Chinese state-linked ownership across eleven Mediterranean container terminals, assessed against NATO logistical exposure. Forty pages, completed July 2026.
To what extent do Chinese commercial investments in Mediterranean port infrastructure create strategic dependencies relevant to NATO logistics?
It is a question usually answered in one of two unsatisfying ways. Either Chinese equity is treated as equivalent to Chinese control, which produces alarm the corporate structures do not support. Or the absence of naval port calls is treated as the absence of a problem, which ignores what commercial concentration does on its own.
High confidence. There is material commercial concentration within a specific corporate perimeter, dominated by COSCO Shipping Ports, and three positions carry high observational utility as a function of their location.
Moderate confidence. That concentration has dependency implications for European logistical resilience.
Not supported. Publicly available evidence does not substantiate claims of routine PLA or PLAN operational use of any position in the network. The commercial dependency argument is far better evidenced than the direct military use argument, and the report says so.
Eleven positions of analytical significance, across three ownership tiers, five Chinese entities and seven jurisdictions.
| Tier | Positions | Count |
|---|---|---|
| Tier 1 Majority ownership, direct board and operational authority |
Piraeus, Valencia, Bilbao, Kumport | 4 |
| Tier 2 Significant influence, board seats and veto rights without control |
Vado Ligure, Suez Canal Container Terminal, Fos-sur-Mer, Malta Freeport, Thessaloniki | 5 |
| Tier 3 Operational concession without controlling equity |
Haifa Bay, now a mixed Tier 2 and 3 configuration after the July 2025 partial divestment | 1 |
| Separate category Hong Kong listed private operator, not SASAC controlled |
Barcelona (BEST), held by Hutchison Port Holdings | 1 |
The count rises to fourteen if the three Atlantic-coast French terminals in the same China Merchants portfolio are included. They are part of the same corporate network but excluded on geographical grounds. Published figures for this network vary between roughly ten and fifteen positions, and the variation is almost entirely a function of inclusion rules like these, which is why the report states its own.
The four SASAC-controlled entities are ultimately subject to central government oversight but maintain distinct corporate boards, separate reporting lines and competing commercial priorities. Hutchison, Hong Kong listed, has a qualitatively different relationship with Beijing. Treating "Chinese-linked" as an undifferentiated category overstates the coherence of the network.
APM Terminals holds majority operational stakes alongside COSCO at Vado Ligure and at Port Said. CMA CGM holds significant minority stakes at Valencia and Bilbao. The July 2025 sale of a 25 per cent interest in the Haifa Bay terminal to Israel's Noy Fund extends the pattern into the concession model.
The consequence is direct. Any assessment using Chinese equity share as a proxy for operational control systematically overstates the extent to which Chinese state direction can be exercised at terminal level.
The Naples exit in 2016, the failed engagement at Trieste, Italy's withdrawal from the Belt and Road Initiative in December 2023 and the partial Haifa divestment in July 2025 form a pattern of contraction and unrealised engagement. There have been no new Tier 1 acquisitions since October 2021.
Whether that is a stable plateau or an early phase of contraction cannot be established from the present evidence, and the report grades that interpretation low rather than choosing the more interesting reading.
The 2024 Red Sea disruption is the natural experiment. Piraeus throughput fell 7.8 per cent as Asia-Europe flows diverted. Over the same period throughput rose 16.8 per cent at COSCO's Spanish terminals and 55.3 per cent at CSP Zeebrugge.
The rerouted traffic never left the COSCO network. Dependence should therefore be assessed across the corporate portfolio rather than at any single terminal, because network-level flexibility keeps cargo inside the same perimeter even as individual port dependence softens.
This is the distinction the report rests on, and collapsing the two into a single "military utility" rating misstates every case where they diverge.
| Dimension | What it measures | Finding across the network |
|---|---|---|
| Operational | Capacity to move military materiel, host military vessels or supply state forces | Low or low to moderate at every case. No case rated high. |
| Observational | Structural access to maritime domain data, vessel identification and transit patterns, by virtue of location | Low at most cases. High at three: Kumport on the Bosphorus southern approach, Port Said at the northern entrance to Suez, and Haifa Bay, co-located with an Israeli naval facility. |
Those three are also the cases most frequently invoked in secondary reporting, usually without the distinction being drawn. Note the separate grading: confidence that the observational potential exists is high, because it is a structural feature of geography. Confidence that it has been operationalised into intelligence flow to Chinese state entities is low, and that gap is stated rather than bridged.
Each case carries three separate confidence ratings: commercial influence, military utility, and evidence of PLA operational use. They are not collapsed into one, because the evidence supporting each is qualitatively different. Piraeus, for instance, grades high, moderate and low respectively. A single overall rating would have concealed exactly the thing worth knowing.
The findings are set against the alternative explanations that would have to hold for them to be wrong, rather than presented as the only available reading. Ownership tiers, throughput figures and shareholder structures are taken from primary corporate filings, principally HKEX disclosures and operator releases, with the source recorded per row.
It does not show military use. It does not establish what data, if any, flows from these terminals to Chinese state entities. It cannot resolve whether the network has plateaued or begun to contract. Each of those is recorded as a limit on the finding rather than filled with the more publishable inference.
Two positions where beneficial ownership could not be established from open sources are recorded as intelligence gaps. Those two are the ones I would want a reader to press me on, which is why they are here rather than buried in an appendix.
The study is unpublished. It is available in full on request, and I am happy to talk through the method with anyone assessing it.