The Submarine Cable in the Sky: When the Line Is Cut, Who Keeps Taiwan Connected to the World

Low-earth-orbit satellites fly at an altitude of roughly 600 km — close to Earth, with low latency — but the drawback is that each satellite covers only a small area, so it takes "laying down a great many of them" to weave a complete network ^[1][9]. This is exactly Starlink's approach: weaving a globe-covering network out of tens of thousands of small satellites.

How outsized is its scale? As of June 2026, Starlink has more than 10,000 satellites in orbit (roughly 10,413 launched, 10,397 operational), covering more than a hundred countries; the U.S. FCC has already approved the first batch of 7,500 satellites for its second-generation constellation, with total filings running as high as 42,000 ^[9][13]. By comparison, Europe's "runner-up," Eutelsat OneWeb, has a constellation of only 648 satellites ^[9].

This gap in numbers spells out one harsh fact up front: when it comes to "global-class broadband," Starlink is in a class of its own, leading by a cliff-edge margin. Any plan that sets "building a constellation to replace Starlink" as its goal is unrealistic in the near term, given the cost, timeline, and launch capacity involved (confidence: high). Understanding this premise is what makes it possible to see what Taiwan is actually doing — and why it did not choose the biggest option.

The centerpiece of Taiwan's own LEO satellite program is the Taiwan Space Agency's (TASA) Beyond 5G (B5G) LEO communications experimental satellite. Frankly, its scale is not in the same league as Starlink's: it currently consists of two high-performance experimental satellites (designated 1A and 1B), deployed in an orbit of roughly 600 km at an inclination of 44 to 50 degrees, with the goal of "validating Taiwan's self-developed communications payload and ground communications equipment" ^[1].

Pay attention to the words "experimental" and "validating" — they are the key to understanding this whole story.

On timing, the first B5G satellite was originally planned for a series of launches starting in 2026, but that was pushed back to 2027, and the second satellite (1B) will not arrive until around 2030 ^[2]. The delay is a fact (confidence: high), and it is not the first time this program has slipped. The division of payload work is also an honest reflection of where Taiwan's capabilities stand: 1A's communications payload went to international cooperation, awarded to the U.S. firm CesiumAstro in April 2025; 1B is being built domestically, to demonstrate Taiwan's own aerospace capability ^[2][3]. TASA's stated goal is for roughly 80% of the B5G satellite's components to come from Taiwanese suppliers — but that is a program target, not an already-achieved reality (confidence: medium) ^[3].

Zooming out, Taiwan's "Third Phase Long-Term Space Technology Program" commits a total of NT$25.1 billion from 2019 to 2028, but its main thrust is actually about ten remote-sensing satellites — high-resolution optical imaging, ultra-high-resolution optics, and synthetic aperture radar (SAR) — while the communications constellation is only just getting started ^[4].

So when someone describes Taiwan's own LEO program as "Taiwan's version of Starlink," that claim needs a discount applied: at this stage it is "validation and training," not yet a resilience mainstay capable of carrying full bandwidth in wartime. Misplacing its value will only leave both expectations unmet. Its real significance only becomes clear from another angle.

If Starlink is this strong, why doesn't Taiwan just use it directly?

There are two layers to the sticking point. The first is regulatory: Taiwan's Telecommunications Management Act requires LEO satellite service providers to register as telecommunications enterprises, requires the chairperson to hold Republic of China nationality, and caps direct foreign ownership at 49% ^[5]. SpaceX's global practice, meanwhile, is to set up 100%-owned subsidiaries; the two sides have been deadlocked on this red line for five years. SpaceX has stated publicly that it is not operating in Taiwan because the competent authority will not allow foreign ownership above 50%, saying "SpaceX has not accepted this condition at any of its operating locations," while stressing the issue has nothing to do with Russia or China ^[6].

The second layer is the real core issue — trust. Taiwanese officials have pointed out more than once that Musk has enormous commercial interests in China, has called Taiwan "an inalienable part of China," suggested handling it like a special administrative region, and compared Taiwan to Hawaii ^[6]. A Ministry of Digital Affairs official put it more bluntly: "If China can influence Musk, in an emergency he could simply turn off Taiwan's satellite receivers." ^[6]

This is not paranoia. Go back to that night in the Black Sea: the drone boats lost contact and ran aground not because the enemy was too strong, but because the switch for the service sat in the hands of a private individual with enormous interests in Beijing ^[7]. Once you understand Ukraine's experience, Taiwan's wariness about that 49% line stops looking like nationalist purism and starts looking like rational risk pricing.

This is a point that needs to be honestly separated, to avoid self-deception: whose rocket you launch on, and who holds the switch for the service, are two different questions. Taiwan's own domestically built satellites still have to launch on SpaceX rockets — Hon Hai's second-generation "Pearl" satellites lifted off on a Falcon 9 on May 3, 2026 ^[12]. Dependence on launch capacity is a real issue, but it is "one-time"; whether communications service can be switched off remotely in wartime is the "ongoing" matter of life and death. It is the latter that Taiwan should prioritize holding onto.

Having decided not to bet on Starlink, Taiwan chose a different path: in November 2023, Chunghwa Telecom signed an exclusive agency agreement with Eutelsat OneWeb for Taiwan, and after more than six months of review, obtained Taiwan's first OneWeb commercial license in June 2025 ^[8].

Why OneWeb? The key is not superior technology — its entire constellation has only 648 satellites, with far less coverage density and bandwidth than Starlink ^[9]. The key lies in its business model: OneWeb operates through a "telecom-operator agency" structure, willing to accept ownership and operation by a local telecom carrier, which lines up neatly with Taiwan's sovereignty preferences and regulatory requirements ^[8]. Taiwan would rather have a line with narrower bandwidth but a switch held by its own people, than the widest line available that could be switched off remotely by someone else.

The cost is also real: OneWeb's coverage of Taiwan currently relies on relaying signals through three nearby ground stations in Japan, Guam, and Thailand, and building a single ground station costs roughly US$15 million (not including operations and maintenance) ^[8]; the initial service targets government and enterprise users, positioned as a "data network backup" rather than a mass consumer service ^[8]. It can weave together a "good-enough safety net," but if treated as the sole solution in wartime, its bandwidth would likely fall short.

So the real answer emerges: it is not a binary choice between "build it yourself" and "buy it," but multiple sources, multiple layers — Taiwan's own B5G program carries wartime core backup and supply-chain training, the OneWeb agency deal supplies everyday resilience for enterprise and government, and the door is kept open, under the right conditions, to add more constellations in the future (even Starlink, if conditions change). This is exactly an extension of the "sea, land, and sky" logic in our submarine cable report: submarine cables, mobile networks, microwave links, and LEO satellites all back each other up in layers. LEO satellites are not meant to carry the entire network alone — they are meant to hold up the single most critical line of communication if the submarine cables are ever cut entirely.

Having talked this much about sovereignty, it is easy to overlook a more down-to-earth question: what can Taiwan actually earn out of this space race?

The answer is not up in the sky, but on the ground. Taiwan's supply chain is toughest at electronics manufacturing and high-frequency communications — and getting a satellite network to actually work requires large quantities of ground equipment, antennas, and RF components. This is exactly the sweet spot where Taiwanese manufacturers have staked their claim.

The most representative case is Universal Microwave Technology: it makes high-frequency microwave and millimeter-wave components (filters, antenna modules, power amplifiers, RF subsystems), and has already broken into the supply chains of all four major LEO constellations — SpaceX, Amazon (Kuiper), Eutelsat OneWeb, and Telesat ^[10]. In 2025 its full-year revenue hit a record NT$2.452 billion, with LEO satellite products climbing to nearly 70% of the fourth quarter's share; undelivered orders on hand exceed NT$1.8 billion, and analysts even estimate 2026 revenue could grow a further roughly 90% (this is an analyst forecast, not an accomplished fact — flagged as inference) ^[10].

But on the same track, the scenery is worlds apart. Microelectronics Technology, another stock likewise labeled a "LEO satellite concept play," posted a 2025 after-tax net loss of NT$1.146 billion, a loss of NT$4.55 per share, with net worth below half of paid-in capital, and was moved into full-cash-delivery trading starting March 2026 ^[11]. One company hitting a record high, the other pushed into full-cash-delivery trading — the contrast is a splash of cold water, reminding everyone: a "concept" is not the same as "profit," and this industry does carry out real culling.

Hon Hai (Foxconn) has taken yet another path: from its first-generation "Pearl" CubeSat launched in 2023 to the second-generation PEARL-1A/1B validating a Ka-band inter-satellite link (ISL) in May 2026, its focus is on satellite "assembly, integration, and test" (AIT) ^[12]. Taiwan's actual position is therefore clear: strong in ground equipment and high-frequency components, weak in system integration and the payload core. Recognizing this is what tells the industry where to put its effort — and that, in turn, lines up neatly with the sovereignty goal of "building up a self-reliant supply chain."

For the state: govern LEO satellites as national-security-grade critical infrastructure, with the core principle that "control of the service switch must remain controllable." Prioritize resources on self-developed payload validation (B5G, the wartime core backup) and on a supply chain that can actually be monetized, rather than chasing an unrealistic goal of building everything in-house. External procurement should follow "multiple sources, multiple layers": the OneWeb agency deal is the current mainstay, but Taiwan should avoid letting a single substitute foreign vendor become a new single point of dependence, and should keep the door open, under the right conditions, to combining multiple constellations (including Starlink, should conditions change in the future). At the same time, treat dependence on launch capacity (the launch side) and control over the service (the operating side) as separate layers, so a one-time risk is not mistaken for an ongoing one.

For industry intermediaries (telecom/satellite/equipment): Taiwan's opportunity lies in ground equipment and high-frequency components, not in competing with Starlink on constellation scale. Put resources behind the strengths already validated by all four major constellations — such as high-frequency RF, antennas, and ground-station integration — and help the government build out a multi-source backup architecture. But conduct honest technical and financial due diligence: "LEO concept" stocks are everywhere, and only the companies that can actually deliver and actually turn a profit are worth betting on; the contrast between Microelectronics Technology and Universal Microwave Technology is a warning bell.

For SMEs: businesses that depend on the cloud, cross-border payment flows, and remote collaboration should treat "communications could be cut" as a basic operating assumption, not a low-probability accident. Prepare a satellite or offline contingency plan in advance, and confirm that your telecom or cloud provider has multi-path routing and satellite backup; amid the panic and information warfare that would accompany an outage, don't let a single technical incident be amplified by rumor into a collapse of confidence in Taiwan's resilience — this, too, is a lesson carried over from our submarine cable report.

The threats Taiwan spends the most effort guarding against are missiles in the sky and warships at sea; but what is most likely to leave the world unable to reach Taiwan first is those few silent lines on the seabed — and that line in the sky which Taiwan does not yet fully hold in its own hands.

On the surface, the story of LEO satellites is a race over satellite counts and bandwidth; underneath, it is a much older question: at the most critical moment, who holds the switch that decides whether you can speak? Ukraine's drone boats gave a brutal answer. Taiwan's response is neither to spitefully refuse the strongest player out of pride, nor to naively believe it can single-handedly weave a network of tens of thousands of satellites; instead, it is this: weave the single most critical stretch of line yourself, rent a sufficient stretch from a trustworthy partner, and never hand every line over to the same person. Sovereignty was never about owning everything — it is about this: even if you are cut, the line still holds; and even if the line is cut, the switch remains in your own hands.