Subsea Cables & Internet Infrastructure

GNC is Guam's main carrier neutral data center. It doubles as a cable landing station. Usable colo space is 650 square meters. It has access to 2 megawatts. Guam's main energy source is diesel. A couple cables house their network equipment at the facility, including the Japan-Guam-Australia North and South networks (Google). The main problem for carriers considering Guam is that the PTT controls metro connectivity between the cable landing stations and data centers. The incumbent TeleGuam is a problem because it charges 10 to 20 times the kilometer rate as in developed Western cities like Amsterdam. A 100G metro wave ranges from $5K to $8K per month. That is more than 20% the cost of 100G waves from Singapore to Guam. One can get a dark fibre pair between Amsterdam data centres for 300 to 400 Euros a month. Guam offers the Pacific something it badly needs, a third telecom hub. Not an important place for peering, but ...

As most of you know, Google is building a subsea cable ring between Christmas Island and Australia. I speculated last week that Christmas Island might be where the planned South Africa to Australia Umoja cable would land. It turns out that the new cable, Dhivaru, will connect the new rising Middle East subsea cable hub of Oman to Christmas Island with a stop in the Maldives. The term 'dhivaru' refers to the rope used to control the sail on traditional Maldivian ships. So the Google plan is quite clear. Google is creating an Indian Ocean subsea cable ring connecting Africa to Australia via Umoja and the Middle East to Australia via Dhivaru and the Bosun cable linking Christmas Island to Darwin. I think that all these Google cables will be 16 or 24 fibre pairs pairs. Certainly not less than 16, but not exceeding 24, as traffic cannot justify it. At first glance there are losers and scorned parties. None of these cables land in India, which might reflect India's subsea cable r...

Tabua is part of Google's grand Pacific Initiative, a project to build a mesh-like web of subsea cables connecting Japan, the US, Australia, and many Pacific islands. The islands include Guam, Fiji, Hawaii, and French Polynesia. These islands play a crucial role: they provide power to keep throughput at higher levels than otherwise possible. Another key role for the islands is as telecom switching hubs with each cable landing station serving several high capacity cables.  Tabua is a standard 16 fibre pair spatial division multiplexing cable with two branches landing on the Australian and American sides. Design throughput is 17 Tbps per fibre pair. This dual branch approach has become popular because if the Queensland branch is damaged, traffic can be switched to the New Wales CLS with fibre linking the two cable landing stations. Similarly, on the US side, it lands in Hawaii and also Los Angeles. If the Hawaii/LA segment fails, then the traffic can be routed via other cables landin...

LS1/PAIX Accra; 100G; Three Years; $25K MRC. London/Lagos; 10G; Three Years; $10K MRC. 2Africa cross connect charges are limited to $150 per month and the cable's reliability will be much better than SAT-3, MainOne, WACS, and ACE. It is buried deeper with better designed back haul and avoids the Congo and the Le Trou Sans Fond canyons. Note the Red Sea segment from Egypt to Oman is incomplete. There is no schedule for its completion.

Zuckerberg should pull his head out of his ass. Yann LeCun's META departure underscores just how much Mark Zuckerberg has pivoted away from seeking AI innovation to chasing the latest fad, namely large language models like ChatGPT. Zuckerberg has a history of bad decisions. The stillborn Metaverse is just one example. Yann LeCun is one of the AI Godfathers. He used convolutional neural nets to greatly improve optical character recognition and computer vision. In contrast, large language models are an obvious dead end due to their creative writing tendencies and amoeba-like reasoning abilities (no offense, amoebas, you actaully excel ChatGPT). Fundamental breakthroughs are not achieved by scaling up digital parrots. Yann has repeatedly noted that the best AI models, the large language models, appear to be just regurgitation machines without any ability to reach new conclusions. They cannot identify or correct their own mistakes or realize that their approach is failing and adopt a n...

- 2Africa is much more expensive than Equiano. The 2Africa 100G pricing is $25K and above excluding tails for Lisbon to Lagos. In contrast, Equiano 100G pricing is below $20K now. Similarly, Equiano 10G pricing gravitates around $5K versus $10K on the same route for 2Africa.  The reason for this disparity is that the 144 Tbps Equiano cable primarily serves South Africa, Portugal, and Nigeria. In contrast, the 180 Tbps 2Africa network serves over 30 countries and Facebook kept 4 of the 16 pairs for itself. Note that the 2Africa map does not include the Pearls extension of 2Africa to the Persian Gulf, Pakistan, and Mumbai.  Another sign of the impending capacity crunch is the unwillingness of 2Africa consortium members to sell IRUs. An IRU is a long term capacity sale ranging typically dffrom 10 years to life of system. Carriers will not sell IRUs if they expect future capacity shortages or think they are likely. Many of these carriers have transit backbones that they must keep ...

Google's Bosun cable will connect Christmas Island, located in the East Indian Ocean, to Darwin, Australia, site of a large military base with rotating contingents of Japanese and American soldiers. The project was announced near the end of 2024. At the time it struck me as a bit strange. Google's new cables across the Pacific will do a lot of island hopping. This allows the power to be boosted, the islands can serve as traffic switching centers if they are hosting multiple cables, and complete OEO regeneration can be done. Voltage drops as electricity flows through the copper or aluminum current conductor. So the advantage of powering a cable at intermediate points is clear. It enables higher end-to-end transmission throughput. A key aspect of Google's Pacific projects is better resiliency. The easiest way to do that is a put a small prefabricated modular CLS on an island and land multiple cables there. There Layer 3 switching can divert traffic in case a ca...