Meta aims to double Aurora's 497 terabits a second on cables it owns

One petabit per second is a thousand terabits, or a million gigabits ^[2], and Aurora, the US-to-Denmark system Meta expects in service in 2028, is designed for about 497 terabits ^[5]. The petabit figure Meta's engineers describe is therefore about 2.01 times Aurora's design capacity ^[12]. Waterworth's published architecture is 24 fibre pairs ^[4]. Spread a petabit across 24 pairs and each pair carries roughly 41.7 terabits ^[13]. But CryptoBriefing attaches the 24-pair count to Waterworth and the petabit work to longer routes in general ^[3][1]. That per-pair number is arithmetic on two separate figures. It is not a spec.

The ownership story is the part a buyer of Meta stock has to take on the publisher's word. CryptoBriefing says tech companies leased capacity on consortium-built cables for years and that hyperscalers including Meta, Google and Microsoft have moved to building and owning systems outright ^[8]. Subsea cables carry more than 95% of intercontinental internet traffic ^[7]. Owning changes the timing of the cash. Aurora carries nothing until 2028 ^[5], so the money goes out years ahead of the traffic. What Meta is not doing is buying leased capacity it could resize annually.

CryptoBriefing called Project Waterworth a multi-billion dollar investment ^[9]. Waterworth runs more than 50,000 kilometres, with landing points on five continents including the United States, India, Brazil and South Africa ^[3], 24 fibre pairs, and protection rated to 7,000 metres of water ^[4]. Waterworth's route runs about 6.9 times Aurora's 7,268 kilometres ^[14].

Petabit-class cables are an industry-wide target for roughly 2030 ^[6], two years after Aurora is due to light up ^[15]. Meta reaching a petabit would put it on the same clock as the rest of submarine telecom. Google has been investing in subsea cable for over a decade, with Equiano between Europe and Africa and Grace Hopper between the US, the UK and Spain. Microsoft has partnered on multiple projects, and Amazon has expanded AWS's subsea footprint ^[10].

In my view the durable asset is the route and the landing rights. Where a cable lands and who owns it draws regulators and national security agencies ^[11], and a permit is harder to replicate than a fibre pair. The counter-thesis is straightforward. If cross-ocean demand grows slower than 24-pair builds add capacity, Meta ends up a wholesaler of transit to the consortiums it stopped renting from ^[8]. The third path is duller still: permits or engineering push the 2028 date out. A capex breakdown putting the subsea programme at a low single-digit share of Meta's capital spending would knock the ownership reading down to a procurement footnote.