Subsea Cables & Internet Infrastructure

The National Institute of IT and Communication Technology (NICT) in Japan has been diligently working for many years on dramatically increasing fibre optic bandwidth by using a wider spectrum range than the traditional C and L bands. The long haul fibre optic networks have relied exclusively on the C and L bands because they offer the lowest optical attenuation for silicon-based fibre optics. Light in these frequency bands fades and loses strength relatively slowly as it passes through fibre optic glass. But as the chart below shows, the C and L bands comprise just a fraction of the available spectrum that lasers can use.  In this trial the NICT achieved 450 Tbps on a repeatered London metro fibre pair connecting Telehouse London to the University of London. Because existing optical amplifiers are designed to work only in the C and L bands, NICT has spent years developing amplifiers optimized for the O, E, S, and U bands. These bands required that the fibre optic strands in the am...

RTD: 135 milliseconds. Term: 1 Year. Only one wave available.

The chart shows US business paid AI subscription growth. Musk's AI company is in last place. He shoved xAI into SpaceX in order to siphon off SpaceX cash flow and IPO proceeds to finance his struggling venture. On its own, no one would lend to xAI given its third tier position in the AI service and its $1 billion per month cash burn. Now Anthropic is leasing a huge chunk of xAI data center capacity to estimate its models. Supposedly worth about $1.25 billion per month. I say supposedly because contracts can vary from rock solid to as soft as jello. Indeed, some industry contracts are really just letters of intent, non-binding expressions of customer interest. In this case both sides can cancel the agreement with only 90 days notice. Again, the Musk cheerleaders are prematurely celebrating. What this deal implies is that there's very little demand for xAI's models. So just as X pivoted from social media platform to AI, it is now pivoting to the AI data center b...

The standard description is that 2Africa is a 16 fibre pair cable with 180 Tbps throughput. It comes right off the 2Africa website. That public figure is a bit misleading. Only the Mediterranean and Red Sea segments have 16 pairs. The all-important West Coast segment has 8. Now these fibre pairs should individually be capable of 20 Tbps or slightly more. So West Coast transmission capacity probably does total 180 Tbps. Indeed, the Mediterranean segment could easily exceed 320 Tbps. And the same for the yet unfinished Red Sea segment. Design Capacity By Segment A. Mediterranean Sea: 16. B. Red Sea: 16. Not finished due to hostilities. C. Red/East Africa: 7. D. Red Sea/India: 9. E. West Coast: 8.

The chart below shows the issue. Total 2025 investments totaled $19.5 billion, yet the 1Q2026 investment alone was $16 billion. So unless the first quarter is an outlier, we are looking at $80 billion spent this year with most of it on an AI business that grew 22% in 2025 versus openAI's 230%. Note that xAI's recent deal with Anthropic does not necessarily improve the situation. This contract involves using xAI's data centers to train Anthropic large language models. Contracts of this nature are often closer to letters of intent than hard revenue commitments. And while such a deal might generate a lot more revenue, it will probably also require Starlink to spend more money on power, buy more GPUs, and the like. In other words, this deal might increase the company's negative cash flows. Starlink's current debt at the end of the first quarter of 2026 was $26 billion. The IPO will provide ample cash to wipe the slate clean, but there is a clear risk given that Musk is ...

1. AI division booked a $2.5 billion loss in operating income in the first quarter of 2026. The AI division began as X, but when Musk got AI religion, he pivoted towards developing the Grok large language model and two large AI data centers. To a large extent, he was searching for a way to make X profitable as the site right wing toxicity caused the loss of billions of dollars in advertising revenue. 2. The AI division is causing SpaceX as a whole to suffer massive bottom line losses. It is offsetting the healthy net income of the Internet service division. 3. Transforming X into an AI company with a large language model and accompanying GPU infrastructure was both a me-too investor move and an act of financial desperation. Musk's acquisition of X led to advertising revenue declining from $4.7 billion in 2022 to $1.8 billion in 2025. It was easy for Musk to take X private because the widening losses reduces the company's equity value. 4. Musk used his voting sh...

As the chart shows, Starlink is on track to lose $20 billion this year. These losses largely reflect the struggling xAI division that Musk shotgun merged with Starlink: "SpaceX had a net loss of $4.28 billion on revenue of $4.69 billion for the first quarter, compared with a net loss of $528 million on revenue of about $4 billion a year earlier, the filing shows." Not only does the company lack a clear path to profitability given the xAI division, it is giving Musk permanent control through a special class of voting shares. This is poor corporate governance and should never be done. No man is infallible. It is a recipe for financial disaster as Musk's plans are completely unrealistic and unprofitable like a lunar base, a million orbital data centers, and a Mars colony.

Tiktok is investing $25 billion in a Thai data center. They recently purchased over 4 terabits of Layer 1 capacity into Bangkok causing prices to skyrocket. I have located some spare capacity at the 10G and 100G level.   Appetizer ...  A point: Songkhla Cable Landing Station. Z point: SG1, Singapore.  Two diversely routed 10G waves.  Term: 3 years.  MRC per wave: $8.5K.

In the past most HK and Singapore traffic to the US was routed via Tokyo due to the lack of direct single cable links. So two cables were required to get to the American West Coast. The total 100G costs of two cable solutions were and are still today high with the range from the lower $30Ks to the low 40Ks. Obviously, the Tokyo routing latency penalty is also very high. It is possible to get 100G two cable solutions in the $20K range, but only a few 100Gs are available at that price point.  In recent years the hyperscalers have recognized that current routing raises cost per bit, latency, and makes Tokyo a single point of failure on their Pacific networks. As a result, new American Tech cables connect Singapore directly to San Jose Equinix or Los Angeles Coresite data centers. For example, META is the lead consortium partner on the new 12 fibre pair Bifrost cable that lands at Grover Beach, California, Rosarita, Mexico, and Winema, Oregon. Keppel owns several fibre pai...

A point: SG1. Z point: FR5. Term: 3 Years. Routing: Avoids Marseille and clocks 139 ms RTD. 

 Term: 1 Year. NRC: $10K. Delivery: 8 to 10 weeks. A pt: SG1. Z pt: LA1. Remark: No Bifrost provider currently offers 10G waves.

The EU has given tens of millions of Euros in preliminary development funds to kick start an Arctic cable that would bypass North America and connect Northern Europe to Japan. There have been ongoing efforts for over a decade to execute such a project, the most recent incarnation is Polar Connect. The idea is to create a new, highly diverse and ultra-low latency route linking Europe to Asia that bypasses the politically unstable Middle East and the less-than-friendly and lukewarm American ally. Although there could be some Russian harassment due to concerns about the cable serving as a surveillance tool via sensors attached it, the path looks downright idyllic in terms of the political environment relative to the Middle East. However, a glance at the map shows the immense challenge. One of the unwritten rules of cable deployment is to avoid shallow waters. Most recently built cables head immediately from their landing points to deep sea as quickly as possible. Ships infe...

There is one problem that no arctic cable advocate ever mentions, namely ice scouring. Neither NorduNet nor the Polar Connect team will ever mention this challenge. It is taboo. Here's the problem. Only 10% of an iceberg lies above sea's surface. Salt water is more dense than fresh water and fresh water ice. Hence icebergs are largely undersea and their tooth can extend as deep as 250 meters. In the shallow waters surrounding the Bering Strait the average depth is well under 60 meters. So the tooth or fang of a large iceberg can carve grooves in the sea bed ranging from 50 centimeters to 20 meters. While most ice gouges aare closer to 50 centimeters than 20 meters, deep groves have accumulated on the sea floor. Moreover, the measured depth is not necessarily the original depth of the groove due to sediment burial over time. An iceberg damaged the North Alaskan Quintillian cable in June 2023. It dug a groove 3 meters deep into the sea floor and severed the buried cable. The site...

The map shows the ultimate challenge for any cable crossing the North Pole. The Bering Strait is the exit for any polar project trying to reach Asia. But the average depth is only 30 to 50 meters. As you learned in school, the Bering Strait was a land bridge during the last Ice Age lasting from 37,500 to 12,000 years ago. The lower the temperature, the more water vapor is deposited as snow and ice on land. Hence the oceans recede as temperatures fall.  The construction rule of thumb is to bury a submarine cable when the sea is a thousand meters or less deep. This means a polar cable must be buried for over two thousand kilometers as the map shows. The North Alaskan coastal Quintillion cable's burial depth varies between two and four meters with a maximum of 12 meters. Deep burial is intended to protect the cable from icebergs scraping the sea floor in shallow water. A dual cable design raises per meter cost because the protect path would traverse the more shallow Russian side of t...

The illustration on top shows the Arctic cable's main predator: icebergs. In waters 500 meters or less deep, floating icebergs carve grooves in the seabed floor. These scars are typically one to two meters deep with the record being 15 meters. A fair number of grooves are carved 5 to 8.5 meters into the floor. Known as ice scouring, icebergs have left marks in water as deep as a thousand meters. These comments apply to both the North Pole and the Antarctic. Cables connecting either region to the rest of the world face this challenge. According to a US Government Geological survey, Canada's Beaufort Sea, highlighted in blue, has at least 2,200 ice scouring marks on its sea floor. Quintillion's cable extends into this region. Any cable linking Europe to Asia via the Arctic must go through the Bering Sea. Geological surveys have shown that ice scouring happens every year in the Bering Sea as wind and currents drive ice floes across waters as shallow as 20 meters. Although the ...

Once upon a time we crammed over 20 cables into the Red Sea, gave a lot of money to Egypt's toll collectors, and patted ourselves on the back. Lack of physical diversity? Hey, my consortium buddies see no problems. Let's go have coffee, tea or a beer. Yes, Israel and a good part of the Arab world were in a quiet war. Yes, most governments in the region have no democratic legitimacy and still monitor their people's emails using deep packet inspection. Yes, there are massive economic inequities. Yes, huge religious tensions between Islamic moderates, theocratic Iranians, Shiites, Sunnis, the secular faction, and Messianic Jewish settlers. But hey, everything is cool, so let's keep doing what we've been doing all along. 🙂  Then the world blew up. Yemen has disintegrated into three factions and the Houthis can destroy any ship entering the Red Sea. It took Omantel a half year to convince the Houthis to allow a cable ship to repair Seacom, AAE1, and EIG. ...

Iceland's New Audur Subsea Cable Farice is building a new 16 to 24 fibre pair cable named after an Icelandic matriarch, Audur The Deep Minded, who sailed from Scotland to Iceland in the 9th century during the first wave of migration to the island. The cable lands in Southeast Iceland and in Scotland near Glasgow. The Icelandic backhaul will be fibre, but the British side is most likely spectrum. Farice views Audur as a replacement for the 22 year old FARICE-1. The ship survey will take place in the summer of 2027 with RFS planned for 2030. Audur falls into the monster capacity cable category. Sixteen fibre pairs can easily achieve 320 Tbps with 24 fibre pairs almost reaching a half petabit per second. What is really striking about this is Iceland's population, which although rapidly increasing due to immigration, is just shy of 400K. Undoubtedly, this reflects Farice's bullish assessment of data center demand driven by cheap hydro power and modest cooling needs. Although Ic...

Quintillion was really formed to build a subsea cable from Japan to Europe via the North Pole. It acquired the assets, mostly human capital and preliminary research, of Arctic Fiber in 2016, and as a first step deployed a North Alaskan subsea cable serving coastal communities plus some terrestrial fibre. See the map below on the left.  No expense was spared protecting it from the harsh environment. Quintillion was buried 3.5 to 4.5 meters deep, probably a record, and for landings a bore pipe was deployed. Nonetheless, it was an ultra-high risk project. Icebergs gouge the sea floor as they float. It is called ice scouring. They carve trenches as deep as 15 meters into the sea floor. The fact is that there is no viable protection against them. Moreover, there were no icebreaker cable ships to fix the cable in case of outages. This meant that outages during autumn, winter or spring could not be fixed until late summer. As an example, Quintillion's most recent outage began in January 2...

Amazon plans to dominate the business market for LEO connectivity. Its Ultra phase array antenna in satellite field tests with corporate beta users has simultaneously clocked 1.2 gigabits down and 400 megabits up. Indeed, LEO management has publicly stated that its download performance will be 2x better than Starlink's and enjoy 6x to 8x better uplink performance. The uplink edge is essential to Amazon's strategy. While residential Internet traffic is lopsided with downloads predominating, business and network applications are often symmetric or nearly. Even Starlink's 400 megabit service requires several hours to upload large gigabyte files.   Amazon is targeting mobile operators and IoT aggregation hubs for its 1 gigabit service. Mobile towers are often long distances from the nearest fibre optic network in many countries. Amazon will carry the local traffic back to the mobile operator's POP or data center. Indeed, Amazon's service is ideal for remote data center...

Arianespace will carry 32 Amazon satellites into orbit on April 28th. Amazon has booked a total of 18 launches that will expand the current 250 LEO constellation to 826 birds. For context, Starlink began service in the Northern latitudes when it had 700 operating satellites. So year's end is when competition begins in earnest. However, the Starlink groupies keep trying to move the goal post. They now claim either the company with the most infrastructure wins or they claim that Amazon cannot credibly offer service unless they have several thousands LEOs up and running. But here's reality. There is no real correlation between the amount of infrastructure and business success in telecom. Nor is there is a first mover advantage in telecom. Equinix and Digital Realty were not first movers in the colo industry. But they dominate today. Free came into the French mobile market in 2007, ten years after full liberalization, yet today has over 20% market share and completely...