Big Tech's Private Oceans Are Everyone Else's Problem
The application layer looks diverse. The substrate looks like an oligopoly — and the oligopoly is drilling into the ocean floor while regulators argue about chatbot disclosures.
The most important infrastructure story of the decade isn’t happening in data centers — it’s happening on the ocean floor.
The FT ran a piece this week on hyperscalers — Google, Meta, Microsoft, Amazon — building their own private subsea cable networks. The frame the piece uses is “national security risk for other countries.” That framing is correct, but it’s also the most obvious way to read it. The more interesting read is what it reveals about how power actually consolidates in a networked world, and why no one in a position to stop it will.
The internet was built on a shared-carrier model. Subsea cables were financed by consortiums — telcos, governments, flag carriers chipping in together, everyone sharing capacity, everyone nominally equal at the layer that mattered most. That model made the internet feel like a commons. It wasn’t, but the fiction was useful. What the hyperscalers are doing now is dissolving the fiction. Google controls or co-owns cables that account for a growing share of global internet capacity. Meta’s 2Africa cable, when fully operational, circles much of the African coastline. These aren’t redundancy plays. They’re moats dug in the seabed.
Here’s the thing everyone is tiptoeing around: when a nation’s internet traffic routes through privately owned infrastructure, the company that owns that infrastructure has leverage the nation doesn’t fully understand until it needs to. This isn’t theoretical. It’s the same dynamic that plays out when a country’s cloud workloads all live in AWS us-east-1. The dependency is invisible until it isn’t. A subsea cable is just that dependency made physical, and made harder to exit. You can migrate off AWS in 18 months with enough pain. You cannot build a transoceanic cable in 18 months for any amount of money.
The Lagarde AI financial-stability warning this week is instructive as a contrast. The ECB is worried that AI poses a huge risk to financial stability — Lagarde’s words, not a paraphrase. Fine. But there is no comparable ECB warning about the fact that European financial infrastructure increasingly routes through American-owned subsea cable. The systemic dependency is already baked in at the physical layer. The AI risk is forward-looking. The cable risk is present-tense.
The “national security” framing is going to push this into a policy conversation about regulation, forced sharing, or government investment in competing infrastructure. That will largely fail. Here’s why. The hyperscalers build cables because they have the capital, the engineering bench, and the traffic volume to justify the economics. No government consortium built in 2026 can match those three things simultaneously. The EU can legislate data sovereignty until it’s blue in the face; if the fastest, most reliable path between Frankfurt and Singapore runs through a Google-owned cable, German banks will use it. Stated preferences don’t override revealed preferences when milliseconds matter.
What the affected countries actually have is negotiating leverage at the landing-station level. A cable is worthless until it plugs into shore. Every country where a cable terminates has, in theory, a chokepoint. The question is whether governments will use that leverage coherently — as a long-term bargaining chip — or blow it on short-term access fees and photo ops with hyperscaler executives. History suggests the latter.
The underrated subplot here is what this means for the countries that got leapfrogged. Africa is the clearest case. Meta’s 2Africa and Google’s Equiano gave Sub-Saharan Africa dramatically better connectivity than the continent would have had waiting for multilateral development bank funding. The price was dependency. That trade might have been correct — better to have the cable with strings attached than no cable at all — but the strings are real, and no one negotiated them very hard.
The internet’s original promise was distribution of power through distribution of connectivity. What’s actually happening is concentration of power through consolidation of the physical layer underneath connectivity. The application layer looks diverse — a thousand apps, a thousand services. The substrate looks like an oligopoly. And the oligopoly is drilling into the ocean floor while the regulators argue about chatbot disclosure requirements.
The market is telling you something here. Hyperscalers are deploying capex at a rate that only makes sense if they expect to control — not just use — the next generation of global infrastructure. That’s not a bet on bandwidth. That’s a bet on leverage.