Most important take away

The internet runs on roughly 500-550 fragile undersea fiber optic cables, and two out of every three new cables are now funded in part or whole by just four U.S. Big Tech companies (Google, Meta, Amazon, Microsoft). This concentration, combined with rising geopolitical tensions, accelerating AI-driven data demand, and the exclusion of Chinese players like HMN/Huawei from joint projects, is reshaping the global subsea cable market and creating both vulnerabilities and major capex tailwinds for the firms that build, lay, and maintain this infrastructure.

Summary

Actionable insights and investment-relevant points from the episode:

• AI is a structural tailwind for subsea cable buildout. Subramanian explicitly states that AI training data centers and the cloud workloads they generate are accelerating the need to lay more cables faster. The earlier worry (circa 5-6 years ago) that cable capacity was already sufficient has been overturned by the AI boom. Investors looking for AI infrastructure exposure beyond chips and data centers should consider the cable layer.

• Big Tech is the customer and the financier. Google, Meta, Amazon, and Microsoft fund ~2/3 of new cables. A transatlantic cable costs ~$500M — meaningful for most investors, trivial for hyperscalers. This vertical integration is bullish for the few companies that physically manufacture, lay, and service cables.

• Bottleneck = pricing power. Only a handful of companies in the world can manufacture these fiber optic cables, and only one company globally makes the specialized software cable-laying ships use to control speed and slack. While no specific stock tickers were named in the episode, the named industry players include:

  • SubCom (U.S. cable-laying contractor) — competes with HMN
  • HMN Technologies (spin-off of Huawei) — effectively sanctioned out of U.S./Western consortium work since the first Trump administration, which has frozen multiple Chinese-Western joint cable projects.
  • Companies behind named cables: Meta’s “2Africa” cable (longest cable in the world, largely Meta-funded).

• Geopolitical chokepoints to watch. The Strait of Hormuz, the Red Sea, and the Suez/Egypt corridor are major cable chokepoints. The episode references a headline that Saudi Arabia, Qatar, and the UAE are financing competing data corridors through Syria, Iraq, and East Africa to bypass the Strait of Hormuz — implying contractor and construction opportunity in alternate Middle East/Africa routing.

• Redundancy and repair cycles are recurring revenue. ~100 cables are cut every year (mostly accidental — ship anchors, fishing trawlers; increasingly intentional/grey-zone activity attributed to Russia, China). Repair fleets and grappling-hook style operations remain a steady, capital-intensive business. Governments (UK, Baltic states, Nordics, Taiwan) are deploying naval assets to protect cables — implying defense/maritime spend uplift.

• China bifurcation risk and opportunity. No new direct China-U.S. subsea cable has been laid in years despite booming traffic. Traffic is rerouted through Singapore and the Philippines, suggesting Southeast Asia cable landing stations and data centers benefit structurally. The risk of “internet bifurcation” — parallel U.S. and Chinese cable systems — implies duplicated capex globally.

• Satellites (Starlink) are complementary, not substitutes. Subramanian is unequivocal: nobody in the industry expects satellites to replace cables. Data demand is growing faster than satellite capacity can. This is a mild caution against overweighting satellite internet as a long-term replacement thesis vs. seabed fiber.

• Cyclicality risk. Cable buildout tracks the broader economy — after the dot-com bust, cable laying froze for ~5 years. AI demand is the current driver; a broad recession would slow the cycle, though the AI structural demand should partially insulate it.

• Frontier markets and sovereign data sovereignty. West African and other underserved nations face a dilemma — accept Meta-funded cables on Meta’s terms (without European-style data residency requirements) or remain underserved. Investors in emerging-market telecoms should monitor government data-sovereignty pushback as a potential headwind to hyperscaler-funded routes.

No specific stock tickers or direct buy/sell recommendations were made in the episode. The actionable thesis is: the AI capex cycle extends meaningfully into the subsea cable supply chain (manufacturers, layers, repair fleets, hyperscalers as owners), and geopolitical chokepoint reshuffling (Middle East alternate corridors, China decoupling) is creating new build-out opportunities outside historical routes.

Chapter Summaries

1. Cold open — subsea cables look like shipping lanes. Tracy and Joe marvel at the physical infrastructure of the internet, the 100+ year-old London-NY cable (origin of the FX “cable” nickname), and the geopolitical vulnerability of subsea cables.

2. Why Subramanian wrote the book. He was inspired by Neal Stephenson’s 1996 Wired essay “Mother Earth Mother Board” and later by the 2022 Tonga volcanic eruption that severed Tonga’s only international cable — a real-world test case for life without the internet.

3. How a modern cable works. Cables on the deep ocean floor are as thin as a human hair, made of highly purified glass, carrying data via wave-division multiplexing (multiple laser frequencies). A handful of companies globally manufacture them; one company makes the specialized cable-laying ship software.

4. Why cables instead of satellites. Satellites can’t possibly carry the volume of data humanity produces (Netflix, Zoom, cloud, AI). Cables will remain the backbone.

5. Who pays — evolution of ownership. State telecoms (80s-90s) → privatized consortia (2000s) → Big Tech (last ~7-8 years). Google, Meta, Amazon, Microsoft now fund ~2 of every 3 new cables. Implications for data privacy and access.

6. Redundancy economics. ~500-550 cables worldwide. ~100 cuts/year, mostly accidental. The system is built with deliberate redundancy — even Google buys capacity on competitors’ cables. Land cables also provide backup.

7. Geography and chokepoints. Egypt (echo of ancient Alexandria), Strait of Hormuz, and the Red Sea concentrate cable traffic. India’s cables still land at Mumbai and Chennai — same locations as the 1800s telegraph.

8. Repair operations. Surprisingly old-school — ships drag grappling hooks along the seabed to snag broken cables, then splice them in onboard clean rooms.

9. Vulnerabilities and bifurcation risk. Big Tech concentration shrinks the “open internet”; West African governments face leverage problems negotiating with Meta. HMN/Huawei sanctioned out of Western consortia; no new direct China-US cables in years. Risk of two parallel cable systems globally.

10. Sabotage and government response. Intentional cable cuts (Russia in the Baltic, fears around Taiwan/China) are rising. UK, Nordic, and Baltic navies are on patrol; private and state spend on cable protection is increasing.

11. Alternate routes and new landing points. Gulf states financing corridors through Syria, Iraq, East Africa to bypass Hormuz. Subramanian argues new landing points (away from century-old Mumbai/Chennai/Alexandria locations) would improve resilience.

12. AI’s impact. AI doesn’t change the trajectory — it accelerates it. More cables, faster, driven by hyperscaler AI workloads. Cable industry is cyclical (dot-com bust paused buildouts ~5 years) but currently in a strong upcycle.

13. Private/military cables. No major corporate (e.g., JP Morgan) builds its own subsea cables, though Michael Lewis’s Flash Boys documented overland private cables. The U.S. and Chinese militaries operate their own dedicated subsea cables.

14. Wrap. Subramanian’s favorite cable: the ~41-mile cable between the Isle of Man and Northern Ireland. Tracy and Joe reflect on the parallels to the dot-com era — overbuild followed by undersupply followed by another boom.