Sovereignty Research

This note updates and refines the analysis in Starlink vs Starshield: Platform Sovereignty and the Collapse of Neutral Infrastructure. The original thesis is substantially supported. Three corrections are required.


Summary of corrections

The layered-separation model is confirmed by new evidence. But three claims in the original paper require revision:

  1. The Ukrainian service was not a separate constellation or classified waveform
  2. Authority was reallocated, not completely transferred
  3. Ukraine did not cause Starshield’s creation

What is now confirmed

In August 2024, the Pentagon added 2,500 existing Starlink terminals in Ukraine to a government-contracted service described as “Starlink over Starshield.” Combined with 500 terminals covered by an earlier arrangement, this brought the total to 3,000 terminals under two U.S. contracts. The stated purpose was internet connectivity in Ukraine, with service extending into 2025.

A subsequent direct clarification from U.S. Space Systems Command materially qualified the initial reporting:

This was not a militarized version of the Starlink satellites.

The command described it as a negotiated service plan “aligned for military use.” It confirmed the service did not use a classified signal and that the terminals employed the same encryption available commercially.

The best-supported technical description is therefore:

Existing Starlink terminals + existing Starlink constellation + a differentiated government service plan and contractual governance layer.

That is strong evidence for the layered model. The distinction existed primarily in the service, entitlement, procurement and control planes, not through a clean duplication of physical infrastructure.


Evidence table

Proposed claimEvidence-based assessment
~3,000 terminals obtained Starshield-related serviceConfirmed for 2024–25
Ukraine used a separate Starshield constellationNot supported for this contract
Access was differentiated at terminal/service levelStrongly supported
The state governed access contractuallySupported, principally through the U.S. government
Control was transferred completely from SpaceX to the stateNot demonstrated
Ukraine caused Starshield to emergeIncorrect as a literal historical claim
The 3,000-terminal arrangement remains active in June 2026Not publicly confirmed

Scale context

The 3,000 terminals were a fraction of Ukraine’s Starlink estate. By early 2025, Ukraine had approximately 47,000 terminals, with most funded by Poland, Germany and other donors rather than the United States. The Starshield-linked agreements represented a military-governed subset within a much larger, heterogeneous Starlink environment — not a wholesale migration into Starshield.

Current continuity is uncertain. Reuters reported on June 2, 2026 that Ukraine’s Defence Ministry did not answer whether it presently uses Starshield.


The strongest new evidence: the 2026 whitelist

In February 2026, Ukraine introduced a nationwide Starlink authorization regime in coordination with SpaceX. Only terminals registered on a government whitelist could operate. Verification used device identifiers (KIT and UTID numbers), and mobility restrictions were applied to authorized devices. Ukraine subsequently reported that approved Ukrainian terminals remained operational while Russian-used terminals were blocked.

This is an exceptionally clear example of the architecture the original paper describes:

shared constellation → shared terminal technology → differentiated identity → state-defined authorization → policy enforcement by the platform operator

It provides more direct evidence of terminal-level governance than the partially undisclosed Starshield contracts do. It also shows that differentiation can include not merely access, but permitted mobility and operating conditions.


The August 2025 procurement signal

A U.S. approval of a possible $150 million extension of satellite communications service for Ukraine’s Starlink terminals provides additional evidence of institutionalization through government procurement. The official notice called the service Starlink, not Starshield — illustrating why product names should not be treated as clean technical or governance boundaries.


What should be reformulated

1. Replace “authority transfer” with “authority reallocation”

The evidence does not show that SpaceX surrendered final technical control or that Ukraine obtained sovereign command over the service. Authority became distributed among:

  • SpaceX as technical operator
  • U.S. agencies as contracting and sponsoring authorities
  • Ukraine as battlefield operator and, under the 2026 whitelist, device-authorizing state

Contracts proceduralized decisions and reduced reliance on informal personal discretion, but they did not eliminate corporate control or geopolitical dependency.

A particularly important implication: contractualization may transfer authority to the sponsor state, not necessarily to the dependent user state. The result may be more institutional, yet still leave Ukraine exposed to upstream American decisions.

2. Do not say that Ukraine created Starshield

Starshield’s national-security development predates Russia’s full-scale invasion. The National Reconnaissance Office awarded SpaceX a $1.8 billion contract in 2021 to build a separate intelligence constellation, with prototypes already under development before 2022.

The defensible causal claim:

Ukraine did not originate Starshield, but it exposed and accelerated the need to apply a government-governed service layer to Starlink communications in active warfare.

3. Qualify “not a separate constellation”

For the Ukrainian “Starlink over Starshield” service, that statement is supported. It should not be generalized to all of Starshield.

Starshield is an umbrella for several government capabilities. Military-specific Starshield terminals may connect both to commercial Starlink satellites and to a separate, more secure Starshield constellation. The NRO surveillance architecture is itself a distinct satellite network.


Revised formulation

Public reporting now confirms that, under two U.S. government contracts, 3,000 existing Starlink terminals in Ukraine received a military-aligned service known as “Starlink over Starshield.” This was not a separate Ukrainian constellation or a publicly confirmed classified waveform. Space Systems Command described it as a negotiated service plan operating through standard Starlink terminals and satellites.

The significance is institutional rather than merely technical. A shared physical system was divided into differentiated service and governance regimes through contracts, terminal entitlements and operational policies.

Ukraine’s subsequent nationwide Starlink whitelist makes this pattern clearer still: access to the same constellation can be governed at the level of terminal identity, authorization and permitted behavior.


The revised thesis

The evidence strongly supports this refined proposition:

Institutional separation does not require complete physical separation. It can be implemented through contracts, service entitlements, terminal identity, access policies and operational control mechanisms layered over shared infrastructure.

The transition is best described as:

private discretion → governance conflict → government contracting → differentiated service regimes → terminal-level authorization

Authority was not transferred completely; it was contractualized and redistributed among SpaceX, the U.S. government and Ukraine.


Implications for AI and cloud infrastructure

For AI and cloud infrastructure, this supports a credible hypothesis: indispensable systems are likely to remain technically shared while access, policy, assurance and decision rights become institutionally segmented.

The Starlink case demonstrates the mechanism. It does not by itself prove that every dual-use system will follow it.

You do not need two constellations to produce two governance regimes. But differentiated contracts do not automatically produce sovereign control: they create a layered allocation of authority across the operator, sponsor state and operational user.