The invisible border that splits the Internet in two

In 2021, American diplomats fanned out across Asia with an unusual message. Not about treaties, not about tariffs, not about troop deployments. They were warning governments about a fiber-optic cable.

SEA-ME-WE 6: 19,000 kilometers of undersea cable running from Marseille to Singapore, passing through the Indian Ocean, touching shore at a dozen countries that together hold roughly a third of the world’s population. HMN Technologies — formerly Huawei Marine Networks, acquired by Chinese firm Hengtong Optic-Electric in 2020 — had entered the lowest bid. By most measures, they were going to win. Then the State Department got involved. American embassies delivered warnings to consortium members: accept HMN’s contract and face consequences, possibly including sanctions. By February 2021, the bidding was reopened. SubCom of New Jersey, whose bid came in higher, won the contract.

Hold that image. The full machinery of US foreign policy — ambassadors, cables, formal diplomatic warnings, the credible threat of financial punishment — deployed over a procurement decision about who lays fiber-optic cable on the ocean floor.

What exactly is at stake in a decision like that? The usual answer is surveillance: Chinese equipment in critical infrastructure means Chinese intelligence access. That’s true, and it matters, and it’s not the whole story. Not even close. What’s actually at stake is something more fundamental: who controls the physical and logical architecture that makes global internet connectivity possible at all, and what happens to that architecture when the countries building it no longer share the same model of what the internet is for.

The fracture of the global internet is not primarily a censorship story. It is an infrastructure story. And the infrastructure is already being divided.

The wrong story

The Great Firewall is real. China’s apparatus for content control is sophisticated, extensively documented, and genuinely effective: DNS poisoning redirects queries for blocked domains to nothing or to Chinese alternatives; IP blocking prevents direct connection to foreign servers; deep packet inspection scans traffic for keywords and disrupts connections that carry them. Hundreds of millions of people in China cannot reach Google, Facebook, or the New York Times without a VPN, and the VPNs work until the government decides they don’t. The Wall exists. It works.

But the Great Firewall operates at the upper layers of internet architecture — the layer where applications talk to each other, where domain names get resolved into addresses, where content is identified and filtered. It is a gate, not a foundation. Aggressively effective as it is, it doesn’t touch the deeper questions of who controls routing infrastructure, who maintains the naming systems that make every addressable destination findable, who sets the technical standards that determine what equipment can connect to what, who owns the cables that carry traffic between continents.

The internet has layers. The one most people experience — web pages, apps, streaming video, email — rides on top of a stack of infrastructure that most people never see and rarely think about. Censorship operates near the top of that stack. The fight that will actually determine whether the internet remains a single system or becomes two separate ones is happening much further down.

When you censor, you decide what can be seen. When you control infrastructure, you decide what can exist.

Two models, one address book

The internet’s naming and numbering system — the infrastructure that assigns every domain name, every IP address, every routing protocol — has been governed since 1998 by ICANN, the Internet Corporation for Assigned Names and Numbers. ICANN was created under contract to the US Department of Commerce. The department’s National Telecommunications and Information Administration (NTIA) held oversight authority over the IANA functions — the actual technical operations that maintain the root zone of the domain name system. This arrangement was always temporary, and always embarrassing. Ostensibly a technical body, ICANN was indisputably an American one.

The competing model runs through the ITU, the International Telecommunication Union — a UN agency founded in 1865 to coordinate telegraph infrastructure among sovereign states. The ITU’s logic is that internet governance belongs to states, not corporations or “multistakeholder communities,” and that each nation has legitimate authority over the internet infrastructure within its borders. It is a model built for government control, dressed in the language of sovereignty.

The conflict broke into the open in December 2012, in Dubai. At the World Conference on International Telecommunications — WCIT-12 — a bloc of countries led by Russia, China, UAE, Saudi Arabia, Algeria, Sudan, and Egypt proposed revisions to the International Telecommunication Regulations that would give states equal rights in managing internet naming, numbering, and addressing. The revised treaty was signed by 89 states. Fifty-five — including the United States, France, Germany, and the United Kingdom — rejected it and walked out. The split was clean and has not healed.

The sovereignty bloc had a real grievance and used it strategically. NTIA’s oversight of ICANN meant, structurally, that the US government held a backstop position over the internet’s address book — and from 2013, Snowden’s revelations established that the NSA had been using that position to access data flowing through American internet infrastructure. The grievance was genuine. The conclusion drawn from it — that states should therefore control their own internet infrastructure — does not follow from the evidence.

The Snowden factor

When PRISM was revealed in June 2013, the US-led multistakeholder model lost significant credibility with swing states. The NSA had access to data flowing through American internet companies — the same infrastructure the multistakeholder model had entrusted to American firms. Brazil's President Dilma Rousseff made it explicit at the UN General Assembly that September: the US had exploited its structural position for mass surveillance. The irony is structural and complete: the strongest argument ever made for state sovereignty over the internet came from US government conduct. The IANA transition in October 2016 removed the most legitimate grievance. The sovereignty bloc was not moved. Which tells you what they were actually after.

The legitimate part of that complaint was addressed in October 2016, when the NTIA contract with ICANN expired and was not renewed. The IANA functions transitioned to a community-steered body after a two-year multistakeholder process. US government backstop authority over the internet’s address book was formally ended. The sovereignty bloc was unmoved. Russia and China continued pressing at ITU plenipotentiary conferences in 2014 in Busan and in 2018 in Dubai; in 2022, at the ITU Plenipotentiary Conference in Bucharest, Russia ran a candidate for Secretary-General and lost to US-backed Doreen Bogdan-Martin by 139 votes to 25. That same year, a CFR task force concluded simply that the era of the global internet is over.

What the sovereignty bloc actually wants is not a fairer internet. It is state authority over information infrastructure — authority that, in practice, means the ability to monitor, filter, and cut flows of information to citizens. The framing is sovereignty. The mechanism would be control. The IANA transition removed the strongest legitimate argument for the sovereignty model. What remains is autocratic preference without the cover.

Who owns the ocean floor

Roughly 95 percent of intercontinental internet data flows through approximately 550 submarine cable systems. Satellites carry the remainder. The cables are mostly invisible — running across ocean floors at depths of up to eight kilometers, sheathed in steel and polyethylene — until they break or until someone threatens to sanction the companies that build them.

HMN Technologies entered submarine cable construction as Huawei Marine Networks around 2008 and built up meaningful market share on the strength of competitive pricing and technical competence. Then the US government started treating cable contracts as a front in the broader technology competition with China. SEA-ME-WE 6 was the sharpest example, but not the only one. US intervention touched at least six cable projects across the Asia-Pacific over four years. The mechanism was consistent: warn consortium members, threaten consequences, wait for the calculus to shift.

The effect on routing is physical and lasting. Carnegie Endowment for International Peace analysis from December 2024 documents the geopolitical pressure reshaping cable deployment in Southeast Asia: China’s extensive claims under the Nine-Dash Line have delayed the SJC2 cable for years and prompted the deliberate rerouting of at least two cable systems — Apricot and Echo — around Indonesia rather than through the South China Sea. The cost of avoidance is real: circumventing the most direct route between Southeast Asia and the Pacific coast of the United States adds cable length and sheathing expense absorbed by transit economies, not by the major powers whose competition created the problem.

The physical infrastructure question is not purely about surveillance, though the surveillance risk is real. Any party with access to a cable — at the cable itself, or at the onshore landing station where it comes ashore — can intercept traffic. A small number of cities and countries host the majority of landing stations; whoever controls those onshore chokepoints determines what interception is structurally possible. The US and UK’s GCHQ understood this. The TEMPORA program, revealed by Snowden, tapped fiber-optic cables at UK landing stations at industrial scale.

But cable ownership also determines routing, latency, and cost. Security and economics don’t separate cleanly. When the US removes a Chinese vendor from a cable consortium, it is simultaneously a security decision, a commercial decision, and a geopolitical decision — embedding American infrastructure preferences into the physical fabric of global connectivity.

On August 7, 2025, the FCC formalized this logic. The Commission adopted a Report and Order establishing a presumption of denial for cable-landing license applications from entities controlled by designated foreign adversaries, and explicitly banned equipment from Huawei, ZTE, China Telecom, and China Mobile in US-connected undersea cables. The rule creates cybersecurity and physical security requirements for all such cables.

The policy is not a content regulation. It’s about who controls the medium — the wires through which every bit of data reaching American shores must pass.

Cable interception: a history

On August 5, 1914 — one day after Britain declared war on Germany — the Royal Navy cut Germany's transatlantic telegraph cables off the coast of Emden. Germany was forced to route communications through cables Britain controlled or could monitor. During the Cold War, the US Navy's Operation IVY BELLS tapped Soviet undersea cables in the Sea of Okhotsk, recording communications that were retrieved by submarine divers. The GCHQ/NSA TEMPORA program, revealed by Snowden in 2013, was not a hack or an exploit — it was systematic tapping of fiber-optic cables at UK landing stations, with the cooperation of the companies that operated them. The architecture of cable interception is as old as the cables themselves. The threat model is not paranoid. It is historical record.

The address book fracture

A cable determines whether data can travel. The Domain Name System determines whether it can be found. DNS is the address book of the internet: type a domain name, and DNS resolves it to an IP address. That resolution process begins with root servers — 13 clusters, predominantly operated by American and European institutions — which sit at the top of the hierarchy and ultimately authorize what is addressable.

Russia has decided it cannot depend on that hierarchy. Federal Law No. 90-FZ, signed May 1, 2019 and effective November 1 of that year, requires deep packet inspection equipment installed at chokepoints under Roskomnadzor control, and establishes a national DNS system capable of resolving Russian domains without referencing international root servers. 90-FZ is not a firewall. It is the architecture of a separable internet — one that can keep running if Russia severs the connection to the global network, and one from which Russia can do the severing.

Russia tested this architecture from June 15 to July 15, 2021. All major Russian telecoms participated. According to preliminary results reported by RBC and confirmed by Russian state media, the tests were assessed as successful, including testing of the capability to physically disconnect Russian internet from the global network. Independent assessment of the claims is limited by the secrecy of the process. What is not limited is the legislative reality: the architecture exists, it is operational, and Russia has demonstrated intent to use it.

China’s approach differs. The Great Firewall filters — DNS poisoning, IP blocking — rather than building routing-layer separation. The structural mechanism is data localization. China’s Cybersecurity Law, effective 2017, requires critical information infrastructure operators to store personal data and “important data” within China. The legal effect is that significant categories of data about Chinese users and operations cannot flow outside China’s borders at all.

Russia’s Federal Law No. 242-FZ, effective September 2015, requires personal data of Russian citizens to be processed on servers physically in Russia. The OECD documented the proliferation of this model in Trade Policy Paper No. 278 from November 2023: by early 2023, over 100 data localization measures had been enacted across 40 countries, with more than two-thirds combining local storage requirements with restrictions on cross-border data flows.

Sovereign DNS plus data localization equals separable network architecture. Not aspirational policy positions — legislated, installed, and in Russia’s case operationally tested. The question of whether the internet will fracture is the wrong question. In significant measure, it already has. What remains open is how deep the separation goes.

How DNS root zones work

When your computer wants to reach a website, it asks a DNS resolver for the IP address of the domain. That resolver, if it doesn't have the answer cached, eventually queries a root name server — one of 13 logical clusters that know which servers are authoritative for each top-level domain. The root zone is maintained by ICANN. If Russia's national DNS stops querying international root servers, two things happen simultaneously: Russian domains become unreachable from outside Russia (because external resolvers can no longer find authoritative servers for .ru domains through the normal hierarchy), and foreign domains become unreachable from inside Russia unless they are mirrored in Russian infrastructure. A content filter blocks specific websites. This is topological separation — the logical equivalent of two separate internets that happen to use the same underlying protocols but cannot, structurally, find each other.

Standards, the deepest layer

The cables carry data. The naming systems find addresses. The standards determine what the data is, how it is formatted, what protocols govern its transmission, and what equipment can process it at all. Standards are the most consequential and least discussed layer of this conflict.

The assumption that technical standards are apolitical is a relic of an era when the organizations that set them were dominated by engineers from a handful of Western countries and Japan. The IETF operates by “rough consensus and running code” — engineer-led, bottom-up, resistant to state capture. The ITU operates by treaty, with voting rights assigned to member states. These are not equivalent governance models, and they have produced genuinely different standards in some areas. The contest over which governance model sets internet standards is, simultaneously, a contest over who controls the architecture of global communications.

China’s engagement with standards bodies is a decades-long strategic investment with measurable results. The Asia Society Policy Institute’s “Stacking the Deck” analysis documents the scale. In ISO — the International Organization for Standardization — Chinese secretariat positions in technical committees and subcommittees grew roughly 58 percent between 2011 and 2021. At the IEC — the International Electrotechnical Commission — Chinese-held secretariat positions grew by roughly two-thirds between 2012 and 2020. By 2020, China held more leadership positions in ISO/IEC combined than Germany or the United States — a first in those organizations’ histories.

In 3GPP, the body that develops 5G technical specifications, Chinese representatives held approximately a quarter of all chair or vice-chair positions as of April 2021, and Chinese voting members outnumbered US members more than two-to-one.

Leadership positions in standards bodies are not ceremonial. They control agendas, determine how time is allocated, and shape what proposals get serious consideration and which die of procedural inattention. The chair of a working group can slow-walk proposals they dislike and accelerate ones they support. Standards get made not by votes, mostly, but by the institutional work that happens before any vote is called.

In September 2019, Huawei, China Mobile, China Unicom, and China’s Ministry of Industry and Information Technology submitted a proposal to ITU-T’s Telecommunications Standardization Advisory Group: “New IP,” a wholesale redesign of foundational internet protocols. ICANN’s technical office analyzed the proposal in October 2020 (OCTO-017) and found that New IP would make pervasive monitoring substantially easier and would shift governance of internet protocols from engineer-led multistakeholder bodies to state-authorized processes. The Internet Society published a critical analysis reaching the same conclusion. The ITU did not advance New IP; at the December 2020 TSAG plenary, discussions were closed and deferred until the 2022 World Telecommunication Standardization Assembly.

Beijing did not retreat. It reformulated. At ITU forums in 2022, China presented IPv6+, an enhanced version of IPv6 developed substantially by Huawei, as an alternative pathway through the same venue. The European Telecommunications Standards Institute had launched a working group on IPv6 Enhanced Innovation in January 2021 with Huawei as a significant participant; by 2022 European concerns about the initiative’s direction had surfaced.

The pattern is legible: propose, encounter resistance, rename, resubmit. Not failure. Strategic persistence — and the distinction matters, because only one of those will stop.

The National Standardization Development Outline — issued October 2021 by the CCP Central Committee and State Council, published by Xinhua, translated by Georgetown’s Center for Security and Emerging Technology — sets near-term and medium-term targets for Chinese participation and leadership in international standards bodies, linking standards-setting to virtually every strategic technology sector. It commits China to increasing the proportion of Chinese-led international standards and expanding leadership in standards bodies, formalizing as state policy the roughly 20 percent annual growth rate in ISO and IEC submissions that Chinese organizations had already been achieving for over a decade.

The Outline is not an engineering document. It is a political claim on the architecture of global communications.

The standards that govern the physical network of 2035 are being designed in committees meeting now. Whatever gets standardized gets built into every router, every base station, every piece of networking equipment manufactured to those specifications. You cannot retrofit a different governance model onto already-deployed infrastructure.

What two Internets cost

Who pays? The IMF estimated in 2023 that severe geoeconomic fragmentation could reduce global output by up to 7 percent of GDP — roughly $7.4 trillion — with scenarios incorporating technology decoupling reaching up to 12 percent in some countries. These figures measure geoeconomic fragmentation broadly — trade restrictions, supply chain disruption, investment barriers — not internet infrastructure specifically. They are an upper bound on a related phenomenon, not a precise estimate of connectivity costs alone. With that scope stated, they are still enormous.

The operational costs are already running. In early 2023, all five of Vietnam’s international submarine cables suffered partial or total damage simultaneously, causing a 75 percent loss of the country’s international data flow over the Lunar New Year period. Providers compensated partly by purchasing additional bandwidth over land cables through China — the cables still functioned, but traffic was rerouted through a neighbor whose relationship with internet freedom is not ambiguous. Three of those five cables failed again in 2024. Cable repairs cost between one and three million dollars per incident, take months, and recur. Vietnam is paying the price of being geographically caught between the competing infrastructure systems. It built connectivity cheaply, using whatever was available. That was rational. It also meant it had no control over the political facts embedded in its infrastructure.

The same dynamic runs at hyperscale. Mastercard spent approximately $350 million building a data center in Pune to comply with the Reserve Bank of India’s data localization mandate for payment data — a direct compliance cost, not an expansion choice. AWS committed $12.7 billion in India through 2030, and Microsoft $3 billion over two years; both companies frame these primarily as cloud and AI infrastructure expansion rather than localization compliance, though India’s data governance requirements shape the context in which both investments are made. These are not discretionary numbers.

Countries paying the highest costs are neither the United States nor China. Both have the population, market size, and technical capacity to sustain separate internet ecosystems. The countries between the blocs cannot. Vietnam, with Chinese-linked cable infrastructure and US routing pressure simultaneously. Indonesia, enforcing its own data localization requirements while hosting both American and Chinese platforms. African countries that built connectivity substantially through Huawei equipment and Chinese-built networks under Belt and Road financing, now facing incompatibility pressures from Western security requirements that demand those networks be replaced.

Infrastructure choices become foreign policy commitments whether intended or not. A country that built its 5G network with Huawei, routes international traffic through HMN cables, and stores cloud data on Alibaba infrastructure has already decided. Those decisions are now baked in — architectural facts with permanent political consequences. The internet’s designers believed they were building something that transcended geopolitics. They were wrong, and we inherit that error one deployment decision at a time.

The cable in the ground

SEA-ME-WE 6 is being laid by SubCom rather than HMN Technologies. The route still runs from Marseille to Singapore, still crosses the Indian Ocean, still touches shore at the same dozen countries. The data still flows. But the ownership structure is different, the routing agreements are different, the political facts embedded in the fiber are different.

The fracture of the global internet is happening at layers most people never see: in procurement decisions for cables on the ocean floor, in votes on technical committee chairmanships, in legislation requiring data to stay within national borders, in testing of sovereign DNS infrastructure on Sunday nights in July. Each of these is individually deniable as routine governance. Collectively, they are the assembly of two incompatible infrastructure systems that currently still interoperate and will, at some point, not.

The question is not whether the split will happen. The architecture for it already exists, in Moscow server rooms and data centers in Mumbai and standards committee agendas in Geneva. The question is how deep the fracture runs before structural incompatibility becomes irreversible — before the differences in cable ownership, DNS root governance, data localization, and technical standards compound into two internets that can no longer find each other. The choices determining that depth are being made now, in the rooms where engineers vote on protocol specifications and diplomats warn cable consortiums about sanctions. They are not being made in public.

The internet was designed as if borders did not exist. The people who designed it were wrong. We are inheriting that error, one cable at a time.

Zastrzeżenie dotyczące Gen AI

Niektóre treści tej strony zostały wygenerowane i/lub edytowane przy pomocy generatywnej sztucznej inteligencji.

Media

France Telecom Marine , cable-laying ship René Descartes , at sunset in the harbor of La Seyne sur Mer (near Toulon , France ) – Wikipedia

Kluczowe źródła i odniesienia

ICANN. “Stewardship of IANA Functions Transitions to Global Internet Community as Contract with U.S. Government Ends.” ICANN Announcement, October 1, 2016. https://www.icann.org/en/announcements/details/stewardship-of-iana-functions-transitions-to-global-internet-community-as-contract-with-us-government-ends-1-10-2016-en

National Telecommunications and Information Administration (NTIA). “Q and A on IANA Stewardship Transition.” US Department of Commerce. https://www.ntia.gov/other-publication/q-and-iana-stewardship-transition-0

Light Reading. “US hardball drove Chinese vendor from huge cable project: reports.” 2021. https://www.lightreading.com/cable-technology/us-hardball-drove-chinese-vendor-from-huge-cable-project-reports

Data Center Dynamics. “Two Chinese operators withdraw from Sea-Me-We 6 subsea cable project as tensions with US grow.” 2021. https://www.datacenterdynamics.com/en/news/two-chinese-operators-withdraw-from-sea-me-we-6-subsea-cable-project-as-tensions-with-us-grow/

Submarine Networks. “SEA-ME-WE 6.” https://www.submarinenetworks.com/en/systems/asia-europe-africa/smw6

Federal Communications Commission. “Review of Submarine Cable Rules — Report and Order and Further Notice of Proposed Rulemaking.” FCC-25-49A1. Adopted August 7, 2025. https://docs.fcc.gov/public/attachments/FCC-25-49A1.pdf

Carnegie Endowment for International Peace. “Subsea Communication Cables in Southeast Asia: A Comprehensive Approach Is Needed.” December 2024. https://carnegieendowment.org/research/2024/12/southeast-asia-undersea-subsea-cables

Council on Foreign Relations Task Force. “Confronting Reality in Cyberspace: Foreign Policy for a Fragmented Internet.” CFR Task Force Report, 2022.

ITU. “World Conference on International Telecommunications (WCIT-12) — Final Acts.” Dubai, December 3–14, 2012. https://www.itu.int/en/wcit-12/Pages/default.aspx

ITU. “Member States elect Doreen Bogdan-Martin as ITU Secretary-General.” ITU Press Release, September 29, 2022. https://www.itu.int/en/mediacentre/Pages/PR-2022-09-29-ITU-SG-elected-Doreen-Bogdan-Martin.aspx

Euronews. “Russia disconnects from internet in tests as it bolsters security.” July 22, 2021. https://www.euronews.com/next/2021/07/22/us-russia-internet

Russian Federal Law No. 90-FZ “On Communications” (amendments establishing the sovereign internet), signed May 1, 2019. Analysis: DGAP (German Council on Foreign Relations).

Russian Federal Law No. 242-FZ on personal data localization, effective September 1, 2015. Analysis: Stanford WILmap.

People’s Republic of China Cybersecurity Law, Article 37, effective June 2017. Analysis: Proskauer Rose LLP client advisory.

OECD Trade Policy Paper No. 278. “Data Localisation Measures: Impacts and Policy Implications.” November 2023.

ICANN Office of the Chief Technology Officer. “OCTO-017: New IP, Shaping Future Network.” October 27, 2020. https://www.icann.org/en/system/files/files/octo-017-27oct20-en.pdf

Internet Society. “Huawei’s ‘New IP’ Proposal FAQ.” February 2022. https://www.internetsociety.org/resources/doc/2022/huaweis-new-ip-proposal-faq/

Internet Society. “Discussion Paper: An Analysis of the ‘New IP’ Proposal to the ITU-T.” April 2020. https://www.internetsociety.org/resources/doc/2020/discussion-paper-an-analysis-of-the-new-ip-proposal-to-the-itu-t/

ETSI. “ETSI Launches New Group on IPv6 Enhanced Innovation.” ETSI Newsroom, January 2021. https://www.etsi.org/newsroom/news/1877-2021-01-etsi-launches-new-group-on-ipv6-enhanced-innovation

Asia Society Policy Institute / Daniel Russel and Blake Berger. “Stacking the Deck: China’s Influence in International Technology Standards Setting.” November 2021. https://asiasociety.org/policy-institute/stacking-deck-chinas-influence-international-technology-standards-setting

Center for Security and Emerging Technology (CSET), Georgetown University. Translation and analysis of “National Standardization Development Outline” (China, October 2021). https://cset.georgetown.edu/publication/the-chinese-communist-party-central-committee-and-the-state-council-publish-the-national-standardization-development-outline/

IMF Finance and Development. “Geoeconomic Fragmentation and the Future of Multilateralism.” June 2023.

IMF Blog. “Geoeconomic Fragmentation: What’s at Stake for the Global Economy.” August 2023.

TeleGeography. “Submarine Cable FAQs.” https://www2.telegeography.com/submarine-cable-faqs-frequently-asked-questions

Privacy International. “GCHQ Tapping into International Fibre Optic Cables, Shares Intel with NSA.” https://privacyinternational.org/long-read/1678/gchq-tapping-international-fibre-optic-cables-shares-intel-nsa

Washington Post. “Brazil’s president condemns NSA spying.” September 24, 2013. https://www.washingtonpost.com/world/national-security/brazils-president-condemns-nsa-spying/2013/09/24/fe1f78ee-2525-11e3-b75d-5b7f66349852_story.html

Analytics India Magazine. “Mastercard Begins RBI Data Localization Compliance By Deleting Indian Users Data.” https://analyticsindiamag.com/mastercard-begins-rbi-data-localization-compliance-by-deleting-indian-users-data/

TechCrunch. “Amazon’s AWS to invest $12.7 billion in India by 2030.” May 17, 2023. https://techcrunch.com/2023/05/17/aws-to-invest-12-7-billion-in-india/

Microsoft Stories India. “Microsoft announces US $3bn investment over two years in India cloud and AI infrastructure.” January 2025. https://news.microsoft.com/en-in/microsoft-announces-us-3bn-investment-over-two-years-in-india-cloud-and-ai-infrastructure-to-accelerate-adoption-of-ai-skilling-and-innovation/

VietnamPlus. “Undersea cable problems affect Vietnam’s Internet connections.” January 3, 2025. https://en.vietnamplus.vn/undersea-cable-problems-affect-vietnams-internet-connections-post307765.vnp

The Register. “Vietnam’s internet again in trouble as 3/5 sub cables cut.” June 18, 2024. https://www.theregister.com/2024/06/18/vietnam_internet_cables/