Summary

  • India's national exchange network has one published traffic figure: a maximum of 1.53 Tbps handled through 31 December 2024, in a ministry annual report that never defines the interval it was measured over.
  • The telecom regulator counted 51,726 Gbps of international internet bandwidth owned by Indian service providers in the same quarter; dividing one by the other gives about 3 percent, which is an author's arithmetic across two unlike metrics, not a share of Indian traffic.
  • Counts of the same network do not agree across official documents: 42 exchanges in December 2022, 77 nodes and 369 members in December 2024, 79 network operations centre locations on an undated operator page, 79 exchange points in a November 2025 interview.

The figure, and the series that is not beside it

The National Internet Exchange of India sits in the directory as a registry organisation, which is accurate and quietly misleading: the same institution runs the country's domestic peering fabric, and the numbers describing that fabric are thinner than the description suggests.

The Ministry of Electronics and Information Technology's 2024-25 annual report states it in one line. "The maximum volume of Internet traffic being handled by NIXI at present is 1.53 tbps." The same passage records 77 functional Internet Exchange Nodes and 369 members including CDN, through 31 December 2024.

That is the whole measurement. No interval is given for the maximum. No per-node breakdown accompanies it. No earlier or later value appears in the documents examined for this article, which is a statement about the record I could read, not a claim that NIXI has never published anything else. A single figure described as a maximum, without its sampling window, cannot distinguish a busy five minutes from a representative hour, and that distinction is the difference between a dimensioning input and a slogan.

A maximum is not a rate

Peak and average answer different questions. An operator buying capacity cares about the sustained load it must carry and the worst-case burst it must absorb without dropping traffic. A maximum tells you the top of something; it does not tell you the width of the distribution underneath, the time of day it occurred, or whether one member's CDN cache refill produced it.

The report's language is careful in a way the coverage around it usually is not. It says maximum volume handled. It does not say capacity, throughput, average, or utilisation. Read strictly, 1.53 Tbps is the highest transfer rate anyone recorded across the exchange at some point up to the cut-off date. Read loosely by a ministry press summary or a conference slide, it becomes "the network carries 1.53 Tbps", which is a different and unsupported sentence.

The denominator a regulator does count

India does count something adjacent, and counts it quarterly. In its performance-indicator report for the quarter ending December 2024, the Telecom Regulatory Authority of India recorded 51,726 Gbps of international internet bandwidth owned by various service providers, against 49,351 Gbps in the previous quarter. The following quarter's report put the March 2025 figure at 52,136 Gbps, against that same 51,726 Gbps.

Run the deltas and the shape changes: 2,375 Gbps added in the December quarter, roughly 4.8 percent, then 410 Gbps added in the March quarter, roughly 0.8 percent. Both percentages are my arithmetic on TRAI's two disclosures, not figures TRAI publishes.

Mind the label. TRAI counts bandwidth owned by service providers. It does not describe the measure as capacity bought abroad, and it is a capacity stock rather than a traffic flow, so it cannot be compared with an exchange's handled traffic without saying so out loud.

Three percent, and what it refuses to prove

Convert the maximum to the same unit and the arithmetic is short. 1.53 Tbps is 1,530 Gbps. Divided by 51,726 Gbps, it is 2.96 percent, near enough to 3 percent.

That number is not a share of India's internet traffic. It is not a capacity-utilisation rate. It is not a measurement of how much Indian traffic peers domestically, because the numerator counts traffic through one exchange network and the denominator counts bandwidth held by providers who also peer elsewhere, buy transit, and cache content inside the country. Neither publisher reports this ratio; it is mine, and the honest use of it is narrow.

What the comparison does establish is the poverty of the public record. The only two numbers that describe India's domestic and international internet at the same date do not measure the same quantity, so the pair cannot settle whether the exchange carries a large share of the country's traffic or a small one. Anyone asserting either conclusion from these documents is asserting more than the documents say.

Three counts of the same footprint

If the traffic figure is single, the inventory figures are plural and they disagree. A Rajya Sabha answer of 9 December 2022 states that "Forty two (42) Internet Exchanges of National Internet Exchange of India (NIXI) have been installed and operational as on date", with an annexure listing 42 locations and a reference to a Sify data centre in Visakhapatnam carrying 1 Gbps between Chennai and Visakhapatnam.

The 2024-25 ministry report gives 77 nodes and 369 members including CDN. NIXI's own operator pages count differently again: 79 network operations centre locations, and a homepage that advertises 20-plus CDNs and 79-plus NOC locations. Then the chief executive, Devesh Tyagi, told ETTelecom in an interview published on 20 November 2025 that NIXI had established 79 exchange points, that "more than 350 ISPs" were connected, and that the exchange was "accommodating data rates of more than 1.5 Tbps", described in the interview as peak data transported to NIXI.

Each label means something slightly different — installed exchange, functional node, network operations centre location, exchange point — and each member count carries a different qualifier: including CDN, or ISPs only. The figures are separated in time and definition, so they are not an arithmetic contradiction. They are also not reconciled anywhere in the material I examined, which leaves a buyer or a regulator unable to tell growth from reclassification.

What the exchange says it is for

NIXI's published routing policy is the clearest statement of mechanism. Its objectives include keeping domestic traffic within India and promoting domestic hosting. It describes regional multilateral peering and bilateral peering with CDNs and other networks. And it says the router exchanges routing information but does not carry transit traffic.

That last sentence matters more than the traffic figure. A peering exchange that does not sell transit competes with international transit on two things only: latency, because the path stays shorter, and cost, because traffic no longer has to be hauled to a foreign exchange and back. The ministry report adds the second lever explicitly, noting that CDNs may connect at NIXI exchange points without port charges.

So the product is a routing shortcut with a waived access fee, aimed at exactly the traffic — cached content and domestic eyeball-to-content paths — that dominates Indian peak hours. Whether it works at scale is the question the single maximum cannot answer.

What an operator can actually decide

An ISP weighing NIXI as a primary path rather than a supplement needs three things: a series showing how traffic moved across reporting periods, per-node capacity at the locations it would use, and a stated measurement convention so two networks can compare notes. Through the documents examined here, it has none of the three.

What it does have is a policy commitment, a wide location list, and a maximum that has been quoted at roughly the same 1.5 Tbps level from December 2024 to November 2025. Read with restraint, that flatness is ambiguous: it is equally consistent with a plateau in demand, a measurement that has simply not been refreshed, and a figure republished because nobody asked for a better one.

Until NIXI or an independent party publishes a series, the domestic fabric of Indian internet is a policy achievement with a single number attached — and India's operators are still sizing their domestic path against international capacity because nothing better exists.