Subsea Cables & Internet Infrastructure

As mentioned in my first post, the subsea network suffered notorious outages, probably in part due to it hugging the coast on the Denmark/Bude segment. These design flaws were common in consortium cables due to politics. Every member wanted a landing. But it is obvious that threading a cable from Bude back out to sea and through the English Channel to Denmark was inviting damage. This is why engineering expertise is not enough in a multi-owner project to ensure a rational design. Even a dumb layman like me can see that the design invited disaster.

TAT-14 was the last traditional carrier consortium cable across the Atlantic. It was decommissioned in 2020. It was a two cable ring design and suffered notorious double outages. Below is a peak at a network monitoring computer terminal. Notice the vintage Sun computer. 

1. Lisbon/Capetown; $5,500; 1 year term; $5,500 NRC. 2. Lagos OADC/Lisbon; $5,000; 1 year term; $5,500 NRC. 3. Lagos OADC/Capetown CT1; $4,750 MRC; year term; $5,500 NRC. Remark: Equiano soars above the rest of African cables in terms of reliability. Plus the only cross connect charges are at carrier neutral data center POPs. 

This new 6 fiber pair cable directly connects Taiwan and the United States. It is a relatively low capacity system clocking 12.8 Tbps per fibre pair or 76.8 Tbps in total. It is the norm in our era for new ocean spanning cables to have at least 12 pairs and as many as 24. In that sense Orca is a throw back to the age of low fibre pair count coherent optical networks as opposed to the current high count spatial division multiplexing approach pioneered by Google's Dunant cable. Most likely META's traffic forecast is insufficient to justify a bigger count as Taiwan is a niche market of only 23 million people. While its economic significance is hard to exaggerate as the world's leading supplier of semiconductor chips, population still determines the size of traffic flows with a few exceptions like Singapore. It would not be surprising given the cable's transmission capacity if META keeps its all for its own use.  META subsidiaries land and hold the cable landing licenses o...

Pundits have wrecklessly proclaimed that Google believes in orbital data centers. A more accurate description is that the hyperscaler is exploring the idea's potential and obstacles by conducting orbital experiments like the SunCatcher satellite. But Google has no delusions that orbital data centers are a real alternative to terrestrial facilities for the next 10 to 20 years. Google runs lots of experiments because it has lots of cash.  At a recent conference Google's head of data center energy expressed skepticism that orbital data centers make sense any time soon. The costs could be as much as 100x their terrestrial counterpartsin the case where maintaining and upgrading hardware is included. To be clear, 100x was not the best estimate, but reflects the fact that greenfield developments involving new and untested technologies generally result in massive overruns. Moreover, costs depend on whether it is a throw away orbital data center where no components are replaced or repai...

Starlink is introducing congestion pricing in rural America in places like Kansas and Missouri despite a 10,000 satellite constellation. There is only one credible explanation: capacity shortages. Moreover, adding satellites will not fix the problem. Physics and mathematics set non-negotiable limits on per subscriber bandwidth. Enormous Satellite Constellations Yield Only A Trickle of Throughput Per Square Kilometer Starlink's V3 satellite can transmit a terabit per second to Earth. Assume generously that Musk has 42,000 of them four or five years in the future. That's 42 petabits of download capacity. The average person or even the business community will be fooled into thinking this is enormous capacity. Assume it is evenly distributed over the earth's surface of 520 million square kilometers for sake of simplicity. That equals only 82 megabits per square kilometer. Not a flood, just a trickle.  Now you can place more satellites in certain orbits to target more densely po...

Cable: 2Africa. A point: JB1, South Africa. Z point: IColo 1, IColo 2, Telehouse or Seacom CLS. Initial Bandwidth: 10G. Granular Layer 1 FlexE service can be upgraded in 1G increments up to 100G. FlexE Protocol 1. Granular Layer 1 with Ethernet wrapping. 2. No jitter. 3. Layer 1 latency. 4. No Ethernet switching. 5. FlexE: https://www.oiforum.com/technical-work/hot-topics/flex-ethernet-flexe-2/.

Google has been the undisputed leader in hyperscaler projects. I believe this reflects its web search and advertising businesses. These involve relatively high traffic flows due to transactions, transfers of personal and web site data, and updating edge servers. Many of the projects listed below are joint projects with network operators and other hyperscalers. Google's solo projects include Dhivaru, Proa, Topaz, SOL, TPU, Boson, Grace Hopper, Junior, Monet, Dunant, Curie, Firmina, Equiano, Nuvem, Umoja, three Indian cables, Bulikula, Tabua, Honomoana, Halailia, Taihai, SOL, TalayLink, Alisios, Canoa, Olaluz, and others.  Facebook's solo projects include Anjana, Waterworth, Petal, Aurora, and Orca. Facebook is more likely to do joint projects rather than take 100% ownership and control. I think this reflects less international traffic because its social media traffic stays within certain regions like North America, Japan, Europe, Africa, etc.

The famous Léon Thévenin cable ship holds the Atlantic record for repairs with 30 over a 29 year period. I believe a repair in this context includes fixing multiple cables during a single outing. The record refers to repairs within the Atlantic Ocean Agreement service area. It is not a cable laying vessel. Léon Charles Thévenin was a French telegraph engineer famous for his Thévenin's theorem, an extension of Ohm's law to complex electrical circuits.

Microsoft has declined to renew a compute agreement with Oracle, partly on the grounds that the Nvidia chips are too old. This supports Michael Burry's contention that GPUs depreciate faster than the five or six years life span that the AI industry uses in accounting treatments. This in turn suggests that AI data center profits are exaggerated. "Microsoft Corp. is winding down an arrangement to rent AI computing from Oracle Corp., part of a larger reassessment of billions of dollars in cloud partnerships. In recent months, Microsoft has been turning off rented computing capacity from Oracle after declining to renew the contract, according to people familiar with the matter. Over time, the computing Microsoft was renting became less useful because it no longer fit the company’s specifications, said the people, who requested anonymity to discuss a private matter. The Oracle hardware used older-generation Nvidia Corp. chips, the people said." Click for more details: https://...

The SLTE footprint has dramatically shrunk over the last quarter century due to Moore's Law and better optics. As the footprint and power requirements fell, the incentive to migrate the SLTEs to carrier neutral data centers increased because colocation costs fell.

 Microsoft has declined to renew a compute agreement with Oracle, partly on the grounds that the Nvidia chips are too old. This supports Michael Burry's contention that GPUs depreciate faster than the five or six years life span that the AI industry uses in accounting treatments. This in turn suggests that AI data center profits are exaggerated. "Microsoft Corp. is winding down an arrangement to rent AI computing from Oracle Corp., part of a larger reassessment of billions of dollars in cloud partnerships. In recent months, Microsoft has been turning off rented computing capacity from Oracle after declining to renew the contract, according to people familiar with the matter. Over time, the computing Microsoft was renting became less useful because it no longer fit the company’s specifications, said the people, who requested anonymity to discuss a private matter. The Oracle hardware used older-generation Nvidia Corp. chips, the people said." Click here for more details: ht...

Recently a client wanted a port and burst Layer 2 service, similar to what Megaport offers in many parts of the world. The customer wanted to pay a fixed port fee and a variable fee for bursting traffic. In the past this setup used Layer 2 Switched Ethernet. The drawback was the cost of Ethernet switching, network management, and the possibility that the provider would severely overbook the service. Making money on such a service is very hard without overbooking. However, most large capacity providers just sell dumb, Layer 1 pipes. So until recently smaller ISPs had few options. Either Megaport or standard chunky Layer 1 waves. The challenge is that many ISPs has peak traffic well below standard 100G wave services, but above 10G. So rightsizing the network is extremely hard. The good news is that the Optical Internetworking Forum developed the FlexE interface standard, which I view as a Layer 1.5 option. It enjoys all the advantages of layer 1 in terms of hard traffic separation, optic...

1. Massive capital expenditures. Much higher capital intensity per rack than a standard data center.  2. Extremely high operating costs due to the power intensity of AI infrastructure. 3. Low customer switching costs shift negotiating power from supplier to customer.  4. Huge gap between capex and revenue flows: 2 to 5 years. 5. Heavy dependence on debt financing and hence vulnerability to rising risk premiums and interest rates. 6. Underlying technology, large language models, is fundamentally flawed according to many computer scientists. This is corroborated by persistent hallucinations and weak problem solving abilities. 7. Large language models show sharply diminishing returns to scale. 8. Luck is required because there are only a few, very large customers for these projects and loss of a single customer can kill a project. Luck is also important because any of these data center projects requires upgrades to the power transmission grid and also generator capacity. 

What a deal!!! Get yourself a promotion, buy now. 🙂 A point: SG1. Z point: Maruyama CLS. Service: Layer 100G Wave. Interface: 100GE. Term: 1 Year. Delivery: Under 4 weeks.

Some claim that Pangram, an AI program, is good at detecting AI written text. So I tested it on Shakespeare. Apparently the Bard was using AI in the 16th century ... Large language models (LLMs) are not logic engines. They are huge nonlinear statistical models with huge number of parameters, in some cases as many as three trillion. LLMs are black boxes. They spit out answers, but it is not clear how they reach those conclusions. Moreover, complicated statistical models are fragile. This means they provide reasonable answers the majority of the time, but will from time to time produce bizarre responses. In contrast, human intelligence is quite robust. Outliers are extremely rare. 

 https://drive.google.com/file/d/1DgndmDXSnGOa5BiWZsVv0pQ_bEx45bhw/view?usp=sharing

Term: three years. NRC: $8,500. Delivery: 4-6 weeks. Remark: Equiano has the lowest latency between Capetown and Lisbon. A point: LS1, Lisbon Equinix. Z point: CT1, Capetown, SA. Customer responsible for cross connects. 

Without an investment grade rating, Anthropic will be unable to tap the funds of most mutual and bond funds despite a trillion dollars in compute commitments over the next three years. This dramatically increases the likelihood of a liquidity crisis and possible bankruptcy. "The FT says the investment giant Blackstone, the anchor investor in a recent $63 billion deal to finance Anthropic’s chips, told investors in the deal that it expects the AI lab to receive a “high” junk rating after its IPO. Previously, the FT notes, Blackstone had said the opposite: that the Claude maker would receive an investment-grade rating (indicating low risk for lenders) before 2027 rolls around." Click here for more details: https://techcrunch.com/2026/10/30/anthropic-expected-to-receive-high-junk-rating-from-credit-agencies-following-ipo-report-says/. 

Free cash flow equals operating cash flow minus capital expenditures. If capex exceeds economic or real depreciation, then that cash could be returned to the owners via dividend or stock buybacks without sacrificing or reducing the company's ability to generate future profits and cash flow. It is a sustainable payout to the owners. Negative free cash flow increases wealth if marginal capex returns are high enough relative to the risk hurdle. Amazon is still quite profitable today, but investors are concerned future returns are inadequate to justify current negative free cash flow. After all, current wealth is being sacrificed in pursuit of future wealth that may be an illusion. The crux of investor angst is that end user AI revenues are tiny relative to investments. Goldman Sachs forecasts global AI capex to reach $1 trillion in 2026 and $3 trillion in 2030. Gartner, the consulting firm estimates $64 billion in 2026 AI end user revenues. The gap between return as measured by end us...