Summary
- A megabit exchanged between two Nepali networks travels over a port and domestic backhaul instead of international transit capacity. Seeing this exchange grow means the border price of connecting one's own users to each other is falling—that is the substantive claim of the announcement.
- The exchange point cites 100 Gbps in April 2026 for how much traffic stays local, but publishes neither port count nor port capacity, neither peak nor average definition. The price of this substitution remains outside the field of view.
- Without the country's total international capacity there is no share, only a floor: 100 Gbps is the minimum traffic no longer using the international transit path, not a percentage of Nepal's total traffic.
What the exchange's page actually says
The sentence comes from the exchange itself. In the version captured as page text of the homepage, it says: 'A Landmark of crossing 100G local NPIX traffic is achieved from this month (April 2026). We have seen spikes earlier too but seems its much consistent now. Thanks to our members in helping to keep the LOCAL TRAFFIC LOCAL and improving the service quality with BETTER LATENCY.' (NPIX homepage)
That is a dense announcement, and it is modest in content. It names a month, an order of magnitude, and explains the rise by member behavior, not new construction, not a purchased capacity increase, not a new site. No sentence claims a new facility went into operation. Whoever reads an investment announcement here reads something other than what is written.
The origin of this quote must be stated clearly: it is page text reported by the research provider, not raw HTML fetched directly by BTW. An earlier capture of the same page could not locate the announcement at that time (second capture). A page changing is normal; it remains a capture discrepancy that may interest the reader, because it marks the difference between a permanently published and a temporarily visible announcement.
What 100 Gbps means in a country like Nepal
An exchange point is not a network. It is a place where networks meet, and it consists of switches where ports are sold or assigned. Traffic passing over such a switch does not leave the country. It does not pay for submarine cable capacity, touch an international transit contract, or wait for capacity expansion at an upstream provider.
For an operator, this shifts the cost curve. An additional megabit between two Nepali customers costs a port and domestic backhaul to that port. An additional megabit over the international path costs port capacity on both sides, transit and reservation of international capacity. This is not a price figure, and this analysis contains no prices—it is the structure of costs, not their level. But it explains why 100 Gbps of local traffic is more than a marketing number: it is a volume that has already passed the more expensive path.
From the same structure follows a second observation, but it remains an open calculation, not speculation. The published figure is a total attributed to the exchange's members; the captured pages state neither the number of member ports nor their capacity nor the topology over which traffic runs. A 100 Gbps total could in principle be carried by a single high-capacity port or by several smaller ports. The registry's VERY SMALL tier is an administrative fee-and-resource classification for APNIC membership, not evidence about typical port sizes or market norms.
Without port count, capacities and topology, a statement about how many ports or members the total spans would be an inference, not a verified fact. What the published total does establish is a floor: in April 2026, according to the exchange, at least 100 Gbps of local traffic was exchanged.
Members without a confirmed count
The members page names World Link Communications, Vianet, Nepal Telecom and NCell as participants and lists Autonomica and Packet Clearing House as special members (NPIX members page). These names are the only clue to who generates the traffic that makes up the 100 Gbps—and they are a summary by the research provider, not quoted page text.
The preserved materials do not settle a member count. An earlier summary prepared by the research provider reports more than 40 distinct entities; a later summary gives the names but no total and quotes no page text. Neither is directly inspected page text; the honest statement is therefore not that the site publishes no total, but that no total can be confirmed from the preserved material. No port count, traffic per member or capacity per port appears either.
This matters for assessment. A list containing Nepal Telecom and NCell says something about the exchange's reach in the country—the state operator and a private mobile network operator are not the kind of participants that treat an exchange as a sideshow. But it remains a list without weighting. Whether the four named are the whole picture, whether it is a selection, whether it reflects today or three years ago: none of that is in the captured page.
The registry line is a fee class, not a scale
The APNIC member database lists Internet Exchange Nepal with country code NP at the VERY SMALL member tier and links to the member website (APNIC member directory). This line is regularly misread. VERY SMALL is an administrative classification derived from the regional registry's fee and resource structure. It says something about the relationship between the exchange and the registry, nothing about interconnection weight, nothing about traffic, nothing about the number of ports.
It is noteworthy that an exchange with a three-digit gigabit volume sits in a small fee class. But the conclusion from that is not a market share, but an observation about capital intensity: an exchange does not grow proportionally to its fee class, because costs lie mainly in switches, sites and operations, not in registered address volume.
The missing denominator
Nothing in the captured and dated pages names Nepal's international internet capacity. Without that magnitude no share can be calculated: whoever does not know how much traffic the country carries abroad in total cannot say what part the exchange has already intercepted. 100 Gbps is therefore a floor of traffic kept local, not a share.
The same applies to the metric. The announcement says the value appears 'much consistent'—but does not define whether 100 Gbps is a peak, a daily average or a monthly average. For an operator sizing capacity, this is not an academic difference: peak load requires reserves, average does not. Likewise missing from the captured pages are a site or PoP list, port capacity data and any measurement of IPv6 deployment. These are gaps in publication, not proof of absence of capability—an exchange can operate ports and sites without inventorying them on a homepage.
What can be said for Nepal
The cautious reading is the defensible one. In April 2026 a national exchange point reported a volume of 100 Gbps of locally exchanged traffic and explicitly attributed that level to member behavior. That is an operating result, not a construction project. It shows that a relevant part of Nepali traffic between Nepali networks already takes the path through the exchange point rather than the international transit chain.
For an operator in the country this means: the decision whether to connect a new customer changes the international purchase volume less than before. For an investor it means that the growth story of such an exchange point is not told in gigabits per year, but in ports, sites and whether the large members are ready to concentrate more traffic there. For a regulator it means that the measurement that would substantiate such a development does not exist: as far as the captured pages go, there is no published survey comparing local exchange and international capacity of the same country in the same time series.
The exchange point itself is listed in BTW's directory as an APNIC membership and numbering-resource trace, not as proof of service or operating capability (directory entry). Whoever places the two entries—the registry line and the announcement on its own page—side by side holds two very different documents: a fee classification and an operating announcement. The second is the substantively relevant one, and it is the one carrying the least detail.
An earlier BTW article described the exchange point's role as host of the SANOG week in Kathmandu. That is a different story: events, workshops, handovers. The interesting question after that week is not who hosted, but what the host publishes as infrastructure. The answer in October 2026 is a date, an order of magnitude, a sentence about members—and a hole where ports, sites and denominator should be.
Two minor provenance points belong to this finding. The news archive of the captured homepage goes back to August 2010; that is the beginning of the archive, not a founding date, and no preserved document names a founding year. The same provider-captured page also lists the exchange as host or co-host of regional events, including SANOG 44 with NPNOG 12 and APrIGF 2026. Event hosting is organizational activity and says nothing about switched traffic, but it is part of what the exchange publishes about itself.
Member Briefing
Read the full trend briefing
Members read the full briefing with the source notes behind every finding.
Strategic Circle
A daily briefing in your language, built from changes we have verified.
Join for USD 20/monthCancel anytime.
Members also get the full research library, a daily briefing on the networks they follow and the private forum.






