Alps Opens a New Chapter in Swiss Supercomputing
How CSCS brought a next-generation research infrastructure into service for AI, medicine, weather forecasting and climate science.
Ā
This article examinesĀ the opening and operation of Alps, the supercomputer run by the Swiss National Supercomputing Centre (CSCS) in Lugano. It also explores a point that can be obscured by headline rankings: Alps is not simply one giant machine, but a configurable research infrastructure built to serve different scientific communities.
Ā
From installation to inauguration
The story began before the ribbon-cutting. CSCS installed the first cabinets in 2020, initially with CPU-based nodes. The system then expanded in stages, adding storage and new compute platforms. Between January and June 2024, CSCS installed 2,688 NVIDIA Grace Hopper nodes, bringing that part of the system to 10,752 Grace Hopper superchips. ETH Zurich officially inaugurated Alps in Lugano on 14 September 2024. The ceremony marked a public milestoneĀ for a facility that had been assembled and expanded over several years.1Ā 2
Ā
The central system is an HPE Cray EX. Each Grace Hopper node combines CPU and GPU resources, allowing a single node to handle computation and data movement for demanding workloads. CSCS lists a 200-gigabit-per-second Slingshot-11 network injection bandwidth per module or GPU, alongside large tiers of fast scratch storage and long-term tape archiving. These components matter together: a scientific job must move data to and from processors efficiently, not just perform arithmetic quickly.2
Ā
Performance figures need a date and a benchmark. At the initial expansion stage, Alps ranked sixth on the June 2024 TOP500 list. A later research paper reports a measured HPL result of about 434 petaflops and seventh place in the November 2024 TOP500. Those rankings are snapshots based on a particular benchmark, not a permanent measure of every scientific workload Alps can run.1Ā 3
Ā
Ā
How Alps is operatedāand what it Delves into
The evolutionĀ from a conventional, fixed-purpose supercomputer to a flexible research platform is central to Alpsā design. CSCS organizes resources into software-defined clusters, or vClusters. A vCluster can be configured for a research communityās needs, with its own computing resources and services. CSCS describes platforms for machine learning and AI, climate and weather, and general high-performance computing. This model gives researchers a more tailored environment while helping the centre manage shared infrastructure and confidentiality requirements.2Ā 3
Ā
The relentless evolutionĀ of scientific workloads also explains why Alps is heterogeneous. Alongside its large Grace Hopper installation, the system page lists CPU-only nodes and other GPU platforms, including NVIDIA A100 and AMD accelerators. Different combinations can suit different software and data needs; āmore powerfulā is not always the same as āmore suitable.ā Network, storage, software, and operational support all affect how quickly a research team can turn a question into a result.2 3
Ā
Furthermore, Alps extends beyond a single room in Lugano. CSCS describes a geographically distributed infrastructure that includes computing in Lausanne and Bologna, with data-archiving resources at the Paul Scherrer Institute. One operational benefit is resilience: the AlpsE extension in Lausanne can act as a failover for critical MeteoSwiss numerical weather-forecasting services if the Lugano site is unavailable. At Bologna, AlpsB connects researchers more directly to European weather and climate archives. The arrangement links compute capacity, data access and continuity of service across locations.2Ā 7
Ā
Science in operation
Weather forecasting is one concrete example. ETH Zurich and CSCS report that MeteoSwiss uses Alps to produce forecasts, and the 2024 inauguration announcement noted a move to a higher-resolution model better suited to Switzerlandās steep mountains and valleys. Higher resolution can represent smaller-scale features, but it also increases computational and data demandsāwhich is exactly the kind of work a national supercomputing facility is designed to support.1Ā 4
Ā
The rapid evolutionĀ of AI adds another use case. CSCS and ETH Zurich describe Alps as a resource for training complex models, including research relevant to medicine and climate science. This is a statement about research capacity, not a claim that a supercomputer alone delivers a clinical breakthrough: biomedical applications still depend on suitable data, scientific validation and specialist interpretation.1
Ā
Climate work increasingly depends on both computing and large, carefully organized datasets. In 2026, ETH Zurich and CSCS reported that around 100 petabytes of publicly available NASA Earth-observation data had been copied to CSCS. Their stated goal is to bring data, Alps computing and analytical methods closer together so researchers can train AI weather models and investigate environmental hazards. The data collection is being updated as new material becomes available; its value lies in making research data accessible for analysis, not simply in storing a headline-sized number of files.4
Ā
A new research Landscape
Alps is part of a broader shift in Switzerlandās scientific Landscape: large-scale computing is becoming a shared, service-oriented resource, rather than a single fixed machine reserved for one kind of calculation. Its distributed footprint also brings compute closer to data and helps protect essential services against a site outage.2Ā 7
Ā
Ā
Closing Thoughts
Alps is significant not just because of its peak performance, but because CSCS has built an operating model around the varied needs of research. A weather service may need dependable daily forecasts. An AI group may need access to large GPU clusters. A climate team may need both historical observations and the ability to process them at scale. The infrastructure is most useful when hardware, software and data access are planned together.
Ā
What does this look like in practice?
In practical terms, a researcher does not simply reserve āthe whole supercomputer.ā A project applies for resources, uses the platform and software environment best suited to its work, and submits jobs to the relevant cluster. For a climate project, that may mean working with a dedicated platform and large datasets; for an AI project, it may mean scheduling GPU-intensive training. CSCSās user webinar offers a practical walkthrough of these processes.6
Ā
Why does this make a difference?
It mattersĀ because national research infrastructure can change which scientific questions are feasible to ask. Better-resolved forecasts can help represent local terrain; high-performance AI systems can make larger model experiments possible; and colocating data with computing can reduce friction for research. None of these capabilities guarantees a discovery. They give researchers more room to test ideas, compare methods and produce evidence that can support decisions in science and society.
Ā
Reference videos
Official Inauguration of Alps ā CSCS on Vimeo. A short video from the centreās inauguration coverage.5
Getting Started on Alps Platforms ā CSCS webinar. An operational orientation for users of the machine-learning, climate-and-weather, and HPC platforms. The session covers allocations, job submission, filesystems and support resources.6
References
Comments