Call: 1-877-697-2926

How Big Tech Fuels the AI Boom with Massive Undersea Cable Investments

Artificial intelligence isn't just reshaping software—it's redefining the internet's physical foundation. As AI models grow in complexity and computing intensity, the need to move massive volumes of data at high speed across the globe has surged. Whether training a large language model or powering real-time recommendations, data transfer infrastructure has quickly become a strategic asset.

At the center of this shift, Big Tech—led by Google, Amazon, Meta, and Microsoft—is accelerating its investments in transoceanic fiber-optic networks. These companies no longer just lease bandwidth; they build, own, and operate submarine cable systems that stretch thousands of miles across ocean floors. Their motives go beyond connectivity. AI workloads generate unprecedented volumes of data, and moving it efficiently requires bypassing traditional telecom bottlenecks.

This wave of infrastructure investment isn’t reactive—it’s foundational. Fiber routes are being selected and designed based on AI architecture needs, placing cable landing stations closer to hyperscale data centers and optimizing paths for latency-sensitive machine learning traffic. The result is a global mesh of undersea lines tailored to the demands of generative AI, large-scale inference, and edge computing. This piece will unpack the intertwined rise of AI and submarine cable investment, tracing how Big Tech is transforming the planet’s digital plumbing to match the scale of their algorithms.

Big Tech’s Growing Appetite for Infrastructure

Google, Amazon, and Meta: From Platforms to Infrastructure Titans

In less than a decade, the leading names in Big Tech have evolved from software and platform providers into commanding forces in global infrastructure. Google, Amazon, and Meta now rank among the largest investors in undersea cable systems—assets once dominated by telecom conglomerates. Each has adopted distinct strategies, but they converge on one point: control the pipes, and you control the data.

Google leads the pack in privately-owned subsea investments. It has directly funded or co-owned more than 20 cable systems spanning hundreds of thousands of kilometers. Projects like Dunant (U.S. to France), Equiano (Portugal to South Africa), and Firmina (U.S. to Argentina) highlight its end-to-end infrastructure ambitions. Google doesn’t just lease bandwidth—it builds the pathways.

Amazon Web Services (AWS) has rapidly caught up, opting for co-ownership stakes and strategic partnerships. The company invests in consortium cables but has also invested in private routes for redundancy and performance. Cables such as Jupiter (linking the U.S. with Japan and the Philippines) and Hawaiki (connecting Australia, New Zealand, and the U.S.) contribute to AWS's global data backbone.

Meta, formerly Facebook, has pursued aggressive cable expansion to support its social platforms and next-gen projects like the metaverse. It co-owns over a dozen cable systems, including 2Africa—an ambitious 45,000-kilometer ring around the African continent. For Meta, cable ownership is about future-proofing access to bandwidth-hungry applications and reaching billions of users in underserved regions.

Microsoft: The Cloud Giant’s Alternative Route

While Google and Meta emphasize physical cable ownership, Microsoft has taken a measured approach, focusing more on data center proliferation and strategic connectivity nodes. Its stake in projects like Marea—a joint venture with Facebook and Telxius connecting Virginia to Spain—demonstrates alignment with partners rather than full independence. Microsoft’s priority remains ensuring seamless Azure cloud delivery to enterprises worldwide through layered infrastructure strategies.

From Bandwidth Leasing to Cable Ownership

The industry has shifted. In the 2000s, tech firms leased transoceanic capacity from telecom companies. Today, they build their own or co-develop routes to guarantee performance, security, and scalability. By 2023, content providers like Google, Facebook, Amazon, and Microsoft accounted for approximately 69% of global international bandwidth usage, according to TeleGeography. And the more they consume, the more they invest.

Owning cables eliminates reliance on third parties, reduces operational costs, and accelerates rollout of latency-sensitive services—AI inferencing, real-time analytics, streaming, and cloud computing all benefit. The more data flows through Big Tech systems, the more owning the ground—or seabed—it travels on becomes a strategic necessity.

So where does this leave carriers and traditional infrastructure players? In many cases, no longer in the driver’s seat.

The Rise of AI Workloads and the Surge in Data Demands

Generative AI applications aren't just pushing the boundaries of innovation—they’re stretching the limits of existing digital infrastructure. From large language models (LLMs) like GPT-4 to advanced vision-language systems, the sheer volume of data required for training and inference has skyrocketed. These models ingest vast troves of text, images, audio, and sensor data, requiring rapid, high-capacity intercontinental data transmission. At the center of this escalation lies one critical component: bandwidth.

Bandwidth: Strained by Training, Stressed by Inference

Training a single state-of-the-art LLM involves petabytes of data. For example, GPT-3 was trained using around 300 billion tokens of text, and GPT-4 likely exceeded that figure by a wide margin. This training doesn’t happen in one place. Instead, it’s distributed across global clusters that depend on ultra-fast, low-latency data linkage. This is where undersea cables carry the weight.

Once models are deployed, the pressure shifts. Inference, the stage where users interact with AI, generates trillions of individual queries. Each one requires access to model parameters stored in multiple regions, returning real-time outputs. Between 2020 and 2023, average data center traffic attributable to AI workloads grew from 1.2 exabytes per month to over 5 exabytes, according to Cisco's Global Cloud Index.

From Bulk Transfers to Global Synchronization

Training traffic resembles bulk shipping: massive, periodic data movements during pre-scheduled model updates. In contrast, inference mirrors global air traffic—lightweight, high-frequency, constant. Both demand distinct routing optimization and capacity management.

As corporations push toward multi-region architectures to minimize latency and build fault tolerance, data synchronization rates across global clusters have spiked. AI-enabled applications like autonomous vehicles, real-time translation, or generative design tools need fresh model access every millisecond. That translates into terabits of persistent intercontinental exchange, round-the-clock.

Anticipating the Next Generation of Demand

It’s not just current bandwidth that matters—it’s the scale of what's coming. OpenAI’s CEO has predicted model training costs reaching hundreds of millions of dollars per run, which implies far higher underlying data movement. When planning for such growth, data infrastructure teams focus on pipeline resiliency, latency compression, and most critically, scalable throughput.

Google’s Grace Hopper cable, Meta’s 2Africa project, and Microsoft's participation in MAREA illustrate this forward-looking capacity planning. These are not speculative investments—they’re calculated moves to lay the foundation for an AI landscape where model runs, edge inference, and data localization all demand concurrency on a global scale.

In every scenario, subsea networks act as the circulatory system enabling AI to function across borders with zero tolerance for lag. The faster AI scales, the deeper Big Tech must dig—literally—beneath the ocean floor.

Why Undersea Cables Matter More Than Ever

The Physical Backbone of the Global Internet

Despite the ubiquity of wireless connections, over 95% of all intercontinental data — from emails to high-frequency trades — moves through a planet-spanning mesh of fiber-optic undersea cables. Satellites may offer reach, but subsea systems deliver far superior bandwidth and latency. These cables, laid thousands of meters beneath ocean surfaces, form the physical infrastructure that makes real-time digital communication possible across continents.

Historically, submarine cables served telecom carriers and hyperscale cloud providers. Today, Big Tech companies like Google, Meta, Amazon, and Microsoft are not just users but owners — actively financing and managing new cables to support their expanding cloud ecosystems and AI operations.

Data Applications Are Getting Heavier

Bandwidth consumption is no longer measured in megabits or gigabits. AI training datasets can exceed petabytes of information; a single self-driving car can generate up to 25 terabytes per hour. High-fidelity VR, 8K streaming, and cloud gaming intensify the load further. These applications demand consistency, speed, and massive data transfer capacity—factors undersea cables are uniquely equipped to handle.

As data payloads grow denser, connectivity infrastructure must evolve to keep pace. Subsea cables, engineered for both distance and capacity, carry the weight of this transition.

AI Is Reshaping Network Priorities: Throughput Over Latency

Up until now, latency optimization was the driving force behind submarine cable routing. Financial markets and mission-critical enterprise applications prioritized split-second speed advantages. AI changes that equation. What matters most for machine learning systems is throughput — moving gigantic datasets between cloud regions to fuel training processes.

Take Google's transoceanic Grace Hopper cable, linking the US and Europe. It introduces novel technology like optical switching at the fiber pair level, which increases flexibility in routing and bandwidth allocation. AI systems benefit far more from scalable, resilient, high-throughput pipelines than from ultra-low-latency routes.

Network engineering now centers around moving more data, more efficiently. That shift will define the next generation of subsea expansions — not simply new routes, but high-capacity corridors optimized for AI-native infrastructures.

Investment Trends: The New Arms Race Beneath the Sea

The flow of capital into subsea infrastructure has accelerated, led by the world’s most powerful tech firms. No longer satisfied with leasing bandwidth from traditional telecom firms, companies like Google, Amazon, Meta, and Microsoft are laying down their own global arteries of connectivity. These investments aren't auxiliary; they're core to the future of AI-enabled services and the hyperscale cloud economy.

Billions Flow into Digital Infrastructure

Between 2016 and 2023, private sector investments in subsea cable systems more than doubled, with a concentration around transcontinental routes vital for AI workloads. According to TeleGeography, private network operators accounted for over 60% of new submarine cable deployments in 2023, a sharp shift from a telecom-centric model that dominated until the early 2010s.

The driving force? Control. Owning cable infrastructure means controlling latency, capacity, and security—three pillars behind AI-driven, latency-intolerant applications like real-time translation, autonomous vehicle coordination, and cloud-based large-language model inference.

Google’s Projects: Equiano and Dunant

Google spearheaded its transition from tenant to builder with high-capacity cable investments such as Equiano and Dunant. The Equiano cable, stretching from Portugal to South Africa, offers a capacity of 144 Tbps, leveraging space-division multiplexing (SDM) to increase scalability. In contrast, the Dunant cable, connecting the U.S. and France, was the first to achieve 12-fiber-pair SDM design across the Atlantic, setting a technological precedent in 2020.

Both cables are engineered to move vast amounts of data, enabling real-time replication of AI models across continents. These systems aren't just fast—they're optimized for the distribution and synchronization of massive cloud-based compute operations.

Amazon’s Quiet yet Expanding Role

Amazon has traditionally kept a lower profile compared to peers, but reports from Bloomberg and capacity market analytics firms indicate that Amazon Web Services has increasingly financed exclusive cable paths. The company reportedly invested in fiber infrastructure linking Southeast Asia to its U.S. and Oceania cloud regions, aimed at bolstering the performance of its AI-heavy services and cloud products.

By investing privately in subsea infrastructure, Amazon cuts dependence on public transit networks, reduces risk of congestion, and achieves deterministic network traffic patterns—all essential for high-availability ML services like SageMaker and Bedrock.

Meta’s Strategic Expansion with 2Africa

Meta, through its consortium leadership in the 2Africa cable project, has taken a continent-scale approach. Spanning over 45,000 km, it will connect 33 countries in Africa, the Middle East, and Europe. Once completed, it will become one of the longest subsea cable systems ever deployed. This initiative aligns with Meta’s ambition to serve as the digital infrastructure backbone for emerging markets.

With future AI features inside WhatsApp, Instagram, and Horizon Worlds in development, Meta’s strategy goes beyond inclusion—it targets performance parity between developed nations and underserved regions.

The Rise of Hyperscalers as Primary Builders

Traditional telecoms no longer lead the race. Hyperscalers—defined as cloud firms operating at massive global scale—now fund, design, and even co-own cable systems once left to carriers. This shift marks a foundational change. According to Dell’Oro Group, over 65% of new intercontinental cables in 2022 had at least one hyperscaler as an anchor investor.

This realignment redefines the economics of subsea communications. While telcos focused on subscriber growth and bandwidth resale, tech firms engineer cables to optimize machine-to-machine traffic and cross-regional compute integration.

So, who truly owns the internet's veins today? Increasingly, it’s the AI platforms themselves—structured not just to deliver ads or store data, but to interpret, translate, and reason at scale in near-real time.

How the Market is Responding to the Subsea Demand

Build Schedules Accelerate, and So Do the Announcements

The number of newly planned and deployed subsea cables has surged dramatically. Between 2016 and 2023, the average number of active subsea cable systems increased from 285 to over 430 globally, according to TeleGeography. Each new cable system typically spans thousands of kilometers and includes multi-terabit capacity—specifically engineered to satisfy hyperscale data traffic driven by AI training and inferencing.

Announcements once spaced years apart now follow in rapid succession. 2023 alone saw Big Tech players involved in over 20 new transoceanic projects, with Google, Meta, and Microsoft leading the pack. These cables, including Firmina, Apricot, and Echo, are designed not only for raw throughput but low latency, synchronous compute, and redundancy.

Real Estate Demand Spikes at Key Landing Sites

Cable landings have shifted from sleepy coastal towns to white-hot market zones. Strategic geographies—like Mombasa, Bude, Fortaleza, and Singapore—are seeing intense acquisition and development activity. Demand is boiling over not only for cable landing stations (CLS) themselves but the interconnection hubs that link them with data centers inland.

Equinix, Digital Realty, and EdgeConneX are ramping up expansion efforts to co-locate with new cable landings, offering meet-me rooms, cross-connects, and edge-ready infrastructure tailored for AI and content delivery. In regions such as West Africa and Southeast Asia, CLS development is tied directly to countries’ digital strategies and their ability to attract AI companies aiming to train and deploy models closer to emerging markets.

Bandwidth Prices Flatten; Transit Becomes Strategic

The influx of new capacity is already reshaping economics. Transatlantic and transpacific bandwidth prices, which dropped annually by 20–30% over the past decade, continue that downward spiral as fresh capacity floods the market. For example, 100 Gbps wavelength prices across the Atlantic fell by more than 25% between 2022 and 2023, according to the latest pricing index from TeleGeography.

However, the impact extends beyond cost. IP transit markets are undergoing strategic recalibration. Tier 1 providers and CDNs must now factor infrastructure ownership and route redundancy into their SLA negotiations, as hyperscalers consolidate control over entire segments of the physical internet. This realignment reallocates power toward those with direct investments in subsea systems rather than those simply paying for bandwidth.

Public Agencies Step in with Policy and Oversight

Watchdogs and regulatory bodies have not remained passive observers. Government and international agencies—including the U.S. Federal Communications Commission, the European Commission, and the International Telecommunication Union—have taken keen interest in the shifting capital flows towards subsea infrastructure. Their influence is tangible in both permitting processes and national interest evaluations.

In 2023, the U.S. Team Telecom increased scrutiny of cable systems involving foreign state-owned entities, directly impacting build schedules for projects touching American territory. The European Union, meanwhile, ramped up funding for pan-European digital corridors through its Connecting Europe Facility (CEF Digital), allocating €258 million to projects, many with subsea elements.

This layered oversight creates a new dynamic where private investment must navigate state-backed digital ambitions. As the subsea boom reshapes digital geography, policymakers are responding by reasserting influence over how, where, and by whom cables are built and owned.

Public vs. Private Ownership: A Power Shift in Infrastructure

From Consortiums to Control: Redefining the Ownership Model

For decades, international subsea cables followed a predictable pattern: built and shared by telecom consortiums, collectively funded by multiple operators, and managed through agreements that ensured shared access and governance. This model prioritized broad interconnection and regulatory compliance, aligning with traditional policies favoring open telecommunications infrastructure.

That model is fading. Amazon, Google, Meta, and Microsoft no longer rely solely on third-party telcos to reach global data centers. They’re now bankrolling private subsea cable systems, securing majority or full ownership stakes. According to TeleGeography, as of 2023, over half of all new subsea cable investment involves Big Tech firms — a stark shift from less than 10% a decade earlier.

Why Direct Ownership Changes the Game

Owning a private cable grants far more than just bandwidth. It gives Big Tech the power to route data wherever and however they want, bypassing traditional chokepoints. Instead of leasing capacity from a shared system, they set the rules — managing IP traffic flow, latency optimization, and regional connectivity with precision. This unlocks not only performance gains but also cost efficiencies over time, especially when powering AI-driven services that require low-latency, high-throughput links across continents.

Private investment also enables faster deployment cycles. Consortium projects often stall due to multi-stakeholder disagreements; private ownership eliminates that friction. In 2021, Google announced it would install three undersea cables — Grace Hopper, Equiano, and Dunant — each privately funded and independently operated. With total control, rollout timelines accelerated, and strategic landing points could be chosen to align with internal data needs, not collective priorities.

Neutrality on the Line

One consequence of this shift lies in interconnection neutrality. Telco-operated cables historically committed to equitable access, selling capacity to a range of service providers, cloud platforms, and network carriers. But when Big Tech controls the infrastructure, they decide who connects and at what terms. Regulatory oversight becomes murkier when infrastructure is private and spans international waters.

Critics argue that private ownership risks fragmenting the internet’s backbone, allowing dominant players to create closed-loop ecosystems. If major cloud providers prioritize their own platforms and partners, smaller networks could be priced out or strategically excluded. This reshapes not only infrastructure access but also the competitive dynamics of AI services delivered over these networks.

The shift from public consortia to private ownership marks more than an investment trend — it redefines who holds the keys to digital infrastructure, how data moves across continents, and who gets to participate in that flow.

Geopolitical Tensions: The New Cold War of Connectivity

Undersea Cables as Critical National Infrastructure

Subsea fiber-optic cables now form the nervous system of global communications. Powering over 95% of international internet traffic, they underpin AI-driven data flows, cross-border cloud computing, and real-time financial transactions. Their role has shifted from neutral conduits to assets of strategic importance. Nations now classify them alongside roads, ports, and power grids—as infrastructure with direct implications for national security.

Scrutiny Across Borders: Routes, Equipment, and Investors

Governments have intensified scrutiny over who builds, owns, and operates undersea infrastructure. Cable routes that pass through or near waters of geopolitical tension attract particular attention. The origin of cable equipment, often manufactured by firms in strategic competitor nations, has also become a focal point. Investment sources are no longer evaluated solely by capital strength but by political alignment.

The U.S. Committee for the Assessment of Foreign Participation in the United States Telecommunications Services Sector (Team Telecom) has blocked or delayed multiple projects involving Chinese stakeholders. Similar vetting is underway across Asia-Pacific and Europe, especially in the context of AI-scale data latencies and throughput demands.

China-US Cable Disputes and European Regulatory Responses

Tensions between the U.S. and China have transformed undersea cable projects into diplomatic flashpoints. In 2020, the Pacific Light Cable Network, a high-capacity subsea cable backed by Google and Meta (then Facebook), was denied permission by U.S. regulators to connect directly with Hong Kong, citing national security concerns. Project stakeholders were forced to reroute and terminate the cable at the Philippines and Taiwan instead.

In Europe, regulation isn't just reactive—it’s structural. The European Union's Digital Decade policy framework includes provisions to strengthen sovereignty over digital infrastructure. Member states have pushed for supply chain transparency and proposed joint development funds to oversee every phase of subsea construction—design, manufacturing, deployment, and long-term ownership.

Digital Sovereignty as a Strategic Doctrine

Countries are reshaping their internet policies around the concept of digital sovereignty—the right to control the hardware, traffic, and services that traverse national networks. It's no longer about content moderation or data privacy. It extends deep into the physical layer of the internet stack—who builds the cables, where they land, and who monitors their endpoints.

This approach has direct implications for AI, where latency-sensitive applications—such as autonomous systems, algorithmic trading, and military-grade simulations—depend on unimpeded, secure data corridors. Nations now negotiate not only bandwidth, but geopolitical leverage, cable landing rights, and long-term strategic autonomy in the AI economy.

Data Sovereignty and AI: Who Controls the Flow?

Data Residency Laws Are Redrawing the Map

Countries enforcing stricter data residency requirements are reshaping how digital infrastructure is designed. Nations like India, Russia, and China now demand that personal and sensitive data be stored and processed within national borders. These policies, originally targeted at privacy and national security, directly influence how subsea cables are routed and utilized. Data that once flowed freely from region to region now faces segmented pathways, effectively creating digital borders in the subsea network.

This regulatory shift isn’t abstract—it's forcing concrete investment realignments. Meta, for instance, has ramped up plans for region-specific data centers in Singapore and Denmark to maintain compliance with local storage mandates. The monetary cost is high, but the penalty for non-compliance—fines, restrictions, or even expulsion—is higher.

Big Tech’s Strategy: Build Redundancy. Everywhere.

To sidestep single points of failure and maintain high AI performance standards, hyperscalers like Google and Microsoft are embedding regional redundancy into their subsea investments. This means building multiple, often overlapping, cable routes complemented by nearby data centers. It's less about having a single fast lane and more about developing an interconnected mesh of highways that can reroute traffic as compliance or latency demands shift.

This redundancy is not only a response to legal intricacies—it's a hedge against geopolitical instability, natural disasters, and the escalating complexity of AI performance tuning across diverse regions.

The Border Problem: AI Doesn’t Travel Well

Unlike generic cloud computing, AI workloads require massive datasets, high-performance compute clusters, and continuous retraining cycles. Pushing this type of data across borders is logarithmically more difficult when data residency or export restrictions apply. Large language models, for instance, need context-rich regional data to fine-tune responses. But when that data is locked behind borders, training has to happen in-country—often at a higher operational cost and with lower resource utilization.

Latency-sensitive applications like autonomous driving simulations, real-time fraud detection, or multilingual recommendation engines suffer when constrained by fragmented infrastructure. This slows down deployment timelines and increases cost, especially when the model training and inference must comply simultaneously with GDPR in the EU and CCPA in California.

Compliance vs. Commercial Logic

There's an ongoing tension between regulatory compliance and the optimization goals of Big Tech. Commercial logic dictates centralizing AI training in hyperscale clusters located where energy is cheap and compute is fast—typically the U.S., Scandinavia, or parts of Canada. Compliance, however, demands that training datasets stay within national boundaries, leading to duplicated expenditures and artificial inefficiencies.

In response, companies are engaging in regulatory arbitrage—routing data first through compliant jurisdictions before transferring—with encryption and processing mechanisms broken into stages. It's more expensive, and slower, but it achieves legal defensibility.

As the regulatory landscape continues to splinter, expect AI infrastructure strategies to increasingly mirror the regulatory borders of the nation-states they touch. Undersea cables, once seen as neutral veins of global commerce, now act as contested territory in a war over who gets to control the flow of data-driven intelligence.

The Future of AI-Speed Internet Lies Underwater

AI has stopped being a buzzword and started dictating how the world’s digital infrastructure is built. Every new model, every neural net, every multimodal system with billions of parameters demands more compute—paired with faster, more reliable global connectivity. That demand is rewriting how the internet is wired at the ocean floor.

Subsea cable construction, once a domain dominated by telecom consortiums and governments, now runs at the pace of Big Tech vision. Amazon, Google, Meta, and Microsoft no longer wait for partners—they own, fund, and operate their own ocean-spanning highways of data. This shift has created a duopoly of roles: these companies are both the stewards of global connectivity and fierce competitors racing for AI dominance.

As these firms extend their control under the sea, the next frontier is emerging: intelligence-driven infrastructure. Expect AI not just to ride on the network, but to actively route traffic through it. Real-time optimization of fiber-optic pathways, automated rerouting during outages, and predictive workload balancing across continents—these capabilities are moving out of the lab and into the cable landing stations.

Want a glimpse of the future? Look past the data centers and toward the ocean floor. There, commercial ambition meets public interest. Infrastructure strategies intersect with global policy. And every meter of fiber glass laid brings us closer to a planet stitched together by AI-ready bandwidth.