local traffic, attributes it to member behaviour and improved latency, and names several significant Nepali networks plus two international support organisations among its members. It publishes no member count, no port inventory, no peak-versus-average definition and no national bandwidth denominator, so the milestone cannot be converted into a share of Nepali traffic from public sources.

The consequence for anyone outside Nepal who deals with the country's connectivity is that the exchange is now demonstrably carrying three-digit-gigabit volumes of domestic traffic, which is the point at which domestic peering stops being a community courtesy and starts being where a serious network must be present. The consequence for anyone inside is sharper: every gigabit on that fabric is a gigabit not bought from an international transit provider, and the exchange has just told the market how much of that substitution is happening without telling it the price, the port count, or the ceiling.