Recent signals from the AI supply chain, including Nvidia's explosive earnings outlook and South Korea's triple-digit surge in semiconductor exports, have fundamentally validated the bull case for AI computing power: global demand is still in an acceleration phase, far from peaking.

According to analysts at Deutsche Bank, the surging demand for AI compute is now rapidly extending beyond GPUs, HBM, and servers into network switching chips, optical modules, subsea fiber optics, and low-Earth-orbit satellite constellations. Nvidia's fiscal Q2 2027 revenue hit $96.2 billion, up 106% year-over-year, with data center revenue reaching $89 billion, up 117%. The company's Q3 revenue guidance was raised further to $108 billion, plus or minus 2%. This indicates that the industry bottleneck is shifting from pure GPU supply to high-speed interconnect, cross-regional data transmission, power, and data center cluster coordination. Without sufficient network throughput and transmission resilience, even the most powerful accelerators risk becoming isolated "compute islands" that cannot efficiently work together.

The robust demand for AI compute clusters is now manifesting in longer-term, larger-scale infrastructure contracts. Reports show that Anthropic, backed by Nvidia, signed a $35 billion cloud computing agreement with Lambda involving approximately 350 megawatts of capacity. Shortly before, Anthropic committed to paying Nscale $45 billion over six years to lease around 460 megawatts of AI compute powered by Nvidia's Vera Rubin chips. Combined, these two deals total roughly $80 billion, highlighting that leading model companies are transforming from on-demand cloud service buyers into long-term infrastructure tenants that pre-secure chips, power equipment, data center operations, and network capacity.

Deutsche Bank argues that the competition for AI infrastructure has evolved from "who owns the most core computing chips" to "who controls the physical channels of global data flow, network infrastructure architecture, and critical data nodes." Submarine cables and large-scale satellite platforms have therefore simultaneously become scarce assets, digital sovereignty tools, and the new frontier of geopolitical maneuvering. The endgame of AI is not merely compute clusters and power infrastructure—it extends to subsea cables and satellites.

Deutsche Bank warns that the "digital chokepoint"—namely submarine cable and satellite resources—is becoming the critical compute asset in the AI inference era. The 1.8 million kilometers of subsea cables and low-Earth-orbit constellations usher in the second main line of AI infrastructure investment. The analysts, led by Marion Laboure, stated in a report released Wednesday that in the era of "infrastructure realism," digital connectivity infrastructure has transformed into a primary battleground for great power competition. Submarine fiber optic cables and LEO satellite constellations, once viewed as neutral utilities supporting global communications, are now seen as geopolitical strategic assets that confer strategic leverage and resilience, while also introducing vulnerabilities. The race to dominate these digital networks has become a decisive geopolitical frontier, with sovereign states and private hyperscale cloud providers all vying for control over the physical systems underpinning the AI super-revolution.

Submarine cables are indispensable to the global economy, carrying over 99% of intercontinental data traffic through fiber channels laid across 1.8 million kilometers of seabed. An estimated $10 trillion in daily financial transactions flow through these underwater networks, including data from the SWIFT network connecting over 8,300 financial institutions across 195 countries. Despite this, the infrastructure remains highly fragile. Laboure's team noted that a single typical cable outage can cause an immediate 2 percentage point drop in the affected economy's per capita GDP growth rate. Over three to six years, the gap between the disrupted country and its unaffected neighbors widens further, with per capita GDP growth ending up 9 percentage points lower, as economic efficiency, innovation capacity, and labor productivity deteriorate significantly.

Repair challenges amplify these risks—there are only about 40 specialized cable repair ships globally, and the cost per incident ranges from $500,000 to $1 million. Geographic concentration further compounds the danger, as critical routes traverse multiple geopolitical hotspots, including the Red Sea corridor, which carries 17% to 25% of global internet traffic. Driven by AI's immense bandwidth demands, U.S. tech giants have fundamentally reshaped the ownership structure of global connectivity infrastructure. The largest internet content providers and cloud networks now account for nearly three-quarters of all international bandwidth demand, up from less than half in 2016. Of the estimated $14 billion in investment required for new subsea cables expected to come online between 2025 and 2027, hyperscale content providers and cloud giants like Alphabet (NASDAQ: GOOGL), Meta Platforms, Inc. (NASDAQ: META), and Amazon.com (NASDAQ: AMZN)—rather than traditional telecommunications operators—are dominating funding and technical design. This vertical integration allows hyperscalers to build private data corridors connecting their massive data centers while reducing reliance on legacy telecom providers. The European Centre for Development Policy Management has warned that the increasingly concentrated control by U.S. hyperscalers is creating structural dependencies that could undermine Europe's digital sovereignty.

Meanwhile, space has emerged as the newest arena for connectivity dominance. Deutsche Bank notes that Starlink, the high-speed satellite internet empire founded and led by Elon Musk's SpaceX, now accounts for 54% of all operational satellites in orbit, with over 10,800 satellites deployed and plans for a final constellation of 42,000. In response, China has launched two parallel strategic programs: the state-led Guowang constellation and the Shanghai-supported Qianfan constellation, which together plan to deploy 13,000 to 15,000 satellites. The Deutsche Bank report states: "In the era of infrastructure realism, development and growth power are not primarily determined by how many cables or satellites are built, but by control over the most connected nodes and architectures of large-scale network systems—from the seabed to low Earth orbit."

SpaceX stands out as the most direct, vertically integrated core beneficiary on the satellite communications side of Deutsche Bank's "digital infrastructure chokepoint" investment framework. The report emphasizes that true power comes not from the absolute number of cables or satellites, but from control over network nodes, transmission architecture, and access rules. SpaceX simultaneously controls reusable rockets, satellite manufacturing, launch capacity, the Starlink network, ground terminals, and network operations. With Starlink already claiming 54% of global orbital satellites, the company captures not just satellite hardware revenue but long-term "infrastructure rent" from global connectivity networks. Its strategic value is especially pronounced in scenarios involving subsea cable outages, warfare, natural disasters, and remote area coverage: Starlink can serve as a redundant communications layer beyond traditional fiber networks while offering high-value connectivity services to governments, militaries, enterprises, and mobile operators. As AI training and inference drive surging cross-data-center traffic, SpaceX could also integrate Starlink, terrestrial AI clouds, and future orbital compute centers to form a "launch-connect-compute" closed loop.

However, submarine fiber retains absolute advantages in capacity, per-bit cost, and stability, while LEO satellites primarily serve as supplemental, backup, and special-scenario connectivity. Concerns over digital sovereignty, spectrum control, and strategic dependencies on the part of various nations may also constrain Starlink's global expansion. On price targets, Deutsche Bank initiated coverage of SpaceX on July 7, 2026, with a "Buy" rating and a $255 target price. Following the Q2 earnings release, the bank reiterated "Buy" on August 5 but trimmed the target from $255 to $235, citing potentially significant increases in compute buildout and capital expenditures for 2027. Based on a share price of approximately $140.71 as of September 3, the $235 target still implies up to 65% upside for SpaceX. Bullish theses from Wall Street institutions like Goldman Sachs and Morgan Stanley center on: reusable rockets and Starship continuously lowering launch costs; Starlink establishing scaled recurring revenue across consumer, enterprise, and government markets; AI infrastructure and compute leasing emerging as a faster-growing new engine; and vertical integration of terrestrial and orbital AI compute enabling SpaceX to deliver connectivity and computing at lower cost.

Deutsche Bank believes the company's Q2 growth came in significantly ahead of expectations, driven by AI infrastructure and Starlink's enterprise and government business, with a path to a $100 billion annualized revenue run rate shortening. The target reduction, therefore, is not a reversal of the bullish thesis but a risk discount for at least 3 gigawatts of new compute capacity, massive capital expenditures, financing needs, and free cash flow pressure. Musk's social media post on X estimated SpaceX's annual revenue could reach approximately $3.5 trillion by around 2033—more aggressive than his previous "trillion dollars by 2030" target. Based on Wall Street's projected $100 billion revenue for 2027, this implies 35-fold growth over six years, a compound annual growth rate of roughly 81%. However, this aggressive projection remains Musk's personal estimate, not official company guidance.

SpaceX's prospectus explicitly defines "building a continuously expanding space civilization, ultimately progressing toward a Type II civilization capable of harnessing the Sun's total energy" as a long-term paradigm shift. Musk himself has stated that lunar satellite factories, mass accelerators, and AI hardware deployments exceeding 100 terawatts annually would propel humanity toward "non-trivial progress" toward Type II civilization. Thus, orbital AI data centers are not isolated projects but the first layer of infrastructure connecting "Earth-constrained compute—solar system-level energy—multi-planetary civilization." The so-called Kardashev Type II Civilization refers to a society capable of utilizing all of its host star's energy, while Type III would dominate the entire galaxy's energy output—far beyond SpaceX's current narrative.