Subsea Cables & Internet Infrastructure

 Firmina is a spatial division multiplexing 16 fibre pair cable with initial design capacity of 320 Tbps. It is named after a Brazilian abolitionist, Maria Firmina dos Reis, who was Brazil's first novelist. Google is the owner. Telxius has acquired a fibre pair on the system as part of a complex deal that involves providing landing and back haul in Brazil. Right now Google is selling fibre pair and spectrum capacity to recoup its capex. Cirion Technologies has also purchased a pair. Stonepeak Investments, an  infrastructure investor, purchased Lumen's South American assets which operate today as Cirion. Firmina is substantially complete, but no RFS announcement so far.  Distinguishing Features: 1. It is possible to power the entire cable from either the US or Brazilian landing stations in case the other CLS experiences a black out.  2. Firmina is the third South American hyperscaler subsea cable. Google is the owner of all three.  3. Firmina is the first spatial...

Semiconductor lasers are one of the three building blocks of our modern IT world along with fibre optic cable and the Ethernet protocol. Subsea cables use semiconductor lasers to emit light, fibre optic cable to carry and guide it, and the dominant wrapping for optical wavelengths is an Ethernet format. Finally, the IP/TCP protocols enable different networking technologies to operate together as one network.  Einstein conjectured in 1917 that electrons could be manipulated or stimulated to emit wavelengths at a specific frequency. Later Charles Townes at Bell Labs proposed that light wave intensity could be magnified by passing it through a gas filled cavity. Done in the right way the light would trigger the gas to amplify the light without changing frequency or phase before leaving the chamber. In 1959 at Columbia University he created the immediate predecessor of the semiconductor laser, a maser (microwave amplification by stimulated emission), which used microwaves as both input...

In a class of 23 students the probability of at least two having the same birthday is 51%. In a class of 30 students the likelihood rises to 70%. I was introduced to this counterintuitive result in my first graduate school probability course. There were no more than 15 students in the class. Yet  I and another classmate is that course shared a common birthday of March 5, 1962. I will give you the logic behind this result at the end of the post.  This probability calculation is a clear warning human intuition is not likely to be a good guide to whether an event is natural or man-made or a conspiracy. Most people will assume that the chance of a two or more students with the same birthday is 10% to 15%. In case of telecommunications we have at least 450 subsea fibre optic cables in service today. This means that clusters of cable outages like we see now in the Baltic Sea and earlier this year in the Red Sea and off Abidjan are virtually guaranteed to happen due to pure chance al...

ZBLAN is the acronym for a flouride-based optical fibre that uses a combination of zirconium, lathanum, barium, sodium, and flouride that has a theoretical optical loss of only .001 dB loss per kilometer in the infrared frequencies. Good silicon-based fibre generally lies in the .15 to .20 range. Obviously, if perfected, this would constitute a major fibre optic revolution. The theoretical dB loss of 2000 kilometers of ZBLAN fibre, a French invention, would be equivalent to ten kilometers of standard silica fibre. Optical amplifiers would become obsolete for terrestrial networks and many subsea spans. The challenge is that earth's gravity leads to crystals forming when the glass is drawn and allowed to cool. Unlike standard fibre optic glass which is SiO2, silicon dioxide, ZBLAN is a complex mixture of elements with different molecular weights. Convection during the drawing process results in these elements separating into crystals. The glass also becomes extremely brittle as a res...

Long haul optical networks are approaching the Shannon limit in the work horse C band raising the inevitable question of how to keep increasing bandwidth. Right now the 64QAM modulation scheme achieves spectral efficiency of 6.5 bit/s/hz. But the upper bound on an optical channel according to the Shannon limit is 6.9 bit/s/hz. In other words, we are getting damn close. Obviously, it is a technological triumph. In very few fields of technology has mankind reached one of Nature's own speed limits. But the practical question is where do we go from here? If you look at the chart, the answer is obvious. The L band (1563-1630) is adjacent to the C band and current long haul DWDM equipment can handle these frequencies without any substantive modification (just software upgrades). Furthermore, as the chart shows, the L band optical loss is only slightly higher than the C band. The line marked 'experimental' showed the actual measured optical loss as a function of wavelength. The C ...

Equiano is the second highest capacity subsea cable serving Africa at 144 Tbps. It enjoys the lowest latency Lisbon/Capetown. This subsea network is highly resilient with no wet outages since going live in 2022. The cable was buried 2 meters below the sea floor and routed around the subterranean canyons responsible for many subsea cable outages. It is a Platinum Service. 5x 100Gs LS1/CT1. Spectrum IRUs.  Won't last long ...

The C band is usually defined as consisting of wavelengths in the range from 1530 to 1565 nanometers. Virtually all long haul fibre optic telecommunication transmission uses exclusively this band for data payloads. The graph below shows why. Optical loss or attenuation in the infrared frequencies used by most lasers achieves a global minimum in the C band. The solid bar labelled 'experimental' shows actual measured db loss performance. The other lines represent a theoretical decomposition of overall intensity loss into different factors.  Waveguide imperfections are geometric flaws in the fibre due to the production process. These flaws lead to light scattering. An example are irregular surfaces at the nanoscale. It could be surface of the optical core or the cladding that shields it. Fore example, if the optical core is not perfectly round, and it never is, then some light will either escape or be reflected back down the fibre in the opposite direction. For a detailed analysis...

The O band was the original or first official spectrum range in optical fibre optic communications. It includes the wavelengths ranging from 1260 nm to 1360 nm. The first fibre optic cable trials and deployments in the mid-70s were very short range ranging from a few metres to several kilometres. In 1977 the first phone voice traffic traversed a local fibre optic link in Long Beach, California. The Dorset, UK police deployed a fibre optic link in 1975, but unfortunately, I have been unable to ascertain the specific application. NORAD used fibre optic cables to connect computers at its underground Cheyenne Mountain headquarters in 1975. Note that this 1970 experiments used 850 nanometers as the semiconductor lasers were not capable of longer wavelengths. The O band became the de facto standard in the early 80s when the industry migrated from the early multimode fibres to single mode fibres and gallium arsenic enables lasers to to achieve the longer wavelengths of the O band. By 1988 com...

Right now the industry's service performance is really bad. Much of the responsibility lies with the venture capital owners and senior management. In particular there is this prevalent dotcom era notion that fibre equals value. It does not. What generates value are well managed, high performance assets generating a lot of revenue with lean overhead and happy customers. As a result, we are seeing builds by carriers that should be focusing on achieving excellent customer service and network performance, including provisioning. ***One major European long provider is quoting three month delivery of terrestrial 100G waves. They are aggressively expand their network, but apparently their venture owners forgot that the market does not highly value empty networks. On-net circuit provisioning should take two weeks. In exceptional cases as little as one week. It is hard for me to see why provisioning should take longer.  ***Customer service in general is an abomination. If there were a telec...

Retelit and its minority partner the Italian government just made a $700 million Euro bid for Sparkle's subsea cable business. The bid expires January 27th, 2025. More details here . My first impression is that this is not a rousing endorsement of the subsea cable service industry. Sparkle does a billion Euros in annual revenue according to a publicly available corporate presentation with wet assets including initially full ownership of the Mediterranean Blue cable and some ownership on Blue-Raman, Equiano, Seabras1, and Curie. However, voice comprises a significant portion of its revenue which most investors will penalize in their valuation estimates. In general, the financial markets view voice as less valuable than data traffic because it is considered a low growth, low margin business. Data is sexy. Voice is not.  I have always viewed Sparkle as a stodgy company with a 20th century management style with too well dressed senior managers. Rigid old hierarchical corporate structur...

EU Gives Old Boys Club European Consortium 6 Billion Euros for LEO Every major world power bloc wants its own LEO satellite constellation. The EU is no exception. It has agreed to give a European consortium consisting of SES, Eutelsat, and Hipasat, 6 billion Euros to develop and deploy by 2030 a total of 290 LEO and MEO satellites. Obviously the EU wants the security and privacy of a homegrown communication system to be called IRIS that could back up terrestrial and subsea cables. I understand the impulse. The Europeans face an aggressive Russia and a surveillance happy US government with a President who is more comfortable with dictators than the democratically elected.  At the same time it is clear that there will be glut of low lying satellite capacity in the near future. Starlink has obviously a huge lead over the European project. In addition, marketing juggernaut Amazon began deployment in early 2024 of its 3,236 satellite system. Customer acquisition costs often determine th...

Another dark fibre ring deal for my favorite European WAN provider ... It's been a two terabit sales year with EUNetworks and I look forward to wrapping up a few more DF leases before year's end. With any luck, 3 more to be precise.  Key Dark Fibre Features ***Metros including Dublin, Manchester, London, Amsterdam, Paris, Rotterdam, Utrecht, Hamburg, Frankfurt, Dusseldorf, Stuttgart, Munich, and Berlin. ***Rapid turn up. ***Custom routing available. ***Unique street level routing in 18 European cities ***Dedicated Dark Fibre, contention free and secure, linking 545+ data centres and 2700+ buildings ***Diversity between cities in Europe and into data centres ***Suitable for all transmission technologies, with single mode fibre (G.652, G.655, G.654 and G.657) ***With an SLA time to fix of 8 hours. ***Advanced fibre monitoring available, giving better visibility of service degradation and preempting potential service loss.

The planned Bulikala cable connects Google's modular prefabricated Guam CLS to Fiji. It is a part of a grand plan to dramatically increase Pacific subsea throughput and resiliency via a web of island hopping fibre optic cables. These small islands offer diverse network routing. They also offer power, which is the gating factor for throughput over long distances. Bulikala deployment is well underway with a branch recently landing on Tuvalu island, which has only satellite connectivity. The branch is a joint project of Google and the island's PTT. Most Pacific islands are poor due to limited resources, geographic isolation, and poor digital connectivity. They are also threatened by rising water due to global warming. Even the Hawaii island chain is relatively poor with Honolulu being surprisingly run down.  There is a mighty power play at work here. The US government provides aid to these islands sprinkled across the Pacific Ocean for their on-land digital infrastructure while Go...

The planned Bulikala cable connects Google's modular prefabricated Guam CLS to Fiji. The branch is right below Guam on the map at the bottom of this page. It is part of a large scale Google project costing a billion dollars  to dramatically increase Pacific subsea throughput and resiliency via a web of island hopping fibre optic cables. These small islands such as Fiji, Christmas Island, the Marshalls, and Polynesia offer diverse network routing that is particularly valuable in case of a subsea cable segment goes dark. They also offer power, which is the gating factor for throughput over long distances. All power conductors lead to voltage drawdown which limits bandwidth. Boosting power at intermediate points will allow higher transmission rates and lead to better return on the capital invested. The overall plan is to connect Japan, Guam, Hawaii, many islands such as Fiji and French Polynesia to the US in such a way as to increase both throughput via power stops at small islands a...

Both the Dutch government and municipalities have banned data centre building in Amsterdam as well as power upgrades for existing facilities. The initial data centre ban was lifted several years ago, but stringent energy efficiency standards have deterred subsequent building and created an image of a city hostile to business. On top of that, the municipality is denying power upgrades. For example, the only way to get incremental power right now at AM5, arguably the single best peering point in the city, is to get vacated or novated space. A client of mine asked for space there, but Equinix replied that power was maxed out and authorities had frozen the total power available to AM5 at the current level. Some local observers argue these moratoriums largely reflect infrastructure neglect in areas like transmission, but I suspect at least in part it is an attempt to appease the Dutch public over power rates. The current Dutch government qualifies as very right wing populist with a host of ...

Optical wavelengths are composed of fundamental building blocks known as optical container units. The industry has recognized that current wavelength transmission rates are too chunky or insufficiently granular. There are limited number of sizes with many customers struggling to justify the leap from 100G to the relatively new 400G standard. Or even the 10G to 100G leap. On the supplier side the limited size options of wavelengths can lead to stranded spectrum. For example, you might have enough spectrum for 5G or 150G or 650G. In all these cases the stranded capacity represents lost revenue in a cut throat, ferociously competitive wholesale market. In general network operators struggle to achieve their targeted rates of return on capital because they are selling quasi-homogeneous products in a market where price comparisons are easy and buyers have strong incentive to minimize costs.  The ODUflex standard was introduced in part to create wavelengths below the 2.5G level. The funda...

Dutch governments need to stop being wimps and let the data centre industry grow. There is a moratorium on Amsterdam data centre build permits as well as power upgrades. Government leaders are supposed to lead, not cower in their offices afraid of public opinion. I have clients who cannot colocate in Amsterdam facilities such as AM5 because no spare power capacity is available. The Dutch government should long term add nuclear facilities to handle growing long term power demand and beef up the transmission networks as opposed to public-pleasing moratoriums on data centre building. Or pushing solar panels for a wet and rainy climate at high latitude. Picture below of AM5, one of the best peering points in the Netherlands. Clock on  https://thetechcapital.com/a-regulatory-chokehold-is-suffocating-amsterdams-data-centre-growth/ . 

The telecom industry is gently nudging our fearful government leaders who think Russian incompetence has reached the point that they would use a Chinese cargo ship and its anchor for sabotage. I strongly urge the Russian intelligence to sue Western leaders for slandering their professional competency.  If Russia wanted to damage a subsea cable, they would use timer charges and the ship's crew would be a thousand miles away in Moscow drinking vodka and feasting on Beluga caviar when it exploded. Just like the Ukrainian special forces which probably blew up the Nordstream 2 pipeline to ensure all Russian gas to Europe goes through the Ukraine. They rented a Russian ship which was over 800 kilometers away in international waters when the charges were triggered. See https://blog.telegeography.com/what-to-know-about-submarine-cable-breaks? for the full article. 

1. Equiano 100G; LS1, Lisbon/OADC, Lagos; $18.5K MRC; Two Year Term. 2. WACS 100G; LS1/Abidjan CLS; $34.5K MRC; Three Year Term. 3. Equiano 100G; LS1/Capetown Terraco; $20.5K MRC; Two Year Term. 4. 2Africa 100G; LS1/2Africa CLS, Congo; $38.5K MRC, Three Year Term. 5. Route Protected 100G; Capetown CT1-CT2/JB2; $7,800 MRC; Three Year Term.

BW Group purchased the Hawaiki cable in July, 2021 from the Hawaiki Submarine Cable Limited Partnership. The Partnership's original plan was to build a sister cable known as Hawaiki Nui (Great Hawaiki). BW, a Singapore company, has pursued this idea and finally signed earlier this year a Memorandum Of Understanding with TELIN, the international cable subsidiary of the Indonesian PTT. The MOU is really the partnership or consortium agreement and typically only happens once funding has been secured and all parties are fully onboard. The fact that Hawaiki Nui cable was announced in 2021 and just achieved the critical MOU milestone tells me that it has been very difficult to get this project off the ground. My speculation is that the cable's estimated cost is very high because it requires deep burial in the shallow waters of Indonesia; moreover, a very thorough and expensive marine survey is also necessary. Another challenge is that the cable's route requires Indonesi...

Sources tell me that the Iceland Iris cable was originally a Vodafone project that FARICE effectively highjacked. FARICE is a government-owned telecom incumbent with limited regulatory oversight due to legal loopholes. Indeed, Vodafone even built a data center in Reykjavik to serve as the cable landing station that the data center company Borealis eventually bought after Vodafone gave up on the subsea cable project. FARICE used its control of telecom infrastructure to block Vodafone. Here is the initial Vodafone project announcement: https://lnkd.in/d2aYnWnJ. Emerald was Vodafone's name for a high capacity cable project that would connect the US to Ireland and including an Iceland branch. After the branch was dropped, Aquacomms finished the project under the name of AEC-1. The projects reflects the insider nature of Icelandic society. Because Iceland has only 400,000 people competition in many industries is limited and everyone knows everyone and everything. For example, a handful...

Google is leading a project to create two new subsea cables collectively known as Australia Connect. Its partners include the entrepreneurial Subco , Vocus, and NextDC. Subco is a private operator of subsea cables. It owns the Oman-Australia cable and the SMAP cable that when finished will connect Perth, Adelaide, Melbourne, and Sydney. Vocus is a competitive Australian carrier. I strongly suspect the Australian and US militaries are silent partners in the cable for reasons I outline below. The Bosun cable will link Darwin on Australia's Northern Coast to Christmas Island and then continue onward to Singapore. The Interlink cable connects Sydney to Perth and Perth up to Christmas Island. This project has military written all over it because Christmas Island could be used as a surveillance node for the US-Australian-Japanese military alliance. Equipped with radar the island can survey the the Southern approaches to Malaysia, Indonesia, and Singapore. The fact that the island is gett...

Both cables has been repaired, namely the one connecting Lithuania to Sweden and the C-Lion system as well. The emerging consensus within the European subsea cable community is that the German, Finnish, and Estonian authorities are wrong in claiming sabotage. Anchor dragging is a very crude form of sabotage and in this case the Chinese ship Li Peng crossed 13 cables and only damaged two. Secondly, a key requirement of sabotage is a safe get away. A heavy, slow cargo ship is not a good getaway car. Furthermore, the European press got it wrong when it claimed the ship captain is Russian. It is a Chinese ship with a Chinese captain. I have written an article explaining why the sabotage accusations are so flimsy in my opinion: https://subseacables.blogspot.com/2024/11/american-officials-say-no-sabotage-in.html. The ship dragged its anchor probably to steady itself in bad weather and turned off the transponder because the captain knew what it was doing was irresponsible. I suspect the l...

A subsea cable is like a long piece of string. Except the world's cables total about 1.5 million kilometers. Anything shaped like that is inherently indefensible. The Big Huff and Puff is that if we place sensors on or near the cables, presto, problem solved. Not a chance. All it takes is a technologically sophisticated and patient adversary to send out unmanned drones to locate the cables in deep sea where they lie exposed on the ocean's floor. Record the coordinates or drop a homing beacon. When the war is about to go from cold to hot, release underwater drones with charges to attach themselves to these cables and blow them up.  Sensors are of limited value because no country or military alliance has sufficient vessels to station them close enough to every point of possible attack. What is the point of a warning if it is too late to avert the attack? The invention of radar did not stop air strikes or render air power useless. Satellites can give 30 minutes warning of a nuclea...