The invisible threads of power: how undersea cables shape the AI data race
When critical infrastructure becomes corporate property.
I feel like I need to start a series called "things that are keeping me up at night" because the last few articles I wrote are just that.
Undersea cables also fit that category. As you read this, your connection to these words likely traveled thousands of kilometres through cables no thicker than a garden hose, lying on the ocean floor in complete darkness. Believe it or not, these undersea cables (over 400 of them) carry 99% of all intercontinental internet traffic. They’re the invisible infrastructure that makes our hyperconnected world possible, yet most of us never think about them. I didn't either, to be honest with you, until I took a law class. I took that class at Yale Law School and learned a little bit more about them.
These cables account for 95% of the world's international voice and data traffic, handling everything from military communications and government operations to air traffic control and port management. Financial markets utilize undersea cables to transfer trillions of dollars every single day; in 2004 alone, approximately $7.4 trillion a day was traded on cables transmitting data between 208 countries. As you can see, this is why undersea cables really do matter.
But here's the thing about invisible infrastructure: it only stays invisible until someone decides to make it very visible.
When cables are cut, chaos follows
The vulnerability of undersea cables isn't theoretical. It's playing out in real time. In November 2024, two major data cables in the Baltic Sea were severed within 24 hours of each other. One connected Finland to Germany, and the other linked Sweden to Lithuania. When this happened, German Defense Minister Boris Pistorius said, "nobody believes these cables were accidentally cut. We have to assume without knowing for sure yet that this was an act of sabotage."
Read that again. He said sabotage! And he's not totally wrong. The prime suspect of that cable cut was a Chinese cargo ship called the Yi Peng 3, which had left a Russian port and whose tracking data showed that it had slowed to a stop and drifted directly over both cables; in one case for more than an hour. The ship was eventually detained by Danish naval forces and inspected by representatives from four EU countries, along with Swedish police and Chinese officials.
I would love to say that this was the first time this happened, but it wasn't. In October 2023, another Chinese ship called the NewNew Polar Bear dragged its anchor several hundred nautical miles across the Baltic Sea floor, destroying undersea cables and a gas pipeline connecting Finland and Estonia. China later admitted that the vessel was responsible but claimed it was an accident. But the pattern I'm talking about goes beyond Europe.
In early 2025, Taiwan accused a Chinese-owned vessel of cutting an undersea communications cable off the Taiwanese coast. A Taiwanese official put it best: "You need to accidentally drop your anchor on the cable, and then you need to accidentally turn on your engine with the anchor down, and even if you realize your anchor is down, you need to keep the engine moving until you cut the cable."
Which speaks to one of the biggest problems I see with undersea cables: the challenge of attribution. It's incredibly hard to tell if someone's cutting cables intentionally or accidentally, especially when ships can move quickly and may not be properly marked.
There's an average of about 200 cable faults a year according to the International Cable Protection Committee. Most damage is caused by ship anchors or fishing equipment, which makes it the perfect cover for deliberate sabotage. It's like hiding a needle in a haystack of legitimate accidents.
The legal vacuum under the sea
When it comes to protecting undersea cables, international law is stuck in the 19th century (no offense, but I think this is one of the many reasons why people struggle to understand or even believe in international law anymore). The foundational treaty is still the International Convention for the Protection of Submarine Telegraph Cables, concluded in Paris in 1884, clearly well before what our world looks like today.
The 1982 UN Convention on the Law of the Sea incorporated some cable protections and strengthened provisions by making them applicable to conduct "calculated or likely to result in" cable damage. But there are massive gaps. When events damage cables in international waters, "there is currently no effective regime to hold the perpetrator of damage responsible."
I did a bit more reading and found Article 113 of UNCLOS, which provides criminal sanctions for those who willfully or negligently injure undersea cables. But nations can only bring those sanctions if they've passed national legislation implementing the article, and most, including the US, have not. Even more problematic, in my opinion, is that UNCLOS "does not criminalize international theft of undersea cables" in international waters, "leaving nations without a legal basis to prevent and penalize such activity."
So aside from the laws being unclear but also having clear gaps in them, enforcement is equally a challenge. In the Baltic incidents I mentioned, coastal authorities had to seek China's consent to inspect the suspected ships under international law. When cables are cut in the high seas, the most distant zone from any nation's coast, no single country has clear authority to act unilaterally, which brings up even more questions on how you can hold people who are responsible accountable.
The new cable barons
While governments struggle with 140-year-old laws, a handful of tech companies are quietly building their own internet. In the last five years, cables partly owned by Google, Facebook, Microsoft, and Amazon have risen eight-fold. Google is a part or sole owner of around 33 subsea cables, Meta owns more than a dozen, Microsoft owns five, and Amazon owns four. These seem like really small numbers, except Google now owns 1.4% of submarine cables worldwide by length, and 8.5% when including shared ownership. In the past 10 years alone, the amount of international cable capacity used by these four tech giants has surged from 10% to 71%.
So where is all this use going? I was able to find this: traffic between data centers accounted for 77% of transatlantic traffic and 60% of trans-Pacific traffic, and it's growing faster than per-person data consumption. These companies aren't just moving our data; they're increasingly moving their own AI models, training data, and computational workloads between their global data centers.
Meta's latest announcement, Project Waterworth, is a $10 billion cable spanning over 50,000 kilometers across five continents that Meta will solely own and operate. The company explicitly frames it in terms of AI: "As AI continues to transform industries and societies around the world, it's clear that capacity, resilience, and global reach are more important than ever to support leading infrastructure."
The AI stakes
I'm sure you're like me and you're tired of hearing about AI, and I have another post in the works around applications of artificial intelligence that I'm actually excited about. But please bear with me for the next 300 or so words.
This concentration of cable ownership isn't happening in a vacuum. It's happening precisely as AI becomes the dominant paradigm of the internet. As technology companies build up their cloud and AI capabilities and construct more data centers globally, they also invest in more subsea cables. In my opinion, the implications are profound. While ownership of a cable doesn't necessarily mean access to data flowing through it, owners can determine which data flows are prioritized, as well as exploit their control over these chokepoints to favor their own services.
Consider what this means for the global AI race that we're in. Training large language models requires moving massive datasets between facilities. Running AI services demands ultra-low latency connections between data centers. For companies dealing with petabytes of user data daily, shaving milliseconds off transit times can translate into millions in revenue.
Google's Equiano cable isn't just a connectivity project; it's a statement of dominance in Africa's burgeoning digital economy. The same pattern is playing out globally: tech giants are using cable ownership to cement their advantage in emerging markets where AI adoption will explode over the next decade.
What happens next
I remember leaving the lecture that I was in and immediately walking home and having some thoughts around: how can a country protect itself if core infrastructure is owned by tech giants? It's a difficult balance. I'm a huge fan of innovation and startups, but I also believe in public infrastructure. When it comes to cables, I actually see value in tech companies owning some of them; they have the money and incentive to maintain them properly. But countries have equal skin in the game since it's our data flowing through these undersea cables.
We're witnessing a fundamental shift in how the internet works. What began as a global commons governed by shared protocols is becoming a series of private networks owned by a handful of corporations and contested by nation-states. The cables that carry our AI models, financial transactions, and daily communications are no longer just infrastructure; they can be used as weapons, diplomatic leverage, and the backbone that lets tech giants control global markets.
Intergovernmental organizations such as the UN or the International Telecommunication Union need to take undersea cable security seriously and establish some sort of internationally followed protocols under a formalized protection plan, because these invisible threads that connect our world are becoming very visible. I don't know if we'll build new frameworks to govern them before the next cable is cut, whether it's accidental or not, but it's something that we need to keep an eye on and try to fight for.