Taiwan is an island of 35,000 km² located 180 km from the Chinese mainland coast. The People's Republic of China asserts sovereignty over it. A military conflict, a naval blockade, or even a major logistical disruption in the Taiwan Strait would interrupt the production of 90% of the world's advanced chips. The global digital economy — infrastructure, cloud, AI, telecoms — would face a component supply rupture within a matter of months. No buffer stock exists at this scale. No alternative production capacity is immediately available.

The SRE administers servers. Servers contain processors. Processors are made in Taiwan. A conflict in the Taiwan Strait does not translate into an abstract geopolitical event — it translates into the impossibility of ordering new servers in the 12 to 18 months that follow, the explosion of existing component prices, and the progressive degradation of infrastructure as hardware ages with no possibility of replacement. Digital sovereignty starts with the chips. Not with the cloud contracts.

Manufacturing is not the only bottleneck. Chip design software (EDA — Electronic Design Automation) is dominated by three American actors: Synopsys, Cadence, and Mentor Graphics (acquired by Siemens). The extreme ultraviolet (EUV) lithography machines essential for the finest nodes are produced by a single global actor: ASML, a Dutch company. Without ASML machines, TSMC cannot produce its most advanced chips. The dependency chain has multiple links, each as narrow as the last.

Europe is present in this chain at one significant point: ASML. This presence is real and strategic — without ASML, global advanced chip production stops. But it is also fragile: ASML itself depends on American suppliers for its lasers and subcomponents whose supply chain is not entirely European. Europe is one link in a chain it does not control end to end. STMicroelectronics, the only European chip manufacturer of scale, operates on less advanced nodes — useful for automotive and industrial applications, but not for AI datacentres.

The European Chips Act of 2023 mobilised €43 billion to attempt to bring Europe's share of global semiconductor production to 20% by 2030. This is a signal. It is not yet a capability. Intel has announced factories in Germany and Poland — projects that take 5 to 10 years to build and reach technical maturity. TSMC has announced a factory in Germany for previous-generation nodes. These investments will reduce dependence on certain segments — they will not recreate the leading edge of technology in Europe within the coming decade.

The US government has implemented strict controls on the export of advanced chips to China starting in 2022, progressively extended and tightened in 2023 and 2024. These controls demonstrate a fact that the Opération Dindon corpus has documented from a different angle: the hardware components of digital infrastructure are instruments of foreign policy. A US administration can, by executive order, decide which countries have access to which chips. This power requires no contractual negotiation — it is exercised by unilateral regulatory means.

Europe is not the target of current export controls. But it is not immune to a reconfiguration of American geopolitical priorities either. A trade partner can become a strategic competitor. Today's exemptions are not tomorrow's guarantees. A Europe that depends on chips whose access is controlled by a third-party government is a Europe whose digital sovereignty is conditional — conditional on the lasting goodwill of that third-party government.

The culture of short-cycle hardware renewal (3 to 5 years) is an industry norm that serves manufacturers' interests more than operators'. A properly maintained server can run for 8 to 10 years on stable workloads. Deliberately extending renewal cycles reduces dependence on the immediate availability of new components — and also reduces the exposure surface to supply disruption risk. This is FinOps applied to hardware, with a strategic dimension that standard FinOps analyses do not integrate.

The standardisation of datacentre architectures around a few processor families (Intel x86, AMD EPYC, NVIDIA GPU) creates dependence on converging supply chains. The rise of ARM architectures in datacentres — Ampere Computing, AWS Graviton — and open-source RISC-V processors represents genuine diversification. A heterogeneous hardware estate is more resilient to a specific supplier disruption than a homogeneous one.

Large organisations — states, telecoms operators, critical infrastructure companies — can integrate into their hardware planning the building of strategic stocks of critical components. This is not common practice in the IT industry, but it is standard in other sectors with critical dependencies (energy, pharmaceuticals, food). The awareness that semiconductors are a strategic resource on a par with oil or medicines has not yet translated into storage practice in the digital sector.

Every critical infrastructure risk analysis (business continuity plan, BCP, DRP) should include a "hardware dependence" component documenting the geographic origin of critical components, alternative sourcing lead times, and extended disruption scenarios. This practice is standard for software and contractual risks — it is absent for hardware risks. Adding it costs nothing and makes visible a dependence that is currently ignored.

The RISC-V instruction set architecture is open source — anyone can design and manufacture a RISC-V processor without paying a licence to Intel or ARM. European research and industrial actors are investing in RISC-V for embedded systems and accelerators. Supporting this ecosystem at the level of public procurement policies and research funding creates, over the long term, an alternative to proprietary architectures concentrated in the United States and the United Kingdom.

The Foundation of Iron begins at electricity and hardware (weeks 1 and 6) precisely because understanding the physical layer is the condition for understanding all the layers above. This pedagogy must be extended to the hardware supply chain: where components come from, who makes them, what the alternatives are, how to assess the resilience of a hardware architecture. An SRE who ignores the upstream chain of their infrastructure is an SRE who does not fully understand the risks they are managing.