On 14 October, without warning, Starlink’s military terminals in Ukraine stopped working on platforms travelling at speeds above 160 km/h. The threshold was no coincidence. Above that speed fly the interceptors Kyiv launches against Russia’s jet-powered drones derived from the Iranian Shahed, as well as Ukraine’s medium-range strike drones. On some stretches of the front, Zelensky wrote, connections were already dropping at speeds of between 130 and 160 km/h, while front-line logistics were also slowing down. The cut-off coincided with Washington’s pressure on Kyiv to stop striking Russian refineries, attacks that Trump blamed for driving up oil prices. According to Zelensky, US officials had also threatened to suspend intelligence-sharing and had presented the Starlink problems as technical faults.

For ten days, nobody could say who had made the decision. Trump denied responsibility: “It wasn’t me, and I don’t think it was Elon.” Marco Rubio maintained that the slowdown was not deliberate and that Starshield, the military version of the service, was not even available to Ukraine, while Ukrainian sources said it had in fact been in use for some time. The European Commission asked SpaceX to restore access “urgently”. On 24 October, Musk settled the matter in his own way, claiming that he had closed a “loophole” in the terms of service.

It is the clearest evidence yet of something Europe struggles to admit: Starlink is currently the world’s best satellite connectivity system, but it is not a reliable provider.

In critical infrastructure, whether payments, cloud services, networks or satellites, you do not buy bandwidth: you buy predictability. Whoever signs the contract is paying to know in advance how the service will behave on its worst day, not for its speed on its best. And the old adage holds even for the richest man in the world: either you do business or you do politics. Anyone who tries to do both ends up doing both badly.

Starlink and Ukraine: When the provider becomes an interested party

There was no shortage of precedents. In 2025, Reuters reported that Musk had ordered coverage to be cut in some areas during the Ukrainian counteroffensive in September 2022. SpaceX disputed that account. In March 2025, Musk wrote that without Starlink, the entire Ukrainian front line would collapse. Polish Foreign Minister Radosław Sikorski, whose government pays about $50 million a year for Ukraine’s terminals, replied that Warsaw would look for other suppliers if SpaceX proved to be an “unreliable provider”. Musk gave assurances that Starlink would never switch off its terminals and would never use them “as a bargaining chip”.

Nineteen months later, that promise appears to have been broken. The October restrictions hit precisely the Starshield terminals designed for the fastest drones, that is, those used for military purposes.

Accounts of what happened multiplied. Musk wrote that the restrictions on Starshield were set by the US military and, ultimately, by the president. According to the Wall Street Journal, Musk ordered the cut-off after discussing it with Trump and lifted it when Trump personally asked him to. Reuters reported that the Pentagon had received a message from the White House and that both the Biden and Trump administrations had repeatedly asked SpaceX to switch Starlink on or off in various countries.

On 24 October, with the service working again at speeds of over 400 km/h, Musk accused the Ukrainian military of using Starlink for offensive drone strikes in breach of the terms of service. The contractual loophole, he wrote, had been closed the previous week. Meanwhile, a bipartisan group of US senators asked the Pentagon’s inspector general to investigate. Democratic Representative Jason Crow called it “more than disturbing” that a private individual should wield such power.

For those buying the service, however, who made the decision is almost beside the point. The problem is that the customer did not know what was happening and had no way of knowing. A provider of critical infrastructure must offer rules that are known in advance, written commitments specifying what the service guarantees, and a clearly identified point of contact. Decisions should not be announced after the fact on a social network owned by the provider itself.

The Italian case: A contract that never materialised

Italy came close to signing a deal. In January 2025, Bloomberg reported talks over a five-year, €1.5 billion contract for secure government and defence communications. Palazzo Chigi, the Prime Minister’s office, denied that anything had been signed, but Defence Minister Guido Crosetto told the Chamber of Deputies that the Defence Ministry was duty-bound to explore the option: in low Earth orbit, where satellites circle a few hundred kilometres above the ground and provide fast connections, there were no alternatives.

On 15 October 2026, while the restrictions were still in force, Crosetto himself confirmed that the project had never been revived and that “it has not been discussed for months”, although he acknowledged that the system remained essential for Kyiv.

The problem is ours, but it is also Musk’s. We need the service, but we must ask ourselves whether we can accept a multi-year contract for encrypted communications with a provider that can adjust its service according to its own political judgement.

Telcos, spectrum and towers: Where the real game is played

On 19 October, Deutsche Telekom, Orange, Telefónica and Vodafone signed a letter of intent to explore a joint bid for European spectrum allocated to mobile satellite services (MSS), which connect phones and other mobile terminals directly to satellites. This is not yet a joint venture. The four operators would build the ground infrastructure in Europe without investing directly in a satellite constellation. The satellite partner has yet to be chosen, and existing agreements, including the one between Deutsche Telekom and Starlink, remain in force.

The prize is the 2 GHz band. The current licences, held by the US companies Viasat and EchoStar, expire in May 2027. SpaceX has made no secret of its desire to secure them, while the Commission intends to reserve part of the band for European players.

The same move has already been made in the United States. In May, AT&T, Verizon and T-Mobile announced a joint venture to pool spectrum earmarked for direct satellite-to-smartphone connections. AT&T signed the agreement on 14 September, according to its quarterly report to the SEC, and the company was formally established in early October. The three chief executives have stated that their goal is to have several constellations competing with one another, thereby avoiding dependence on a single provider.

Meanwhile, SpaceX is buying up, piece by piece, everything a telecommunications operator needs. It bought spectrum from EchoStar for $17 billion, and the FCC, the US communications regulator, has authorised 15,000 of its next-generation satellites for direct-to-phone connections. On 23 October, Bloomberg revealed that SpaceX was in talks to gain access to at least 80,000 sites in the United States, roughly half of the country’s 158,500 existing mobile tower sites.

Tower companies rose on the stock market: Crown Castle by 9%, American Tower by 4% and SBA by 3.6%, while the three operators lost ground. According to del Castillo, since 15 September, AT&T shares have fallen by 22%, T-Mobile’s by 19% and Verizon’s by 13%.

IRIS² at the crossroads between Airbus and FCAS

Europe’s first public response to Starlink, and a timid one at that, is called IRIS². In August 2026, the Commission and the SpaceRISE consortium, comprising SES, Eutelsat and Hispasat, signed the implementation agreement. The project has grown to 348 satellites, 66 of them dedicated to defence and security, with the first launches scheduled for 2029 and a total value of €15.6 billion. It began with an estimated cost of €6 billion and an entry-into-service date of 2024, later pushed back to 2030 or 2031. Starlink, meanwhile, has surpassed 10,000 satellites in orbit.

History offers a precedent: Airbus. Born out of a 1967 memorandum and established as a consortium in 1970, with Spain’s CASA joining in 1971, Airbus received public funding for decades. Its subsidies, together with those awarded to Boeing, fuelled one of the longest disputes in the history of the World Trade Organization, which was suspended only after the 2021 truce. Today, Airbus is the only global competitor to the American aircraft industry.

FCAS, the Franco-German next-generation fighter programme, offers the opposite example. In early June 2026, Germany and France shut the project down after nine years because Dassault insisted on being the aircraft’s “architect”, while Airbus refused to accept that arrangement. Paris is now going it alone, while Berlin is weighing whether to join GCAP, the fighter programme Italy is developing with the United Kingdom and Japan, or seek an agreement with Sweden and Spain.

What made the difference was not money but the chain of command. Airbus worked because governments accepted a single industrial lead and long time frames. FCAS foundered on the question of who was in charge.

On satellites, Europe risks repeating the same mistake by pursuing two parallel tracks: IRIS² for governments and the telecommunications operators’ initiative for the commercial market, with a satellite partner still to be chosen. If the two do not converge, their fate may already be sealed.

Either you do business or you do politics

Musk has tried to do both. He led DOGE, the body Trump created to cut federal spending, and left it in May 2025. He broke with Trump, announced plans for a party of his own before shelving them, made peace with him again and, according to Federal Election Commission data reported by Bloomberg, spent more than $445 million on the midterm elections. Musk himself has admitted that, if he could go back, he would have focused on his companies rather than on DOGE.

Even the company’s own filings spell out the risks. In the prospectus for its Nasdaq listing, dated 12 June 2026, SpaceX states that Musk “does not devote his full time and attention” to the company and that his actions and statements may affect its relationships with customers and regulators. The document also refers to the 2024 freeze on Starlink’s assets in Brazil. When a risk appears in the document a company itself hands to investors, it is no longer merely the opinion of its critics.

The consumer market is quick to forgive. Tesla lost almost half its European market share in 2025, partly because of the backlash against its founder’s political positions, but it is staging a strong recovery in 2026. States and armed forces, however, are not buying a car. They sign multi-year contracts, build operational doctrine around a service and cannot switch providers in a week. For them, reliability is not a feature of the product: it is the product.