The North American Network Operators Group held its latest meeting in Miami in mid-October 2026. Its opening keynote, delivered by Telegeography’s Paul Brodsky, examined the forces reshaping global Internet infrastructure, including growing bandwidth needs, hyperscaler-owned cables and low-Earth-orbit (LEO) satellite services.
Internet demand continues to rise, although growth has slowed from the exceptionally high rates seen in the early 1990s. Brodsky’s presentation put current annual growth in international bandwidth at about 30%. That still means adding roughly 2 petabits per second (Pbps) of capacity each year, with the annual requirement projected to exceed 10 Pbps by 2035.
Content delivery networks and replicated service points have helped manage growth by placing content and services closer to users. But the pressure to expand capacity remains, while much of the submarine cable infrastructure installed around the turn of the century is approaching the end of its roughly 25-year service life. About a quarter of the world’s existing subsea cable kilometres are expected to be retired by 2030.
New cable technology has increased capacity per system and lowered the unit cost of bandwidth. The combination of ageing infrastructure, growing demand and the need for more diverse routes is driving new projects. Announced cable investment is expected to total between $5 billion and $6 billion annually from 2026 to 2029, or about $22 billion across the period.
Google, Meta, Microsoft and Amazon are playing a larger role in funding these systems. According to the figures presented, the four companies fund 38% of cable investments for their own private use and participate in consortium funding for another 25%. The remaining 37% has no hyperscaler involvement. The transatlantic route is particularly concentrated, with 85% of investment privately funded by hyperscalers.
The shift reflects the companies’ ability to finance and operate networks for their own services, as well as the risks of relying on a consortium when one customer accounts for a large share of a cable’s capacity leases. It also raises concerns among some governments about foreign private ownership of infrastructure considered critical, including control of repair vessels and maintenance arrangements.
LEO satellite networks are also expanding the options for global connectivity. Unlike geostationary satellites, which orbit at about 36,000 kilometres and introduce significant delays, LEO systems use large constellations of smaller satellites. Their spacecraft hand off user connections as they move, while laser links can relay data across the constellation and help cover oceans with fewer ground stations. The presentation said these services can deliver hundreds of megabits per second, with capacity and delay comparable to terrestrial broadband for users with an antenna.
Brodsky contrasted the scale of investment in satellite platforms with submarine cables. SpaceX has spent more than $20 billion on Starlink, while Amazon’s LEO network is expected to cost $20 billion. Starlink reported $5.25 billion in revenue and $2.4 billion in net profit in the third quarter of 2026, according to the presentation.
The two types of infrastructure serve different roles. LEO services focus on last-mile access, particularly in remote or underserved areas, while submarine cables carry large volumes of data between networks and data platforms. Brodsky described them as complementary rather than direct competitors, while noting that Amazon’s service could challenge Starlink’s position in its market. Its eventual positioning remains to be seen.
Sources: APNIC Blog.