Summary

  • The public peering page prints a monthly recurring charge for each port step — 1G at zero, 10G at 200.-, 100G at 600.- or 650.- depending on transceiver type, 400G at 1'000.- — and no currency symbol appears anywhere in the retained text of that table.
  • The statistics page offers four graphs of "the overall traffic exchanged at SwissIX" over 2-day, 1-month, 1-year and 2-year horizons, and prints no peak value, no sampling interval and no per-site or per-port series.
  • The homepage calls SwissIX "the largest Internet Exchange Point (IXP) in Switzerland in terms of port capacity and bandwidth", while the retained record carries no port count, no lit-port inventory and no measurement of any other Swiss exchange.
  • A description of an 80 percent peak-of-paid-capacity aggregation rule circulates in summarised accounts of the peering page but does not appear in the page text captured here; it is reported below as a description, never as a quotation.

The consequence is narrow and practical. An operator can price a port to the unit. It cannot, from this record, work out how much of that port it may use before the exchange's own constraint bites.

What was read, and how it is dated

The four documents behind this report are the SwissIX homepage, the traffic statistics page, the participants page and the peering page. They were captured in a single window that opened at 22:41:19 UTC and closed at 22:42:52 UTC on 18 October 2026. Where this report quotes a sentence, that sentence appeared in the captured text of the page named. Where it describes something a page is said to contain but whose wording was not captured, it says so explicitly and treats the item as a description rather than evidence.

That distinction is not pedantry. The most consequential capacity statement in this entire record — a rule that keeps aggregate traffic below a stated fraction of paid port capacity — reached this desk only in summarised form. A reader who cannot see whether that rule is a sentence on the page, a clause in a contract, or an artefact of a machine-generated précis cannot price a port against it. The report therefore keeps two ledgers: what the pages say, and what they are reported to say. Nothing in the second ledger is used as a premise for a calculation.

The ladder: a complete menu of port steps

The peering page carries a table headed "Port Pricing – Monthly Recurring Charges (MRC)". The captured rows read as follows, reproduced with the operator's own formatting, including the apostrophe in the 400G figure:

BGP Port Speed Standard Transceiver Types Port MRC
1G LX or BiDi 0.-
10G LR or BiDi 200.-
100G LR 600.-
100G LR4 or BiDi 650.-
400G LR4 1'000.-

Three things follow from that table, and a fourth does not.

First, the ladder is monotonic in port size: four speeds, with the fastest step carrying the highest charge. Second, 1G is priced at zero, which is a commercial signal about the kind of member the exchange wants at the bottom of the range rather than a statement about capacity. Third, the 100G step splits by transceiver type — 600.- for LR optics, 650.- for LR4 or BiDi — so the same nominal port speed carries two different monthly charges depending on what is plugged into it. Anyone reconstructing an exchange's revenue from a public price list would have to know the transceiver mix to do it, and this record supplies no such mix.

What does not follow is currency. The charges are printed as digits followed by a full stop and a hyphen. No currency symbol, no currency code and no reference to a currency appears in the captured table. It is entirely reasonable for a Swiss operator to price in Swiss francs, and readers will assume it, but assumption is not evidence and this report does not convert the figures to any currency. All comparisons below are therefore ratios between printed numbers, which survive the absence of a currency because they are internal to the table.

Two ratios are worth stating, because they are the arithmetic a buyer actually performs. The step from 10G to 100G on LR optics moves the nominal port speed by a factor of ten and the printed monthly charge from 200.- to 600.-, a factor of three. The step from 100G LR to 400G LR4 moves the nominal speed by a factor of four and the printed charge from 600.- to 1'000.-, a factor of about 1.67. These are my own divisions of the operator's printed numbers; the operator publishes no such ratios and states no policy about them.

That arithmetic matters because of what it does to behaviour, not because it says anything about the exchange's margin. Where the charge per unit of nominal port speed falls steeply as the port gets larger, a member with rising traffic has a standing commercial reason to buy the next step up rather than a second small port. The published schedule therefore points a growing member toward fewer, larger ports — which is a statement about the price list, not about any member's conduct, and not about how many of each port size exists.

The load surface: four horizons of one aggregate

The statistics page is short and explicit about what it publishes. Its captured sentence reads: "Below you will find a 2-day, 1-month, 1-year and 2-year traffic graph of the overall traffic exchanged at SwissIX."

Four properties of that sentence deserve to be separated, because each one narrows what the page can be used for.

It publishes graphs. Not a table of numbers, not a downloadable series, not a printed peak. The distinction matters when a reader wants to cite a figure: a plotted line can be looked at, but the value at the top of the curve is not stated in text, and a reader who is not looking at the image cannot recover it.

It publishes four horizons. A two-day window and a two-year window answer different questions. The two-day view shows the shape of a working week; the two-year view shows direction. Neither replaces the other, and the presence of both is a genuine disclosure choice by the operator rather than an omission.

It publishes the overall traffic. The word is the operator's. Whatever the four graphs plot, they plot it summed across the exchange — not by site, not by peering LAN, not by member, not by port. There is no per-site series here and no per-member series.

It publishes no definition. Nothing in the captured text states the sampling interval, states whether the plotted value is a maximum sample or a smoothed average, or states what counts as a peak. A reader who wants to know whether the busiest instant of a week is visible in the two-day graph, or has been averaged away, cannot tell from the page.

So the record contains a real load surface. It is simply an aggregate one, and its definitional terms are not stated. An exchange that publishes an aggregate curve has told the market something true and useful about total traffic direction. It has not told any individual member anything about the port that member rents.

The join that is missing

Put the two sections together and the shape of the gap becomes visible. The price ladder describes capacity commercially: it says what a port of a given nominal speed costs per month. The statistics page describes load in aggregate: it shows what the whole exchange carried over four windows. Between them sits the ratio that decides whether a specific purchase is adequate — the load on a specific port, at the hour that matters, as a fraction of that port's capacity.

That ratio is not in the record. This is not a complaint that the operator withholds a secret. It is a statement about the shape of the disclosure: the two published surfaces are indexed to different objects — a port and the exchange — and no published surface maps one onto the other.

The consequence is easiest to see from the buyer's side. Suppose a member is deciding whether to keep a 10G port or move to 100G. The public record gives it the price difference between the two steps and a curve describing the whole exchange. Neither tells it whether its own port is close to its limit at the hour its own traffic peaks. If the aggregate curve is steep, the member may infer that the exchange is busy, which is not the same as its port being full. If the aggregate curve is flat, the member may infer that there is room, which is likewise not the same.

This is where a rule that keeps aggregate traffic below a stated fraction of paid capacity would change the picture, if it exists in the form described. Applied to a single port's nominal speed, a ceiling expressed as a fraction of paid port capacity converts a port speed into a permitted peak on that port — which is precisely the object the graphs do not supply. My own application of the reported fraction to the published port speeds gives, for illustration, a 10G port permitted to run at about eight tenths of its nominal speed, a 100G port at the same fraction of a larger number, and a 400G port at the same fraction again.

Those derived throughputs are my arithmetic on a description, not figures SwissIX publishes, and this report does not present them as the exchange's numbers.

That is why the provenance of the aggregation rule matters so much. If the rule is page text or contract text, then the exchange does publish a peak-referenced capacity constraint in commercial terms, and the missing piece is only the mapping from the constraint to an individual port's observed load. If the rule exists only in summarised descriptions of the page, then the record contains a price menu and an aggregate curve and nothing that constrains how hard a single port may be driven. The two readings support different purchasing decisions, and the captured evidence cannot yet choose between them.

The superlative that cannot be tested

The homepage makes one comparative claim, and it is specific about its terms: "SwissIX is the largest Internet Exchange Point (IXP) in Switzerland in terms of port capacity and bandwidth." Two quantities are named — port capacity and bandwidth — and the claim is that SwissIX leads on them.

Now test it against the rest of the record. Port capacity would be established by a port inventory or a count of installed or lit ports at each speed. The captured pages carry no such count; the ladder gives prices per port, not quantities of ports. Bandwidth would be established by a traffic figure with a definition — the very thing the statistics page declines to print in text. And "largest in Switzerland" is a comparative, which requires a second Swiss exchange measured on the same terms; the captured pages measure no other exchange at all.

The claim is therefore not false. It is untestable from the record that carries it. That is a statement about the disclosure, not about the market: this report offers no view on SwissIX's share of Swiss interconnection, because the retained documents contain nothing that would let anyone compute one.

This is worth naming precisely because the superlative and the measurement live one link apart. A reader who lands on the homepage and follows the Quick Links to the traffic statistics page arrives at four graphs and no figure. The comparative is asserted at the front door; the surface that could begin to substantiate it, if it carried a port count and a defined peak, is one click away and carries neither.

Membership as a presence list

The participants page is one sentence long in the captured text: "Below you find all peering participants. SCION Peering Mesh peers are on a separate page."

Two things follow. First, the membership surface is split into at least two populations, so any count taken from the main participants page is a count of one class, not of the whole membership. Second, and more important for capacity questions, even a complete participant list would establish presence, not volume. A network that appears in a participant list has connected; the list says nothing about how much traffic it exchanges, whether it is a heavy user of the fabric or a light one, or whether the aggregate curve on the statistics page is produced by many members evenly or by a few large ones.

Concentration is therefore not measurable from this record. If a reader needs to know whether a handful of networks dominate SwissIX traffic, the documents that would settle it are not on the public site as captured: they are per-member traffic statistics if the exchange circulates them to members, ratios in audited annual accounts if the exchange's financial statements break out traffic-related income, or operator presentations at a peering forum where a member discloses its own volumes. Naming those document classes is the honest end point; inferring concentration from a participant count would be inventing a measurement.

The same restraint applies to the eight-datacenter count on the homepage. "SwissIX is present in 8 datacenters across Switzerland" is a count of premises. It does not say which premises hold which equipment, how capacity divides among them, or whether the fibre paths between them are independent of one another. A count of sites is not a measure of diversity.

What the record does not establish

The operator describes itself on the homepage in one line: "We are a non-for-profit association." That is the operator's own characterisation of its legal form. No statutes, no registration extract and no fee schedule were captured in this window, so this report relays the self-description as a self-description and asserts nothing further about SwissIX's legal form, ownership or governance.

Four absences should be stated plainly, each scoped to the four documents captured between 22:41:19 and 22:42:52 UTC on 18 October 2026. There is no numeric peak value in any unit of throughput anywhere in those pages' retained text. There is no stated sampling interval and no statement of whether plotted values are maxima or averages. There is no port count, no lit-port inventory, and no per-site or per-port utilisation series. And there is no historical data file: the statistics surface is a set of rendered graphs, and a rendered graph is not a series a member can query.

Absence in a captured excerpt is not proof of absence from the live site. Two of the four graphs' plotted values were not read as data here, so no throughput figure is stated or estimated in this report for SwissIX traffic. A reader who needs the plotted peak must look at the images themselves, on the date they look.

Falsifiers, named in advance

A finding about a record should say what would overturn it. Five developments would.

A printed peak figure with a stated sampling interval would convert the central finding from "no defined measurement" into "this measurement, defined this way", and the article would have to be rewritten around the definition rather than around its absence.

Any per-port or per-member utilisation series — even a single published top-ten of port utilisation — would move headroom from inference to measurement and would make the price ladder answerable against observed load.

Full text of the peering page including the aggregation clause would confirm or refute the 80 percent description and would settle whether a peak-referenced capacity constraint exists in quotable form.

A published port inventory or lit-port count would let the homepage comparative be tested on its own first term, at least within Switzerland.

A measurement of any other Swiss exchange's port capacity or traffic would let the comparative be tested on its second term. Without one, "largest" remains a claim about a set whose other members have not been measured in this record.

The bounded consequence

The useful conclusion is not that SwissIX discloses too little. It discloses more than many exchanges: a priced ladder across four speeds including a 400G step, an explicit aggregate traffic surface with four horizons, a separated SCION Peering Mesh population, and a named presence in eight Swiss datacenters. The conclusion is that the disclosure is shaped like a commercial menu plus a market-level indicator, and that the arithmetic which decides an individual purchase happens in a layer neither surface addresses.

For a member, that means the public record can price a decision but cannot size it. For a reader tracking Swiss interconnection, it means the published curve should be read as a statement about total traffic direction and not as a capacity statement about any port. Both readings are available from the same four pages, and the difference between them is the difference between a price list and a measurement.

Sources

The four documents behind this report were captured between 22:41:19 and 22:42:52 UTC on 18 October 2026: the SwissIX homepage, the traffic statistics page, the participants page and the peering page. The subject's record in this publication's directory is at SwissIX – Swiss Internet Exchange.