The Squid’s InkSquidGrid’s weekly read of the internet — fresh ink every Monday

Week 39, 2026 · Sep 14 – Sep 20

Covers Monday Sep 14 00:00 to Monday Sep 21 00:00 UTC.

About this week’s count. We started following two gaming companies partway through the week, Nintendo Network and Riot Games, and between them they posted 60 of these reports, none of which could appear in last week's total. Counting only the companies we followed for both full weeks, reports were up 36%, not 47%. One payments company, Xendit, also posted 176 on its own, up from 51 the week before.

How to read this: we measure millions of internet routes around the world. Every speed number below is how many milliseconds (thousandths of a second) slower a route ran than its own normal speed, so higher is worse. Each section says whether things got better or worse than last week.

2,093,450internet routes we measure

722,492,483measurements taken this week up 4% from last week

1,697outages and problems companies reported up 47% from last week

573of them rated major or critical

SG.GS (Singapore network operator)’s rough weekthe week’s story

About the same as last week. SG.GS (Singapore network operator) is one of the large networks that carry internet traffic between countries. Routes through it ran 46.4 milliseconds slower than normal on average this week (last week: 46.7). 851 routes through it were badly slowed (more than 25 milliseconds over normal) at some point. The chart shows how many, hour by hour:

Signed-in customers see the affected routes, the carrier on each side of every changed handoff, and the regional split on the SG.GS (Singapore network operator) assessment page.

The networks that slowed down the most

Large carrier networks, ranked by how much slower than normal their routes ran this week. Big networks get no allowance for their size.

Network	slower than normal	routes badly slowed	last week	compared with last week
SG.GS (Singapore network operator)	46.4 ms	851	46.7 ms	About the same as last week
Bharti Airtel Limited	32.6 ms	1,853	33.2 ms	About the same as last week
FLAG Telecom	31.0 ms	242	30.2 ms	About the same as last week
TransTelecom	26.7 ms	528	26.8 ms	About the same as last week
SEACOM	26.3 ms	480	25.8 ms	About the same as last week
Gateway Communications (Belgium)	26.1 ms	2,378	28.5 ms	Better than last week

“Routes badly slowed” counts routes that ran more than 25 milliseconds over normal at some point this week.

1,697 outages and problems reported by companies

More than last week (up 47%). These are problems online services posted on their own status pages. Last week: 1,156.

SaaS913

Payments & fintech367

AI69

Gaming66

Education & research55

Crypto46

Consumer39

CDN & edge32

Online business software (SaaS) had the most problems again, as it usually does.

What tended to break, in providers’ own words: connectivity ×183 · slowness ×120 · auth & login ×118 · elevated errors ×65

Longest problems to fix: Cloudflare — Incorrect geo location for some Cloudflare WARP users (500 hours, about 21 days); Prezzee — We are experiencing technical issues with Marks and Spencer in UK (499 hours, about 21 days); Huntress — ITDR Ingestion Delay (484 hours, about 20 days).

Ground eventsthe physical world, and whether the internet flinched

Sep 18 · blackout Cuba's power grid collapsed again
On Friday Sep 18 Cuba's electrical grid failed across the whole island, and with the power went most phone and internet service. It is the latest in a run of nationwide blackouts this year, after others in March, July and early August. The government said the cause was under investigation and gave no time for full restoration.
Our fleet’s view: About the same as last week. Routes toward Cuba (16 we measure) ran 4.7 milliseconds slower than normal, versus 4.5 last week, and stayed within a few milliseconds of normal right through the blackout. We could not see it: our measurements reach the big networks that carry traffic toward the island, not the homes and phones that lost power. That is a limit of what we can see, not evidence the lights stayed on.
Fox News on the blackout · Wikipedia: Cuba blackouts, 2024–2026

Sep 14 – 20 · quiet The planet mostly cooperated again
No earthquake reached magnitude 7.0 (per the US Geological Survey) and no storm reached the top two alert levels of GDACS, the UN-backed disaster alert system; India's flood alert ended Sep 16. Two quakes passed magnitude 6, both deep: a 6.5 under Alaska's Aleutian Islands on Sep 17 (14:19 UTC, about 98 km down) and a 6.4 in the highlands of Papua New Guinea on Sep 20 (09:17 UTC, about 104 km down).
Our fleet’s view: Nothing to see in our numbers. We measure no routes into Papua New Guinea, so we cannot say anything about the second quake either way.
USGS event catalogue · GDACS

Ongoing, week 14 of tracking: Iran, better than last week. The government keeps the internet slow and filtered, and the bill that would hand control of the country's links to the outside internet to the Supreme Cyberspace Council has sat with lawmakers since Aug 26 with no vote date. Routes into Iran (55 we measure) ran 4.8 milliseconds slower than normal, down from 9.6 last week.

We cover earthquakes of magnitude 7.0 or more and storms at the top two disaster alert levels. “Our fleet’s view” is what our own measurements saw, not a damage report.

Subsea watchwhere the internet is wet

Since Sep 4 · ongoing Perth's cable fix slips to late October
Last week the first repair was due Sep 16. It did not happen. Vocus, which owns the Australia–Singapore Cable, now expects Indigo West, the other direct Perth–Singapore cable and out since early August, to be back on Oct 24, and its own cable repaired on Nov 4. The order is deliberate: fixing Vocus's cable means switching off a section that is carrying rerouted traffic through Jakarta, so the second path has to be up first.
Our fleet’s view: Worse than last week for Australia. Routes from Australia (10,525 we measure) ran 19.7 milliseconds slower than normal, up from 16.9. Routes from Singapore (14,465) ran 24.3 milliseconds slower, about the same as last week's 23.8. The internet still works, just slower, and will for at least another five weeks.
Vocus repair update, Sep 18

Since Aug 25 · ongoing Vietnam is still down four international cables
No repair has finished yet. Four of the undersea cables that connect Vietnam to the rest of the world are still damaged: SJC2, AAE-1, AAG and IA, taking out about 30% of the country's international capacity. Work on AAE-1 is still expected to start late this month, depending on a repair ship and the weather. SJC2 and AAG still have no repair dates.
Our fleet’s view: About the same as last week. Routes from Vietnam (579 we measure) ran 7.4 milliseconds slower than normal, versus 7.6 last week. Traffic is being rerouted well enough that most people would not notice.
SubTel Forum

Also on the wire this week: ESA Funds Automated Cable Backup Project (2026-09-15).

China’s internet gateways

China’s three big carriers control most traffic in and out of the country. We check how much of each network we can reach from outside China. The percentage is how much we reached; “normal” is its usual level. Lower means more is blocked or unreachable.

China Telecom 50.9% reachable (normal 50.6%)
About normal

China Mobile 82.2% reachable (normal 84.8%)
About normal

China Unicom 89.1% reachable (normal 76.2%)
Better than normal

“About normal” means within 3 percentage points of the usual level.

For the feedpaste-ready

The Squid's Ink — Week 39. The internet, measured, not guessed. The cable story: the fix for Perth's two broken undersea cables to Singapore slipped five weeks. Vocus now expects the first back on Oct 24 and its own on Nov 4. Routes from Australia ran 19.7 milliseconds slower than normal, a little worse than last week. The ground story: Cuba's power grid collapsed again on Sep 18, taking phone and internet service with it. We couldn't see it from outside: the networks carrying traffic toward the island stayed near normal. The backbone story: SG.GS, a Singapore network operator, had the week's roughest ride: routes through it ran 46.4 milliseconds slower than normal, with 851 badly slowed. About the same as last week, and right where Perth's broken cables land. Every number compares a route to its own normal speed, measured by SquidGrid's independent fleet. Full report: https://squidgrid.net/reports

Every figure derives from SquidGrid’s own measurement fleet and its archive of provider-published incidents; provider self-reports are labeled as such. Provider Assessment Reports carry the per-provider detail. The internet was mostly fine this week. We measured anyway.