Subsea Cables & Internet Infrastructure

 1. APG Cable; 100G; Hong Kong/Singapore; $8500 MRC; 1 Year. 2. ADC; 100G; Tokyo/Singapore; $13,850 MRC; 1 Year. 3. ASE; 100G; Tokyo/Singapore; $18,000 MRC; 1 Year. 4. ASE; 100G; HK/Tokyo; $8,500 MRC; 1 Year. 5. ASE; 100G; HK/Singapore; $5500 MRC; 1 Year Remarks: Customers responsible for cross connects. Pricing is the same whether from CLS to CLS or carrier neutral POP to carrier neutral POP.

The emerging computer scientist consensus is that large language models are a dead end and will never achieve general artificial intelligence. It is interesting in this respect that Microsoft just cancelled two gigawatts worth of data center projects. Analysts at TD Cowen, an investment management firm, argue that a data center glut is developing due to too many AI-driven projects. Their view is supported by senior management in the cloud and data center industries. "Earlier this week, Alibaba Group Holding Ltd. Chairman Joe Tsai warned of a potential bubble in data center construction, saying new projects may exceed demand for AI services." Right now the main demand is for AI writing and data summary services via chatbots. It takes a special kind of obtuseness to think these services merit hundreds of billions or even trillions of dollars in data centers, GPUs, and servers.  The herd phenomenon where copy cats pile into a sector due to investor excitement, but without a soli...

The chart shows that foreign tariffs on US goods and services are not much higher than US tariffs on their foreign counterparts. European tariffs on US goods and services are less than 2% higher than their American counterparts. There were no tariffs between the US, Canada, and Mexico prior to Trump's trade war. My forecast is the planned Republican tax cuts guarantee that the US continues to run high trade deficits. In fact, they will grow as higher government deficits raise US interest rates and strengthen the dollar. A higher dollar reduces the price of imported goods and services and raises the price of American exports.  In light of the small tariff differentials no trade agreement is likely to compress the US trade deficit. Now there are always non-tariff barriers that a free trade agreement could dismantle, but some will undoutedly remain because many counties have higher quality and safety  standards than the US. For example, European meat has lower bacter...

1 Year Term; Customer handles cross connects.

The East Asia to America cable connects Taiwan, South Korea, Japan, and the US. It is a standard spatial multiplexing cable with 12 fibre pairs and aggregate design throughput of 192 Tbps. Per fibre pair throughput is relatively limited 16 Tbps due to the cable's long length of 12,500 kilometers together with the lack of any intermediate power sources. E2A is an open cable so the common infrastructure is limited to the wet segment plus the power feed which is typically housed at the cable landing station. Each fibre pair and spectrum owner will install and operate their own submarine termination equipment such as the DWDM kit. Indeed, each owner is free to choose the data centers in which it terminates its capacity. The network design minimizes latency by using a main trunk from Taiwan to Moro, California with branching units to South Korea and Japan. See the map below.  This cable is interesting because to date there have been no direct South Korea to US links. This is the first. ...

 NTT is the lead consortium member on a new Indian subsea cable linking Mumbai and Chennai to Myanmar, Thailand, Malaysia, and Singapore. Like most cables these days Mist is a spatial division multiplexing system. It has 12 fibre pairs with aggregate transmission capacity of 240 Tbps. The DWDM backbone is 400G. Mist is part of a new wave of high capacity, high fibre count subsea cables connecting India to Europe and Southeast Asia that also includes Blue-Raman, IEX, and IAX. Information on their transmission rates is sparse, but their collective throughput is probably at least a half petabit per second and perhaps as high as 750 Tbps. NTT spent about $400 million on the project, which is part of a bigger plan to become a major data center player in India. Indian carriers still largely dominate the local data center business although that is rapidly changing as Equinix makes a concerted push. Undoubtedly, NTT management realized that the main obstacle to thriving Indian data centers...

Looking for new long haul providers from Portugal to Madrid and beyond? I've got them. Lisbon/Madrid/Barcelona/Marseille and Madrid/Paris. These unique routes are in the gas pipeline rights of way and often include new fibre. No Colt nor EXA in the conduit systems.   Just say no to the usual suspects. A 100G wave from LS1 to other Portuguese data centers as well as Madrid is 975 Euros on a 3 year term. You will not be in the same conduit system as Colt or EXA. In fact, most of your Layer 3 upstreams are also absent so this ideal for improving network resiliency via unique physical diversity.

Hollow core transmits light via a central air filled tube in the fibre optic strand versus solid ultra-pure silicon. The benefit is 50% less latency which brings it within 1% of light's maximum speed. High attenuation bedeviled this technology's earliest incarnations as light would bleed and escape through the cladding. A founder of Lumenosity, the preeminent hollow core manufacturer, stopped light bleeding using a reflective layer made of silver. Photonic bandgaps are also used for the same purpose. The fibre offers several performance advantages over traditional solid core fibre: 1. Lower latency since light travels 50% faster via air. 2. Air generates less chromatic dispersion than glass. Chromatic dispersion involves the different frequencies making up a light pulse moving at different speeds. This causes a pulse meant to occupy one time slot to spread and occupy multiple time slots. The data is corrupted. 2. Less nonlinear distortions such as 4 wave mixing, stimulated Bril...

1. Senegal is really hard because several 2African consortium members decided to exclude it from their footprint. But subsea capacity demand is significant because Dakar is a natural telecom hub due to its location along the Northwest African coast. I am working on getting a good subsea 2Africa solution for African ISPs for whom Dakar is part of their plans. 2. If you planning to lease subsea capacity into South Africa or Lagos, best to wait a few months for 2Africa to go live. The combination of 2Africa and Equiano will create a temporary glut. That's the time to strike. 3. I recommend African consortium members on the Equiano and 2Africa cables over their foreign counterparts (Trump will be complaining I am discriminating against white people😄) because they have larger metro footprints. Metro waves are extremely expensive in African cities. It is likely that you can negotiate a subsea deal with an African carrier that eliminates or sharply diminishes any metro charges. I just di...

1. The intra-Asian APG cable has been down for an entire year. Forecast is that it will go live again end of April. 2. Bombing the Houthis is unlikely to get resolve the Red Sea impasse. Air power alone is generally ineffective. Military history says change requires boots on the ground. Indeed, most likely outcome is that the US bombings will delay any resolution. Expect futher delays for 2Africa, Blue-Raman, and SWM6 RFS dates.  3. Got 100G AAE1 wavelength capacity on non-express route for Marseille/Singapore. Available below $25K MRC per month. Big Chinese social media buyer Tiktok looking to take 10x 100Gs on same cable. So you better hurry or end up empty handed. 🙂

Latency: Under 63.3 ms RTD. A point: Singapore CLS. Z point: Maruyama CLS, Japan. Service: 100G Wavelength. MRC: $15.5K. Term: 1 Year. APG has a very short subsea path from the all-important Japanese financial exchanges to their Singapore counterparts. It enjoys lower latency than ASE. But uptime has been poor so you will need redundancy on a physically diverse cable. This is challenging because Pacific Rim cable landing stations often serve a multitude of subsea networks. A typical CLS may be home to as many of 8 cables or more. I can help you in designing your trading and market data networks.

The East Africa capacity shortage is real and chronic. A cable serving the East Coast recently upgraded its network, but sold almost everything before the upgrade was complete. The 100G prices along the coast vary from $35K to $110K per month depending on end points. Not exacly cheap. Peace is not helping ease the crunch as the Chinese carriers are not providing leases into Kenya, but instead offering just IRUs. This eliminates most of the market as the capital required is simply too great for most players except PTTs, OTTs, and mobile carriers. The 2Africa cable is simply insufficient as Facebook kept 4 fibre pairs for itself leaving only 12 pairs for carriers' internal traffic and wholesale sales. The only other significant cables serving the Coastd are Eassy and Seacom. Neither is really high capacity by today's standards. Furthermore, 2Africa's Marseille/Mombasa segment may not be ready until 2026 given Red Sea hostilities. So regional ISPs have little choice but to hau...

1. The eternal East African capacity crunch with 100G pricing ranging from $35K to $110K leads to a new East African cable being announced in quarter two or three of this year. Thanks to Mark Tinka for this brave, but plausible prediction. Peace has proven to be a disappointment for East Africa with capacity owners focusing on selling IRUs into Kenya. 2. In 2026 a consortium decides to build a new West African cable as 2Africa is hopelessly inadequate in terms of footprint and capacity. Facebook is using 4 pairs for its own traffic leaving only 12 pairs for the rest of the market. But 4 pairs is not enough for Facebook's own long term needs. 3. Africa-1, SWM6, and Blue-Raman are all delayed into 2026 due to the threat of Houthi attacks. 4. The collapse of the AI speculative binge begins second half of 2025. AI disappoints because large language models are not capable of logical reasoning nor can they distinguish truth from falsehood. Lack of applications leads to too much money cha...

I am trying to estimate Starlink's total annual depreciation. The reason is quite simple. These LEO satellites have five year life spans. Five years is very short. So depreciation is quite high. Each year a fifth of the fleet is retired. This means depreciation is high relative to revenues and also the steady state capex is quite large. Most telecom infrastructure enjoys a longer life span. The depreciation rate is a bit scary in my opinion.  Starlink had  5,200 operational satellites in May 2024. Construction costs vary from $200K for the earliest models to $800K for the more powerful recent models just being deployed. Let's figure $500K is the average weighted construction cost. Then annual depreciation is roughly $520 million per year.  Right now all you read in the press are good things about Starlink financials. But a private company intending to IPO generally only discloses flattering news. So far we have only been told revenues are in the billions and that the comp...

Naive speculators including a large part of the foreign policy establishment have pushed the idea that Russia is paying Chinese ships to drag their anchors to damage NATO members in the Baltic Sea. Historical records clearly debunk the notion that anchor dragging is a sure sign of intent to sabotage. ***The Chilean flag container ship Aconcagua dragged its anchor while sailing from Philadelphia to New York in 2008. It severed three high capacity Trans-Atlantic cables. The anchor had not been properly secured, which requires doing three separate tasks, and it dropped back into the water. An anchor typically weighs 5 to 10 tons versus 50,000 to 200,000 tons for a cargo ship. This means a ship can drift dragging an anchor even if propulsion has been halted and the anchor dropped. A good example is the infamous Rubymar, an abandoned freight ship, that drifted across the Red Sea for 31 kilometers in the Spring of 2024 causing three key subsea cable outages. In the case of the Aconcagua, the...

The conventional wisdom has been that trawlers are the primary culprit of outages. These ships deploy fishing nets attached to sand skies (my term) that slide along the sea floor. Typically they sink 40 centimeters or even more into the muck catching the very thin fibre optic cables. Sometimes the cables break. In other cases the crew intentionally cuts the cable in its haste to resume harvesting. It is illegal, but the sea has few spectators. BT studied causes of subsea damage in the vicinity of the UK using the AIS (automatic identification system) to track ships involved in incidents. The chart below shows that prior to 2006 fishing was indeed the primary villain. But an evolution has occurred. Global trade expanded over the period while at the same time governments and voluntary organizations like Kingfisher worked hard to protect subsea cables by providing accurate maps of cable routes. Since global trade has only expanded since 2010 it is likely that the trend has continued. Ship...

A few weeks ago Microsoft's CEO warned that much of the AI enthusiasm amounted to hype. Moreover, he is skeptical about general artificial intelligence. He has a point. Microsoft invested $12 billion in OpenAI, but has seen little revenue from integrating its models into the AI Azure services offered on its Cloud platform. In addition, Microsoft has abandoned several site options (lease or land purchase) for new data centers. Just a coincidence? Probably not. Total global AI data center infrastructure spending planned and announced easily exceeds a trillion dollars. Yet Cloud revenues are slowing across providers, not accelerating. AWS has gone from 30% to 40% annual growth down to 10% to 20% in recent years. A similar slowdown has hit Microsoft. Yet at the same time as global revenues are slowing, AI capex has ramped up. There is a great likelihood that the result is a crash with abandoned assets and cancelled projects. Demand is not materializing plus there is a severe shortage o...

The 12 fibre pair spatial division multiplexing cable is a sign of our times. It links together Japan, Taiwan, Guam, Philippines, Singapore, and Indonesia, but excludes Hong Kong and China on security grounds. Consistent with these geopolitical concerns, the cable's main trunk bypasses the South China Sea where China is demanding permit applications (which often exceed a year to get approved), can easily tap cables, and has created artificial military island bases. In a military conflict the fifteen cables traversing the the South China Sea would be at risk. Instead, Apricot takes a long route via Indonesian waters in order to head North along the Philippines' East Coast. The cable's design reflects the potential threat that an aggressive China in conflict with the Philippines and Taiwan poses to Southeast Asian telecommunications.  Although latency is much higher, Apricot offers attractive physical diversity because it avoids the cable congested South China Sea, includes a...

The 12 fibe pair Echo subsea cable aptly illustrates how American Digital Giants are increasingly dominating the Pacific subsea telecommunications market. Carrier consortiums have largely controlled the region's connectivity up to now with the last notable project being the Jupiter cable (RFS 2021) connecting Tokyo to Los Angeles. Google's first solo project was the 16 pair Topaz cable RFS 2024 that directly links Tokyo to Vancouver from where the rest of Canada can be served or the traffic sent South to Seattle. Going forward Google and Facebook own and control all the planned major new cables over the next several years connecting the Asia Rim to the US. Amazon and Microsoft are silient minority investors as well in many of them. Echo is the first cable to seamlessly connect the Singapore telecom hub to the West Coast. Hong Kong and Tokyo were the Pacific Rim's primary telecom hubs up to 2015. But Singapore has seized Hong Kong's telecom crown. This reflects several f...

The 2Africa cable is one of the most ambitious and important subsea projects ever undertaken. It spans a record 45,000 kilometers or 28,000 US miles. As the map shows, it extends from Mumbai to London with many European landings, completely encircles Africa, and provides dense Middle Eastern coverage. 2Africa has a record 46 landings which enables it to serve 33 countries across Europe, Middle East, and  Africa. It also serves India and Pakistan. The cable is unique in having many landings in several countries including 4 in Egypt, 4 in Saudi Arabia, and 4 in South Africa as well as 2 in Kenya, Mozambique, Nigeria, and Spain. A signature theme of the 2Africa project is improving network uptime through physical diversity in the form of multipe, widely separated new landings in key countries. For example, the subsea network brings much needed diversity to Nigeria's telecommunications infrastructure with the first CLS outside Lagos several hundred kilometers to the Southeast. The othe...