Part 1 | Part 2
To assess the environmental impact of a system, its entire lifecycle must be considered. While the first part of this article explored optimisation levers during the operational phase, particularly in terms of energy efficiency, it only addresses part of the equation.
The overall footprint of a supercomputer is also shaped upstream, through component manufacturing and supply chain organisation, and downstream, through the end-of-life management of equipment.
Raw materials: an environmental and ethical challenge for HPC
In the longer term, the availability of certain rare and precious metals used in the electronics industry such as cobalt, nickel and lithium is becoming an increasing concern for supercomputer manufacturing.
These materials, which are essential for component production, may become scarcer, prompting manufacturers to rethink their technological choices. Efforts are therefore underway to reduce the use of so-called 3TG metals (gold, tin, tantalum and tungsten), by optimising hardware architectures, exploring alternative technologies and developing next-generation materials to limit dependence on critical resources.
This approach also involves increased vigilance regarding the origin of materials. Supplier selection plays a key role in ensuring that both environmental and ethical standards are met across the entire supply chain.
The term “3TG” refers to tantalum, tin, tungsten and gold, often described as “conflict minerals”. Their extraction is sometimes linked to armed conflicts and human rights violations, particularly in Central Africa. In response to these ethical concerns, international regulations have been introduced to promote responsible supply chains.
Under European regulations, companies involved in the extraction and trade of these metals are required to carry out due diligence to identify and manage risks within their supply chains, ensuring that their activities do not contribute to fuelling conflicts.
At the same time, alignment with international regulations and standards remains essential. Compliance with directives such as the European Union’s RoHS (Restriction of Hazardous Substances), which strictly regulates the use of dangerous substances in electrical and electronic equipment, helps reduce environmental impact while improving user safety.
The European Union’s RoHS Directive restricts the use of certain hazardous substances in electrical and electronic equipment. Its aim is to protect human health and the environment by limiting the use of harmful materials such as lead, mercury and cadmium, as well as certain flame retardants and plasticisers.
Manufacturers must comply with RoHS requirements to market their products in Europe, encouraging the design and production of safer and more environmentally friendly electronic equipment.
Together, these levers — reducing the use of critical metals, strengthening sourcing requirements and complying with environmental standards — contribute to building a more sustainable path for high-performance computing technologies, by integrating resource-related challenges from the design phase onwards.
European sovereignty: reshoring production as a geopolitical and environmental priority
The production of critical components for supercomputers, particularly circuit boards, now lies at the heart of a European sovereignty strategy. In a context of intense global technological competition, having design, manufacturing and maintenance capabilities for high-performance computing within Europe is essential to secure both civilian and strategic uses and to limit geopolitical dependencies.
As part of this approach, Bull has initiated the progressive reshoring of certain production capacities. For the BullSequana XH3500 system, nearly three-quarters of components are now produced in Europe. Board manufacturing, previously carried out in India, has been relocated to several European sites in France, Romania and Ireland. This evolution strengthens control over supply flows, improves component traceability and ensures regulatory compliance, while reducing the environmental impact of logistics chains: CO₂ emissions associated with material transport decreased by 50% in 2025 compared with 2019.
This strategy is also reflected in major flagship projects such as Alice Recoque, which integrates key European technologies, including the BXI v3 interconnect developed by Bull and, in the longer term, the European SiPearl Rhea2 processor. It is further supported by strengthened industrial investment, with the ambition to double the production capacity of Bull’s supercomputer manufacturing site in Angers. Altogether, this approach aligns with European climate objectives and contributes to a model of digital sovereignty built on high-performance, secure and sustainable computing technologies.
Anticipating end-of-life: tonnes of metals to recover
Another major environmental challenge for supercomputers lies in their end-of-life management. Although these systems can operate for up to 10 to 12 years, they are typically renewed after five years to maintain optimal performance and continue benefiting from the most advanced technologies. Yet, a supercomputer spans large physical infrastructures and comprises several dozen racks, each containing more than a tonne of metals — representing a significant volume of materials to process during renewal.
At end-of-life, equipment is generally taken back by the manufacturer and directed towards specialist recycling channels for the recovery of metals and electronic components. Despite progress in this area, these processes remain complex due to the interconnection of materials, often requiring demanding manual or chemical treatments. Acting upstream, from the design phase, to facilitate dismantling and recycling is therefore essential.
Reuse also offers a valuable alternative: some systems can be refurbished and redeployed as smaller clusters. At Bull, this approach makes it possible to give supercomputers a second life, particularly in Africa and Eastern Europe, extending their use while reducing their environmental impact.
So, can supercomputers really be eco‑responsible?
While it would be unrealistic to speak of complete neutrality for infrastructures that are inherently resource-intensive, the levers explored throughout this article demonstrate that a supercomputer can now be designed in a significantly more responsible way.
Energy efficiency, heat reuse, controlled water consumption, end-of-life anticipation, material choices and industrial reshoring: eco-responsibility does not rely on a single technology, but on a comprehensive approach embedded from the design phase.
The challenge is therefore not to oppose performance and responsibility, but to advance them together. By addressing the entire lifecycle, from manufacturing and operation to reuse and recycling, it becomes possible to sustainably reduce the environmental footprint of high-performance computing, while meeting the growing demand for computing power. In this sense, the eco‑responsible supercomputer is not an abstract objective: it is an industrial and technological trajectory that is already underway.
