# Alps by Swiss National Supercomputing Centre (CSCS)

Canonical page: https://technooptimist.io/computing/alps
Frontier: [Computing](https://technooptimist.io/computing.md) · Field: Supercomputers · Company: [Swiss National Supercomputing Centre (CSCS)](https://technooptimist.io/companies/cscs.md)
Stage: Deploy (4 of 5: Research → Proto → Pilot → Deploy → Scaling)
Updated: last update 2026-09-20, facts checked 2026-10-11
Goal: Give Swiss science one machine that handles simulation, data analysis and AI at frontier scale, including training fully open AI models the country controls, instead of renting compute from US clouds.
Status: In full production; No. 10 on the TOP500 (Jun 2026) and serving Apertus 1.5 to researchers since July 2026.
Next milestone: Apertus 2.0, target 2027

## Milestones

| Status | Milestone | Date | Slip |
| --- | --- | --- | --- |
| Done | First phase: No. 6 on TOP500 (270 PFlop/s) | Jun 2024 |   |
| Done | Inaugurated in Lugano | 14 Sep 2024 |   |
| Done | Full system: 434.9 PFlop/s, No. 7 on TOP500 | Nov 2024 |   |
| Done | Apertus 1.0 open model released, trained on Alps | Sep 2025 |   |
| Done | Apertus 1.5 offered to researchers as a service | Jul 2026 |   |
| Current | Apertus 2.0 | 2027 |   |

## Upcoming events

- **2027**: Apertus 2.0

## Most important updates

- **20 Sep 2026**: CSCS completes Alps' shift to a cloud-style research platform; Apertus 1.5 served since July. CSCS says Alps now runs as an integrated stack of hardware, software and services rather than a fixed mainframe. ([our story](https://technooptimist.io/stories/cscs-completes-alps-shift-to-a-cloud-style-research-platform-apertus-1-fd8dd487.md) · [original source](https://www.netzwoche.ch/news/2026-09-20/mehr-als-ein-supercomputer))
- **Sep 2025**: Swiss AI Initiative releases Apertus, a fully open 70B language model trained on Alps. EPFL, ETH Zurich and CSCS released Apertus in 8B and 70B sizes, trained on 15 trillion tokens covering over 1,000 languages, with weights, data and recipes open. ([our story](https://technooptimist.io/stories/swiss-ai-initiative-releases-apertus-a-fully-open-70b-language-model-a3f3b3c5.md) · [original source](https://opendatascience.com/switzerland-launches-apertus-a-fully-open-ai-model-for-research-and-industry/))
- **14 Sep 2024**: Alps supercomputer inaugurated in Lugano with 10,752 NVIDIA Grace Hopper superchips. ETH Zurich opened Alps at CSCS, an HPE Cray machine with 10,752 GH200 superchips aiming for about half an exaflop of peak performance. ([our story](https://technooptimist.io/stories/alps-supercomputer-inaugurated-in-lugano-with-10-752-nvidia-grace-06f2a079.md) · [original source](https://www.pctipp.ch/themen/news/alps-supercomputer-im-tessin-eingeweiht))

## Current obstacles

- **Falling down the rankings**: Unchanged since late 2024, Alps slid from No. 7 to No. 10 on the TOP500 as larger systems arrive.
- **Big jobs crowd the machine**: Training Apertus took about 40% of Alps for three to four months, competing with weather and science users.

## Physics limits

- **Power sets the size**: At about 7 MW, each further step in performance needs proportionally more electricity and cooling unless chips get more efficient per operation.
- **Communication caps scaling**: Splitting one job over thousands of GPUs means constant data exchange. Network speed and latency, not chip speed, often set how far a job can scale.

## How it works

### Node: Four Grace Hopper modules

Each node holds four GH200 superchips: an Arm CPU and an H100 GPU on one module, sharing memory over a fast link.

### Network: Slingshot ties it together

HPE's Slingshot-11 network links the 2,688 GPU nodes so thousands of GPUs can work on one job, such as training a language model.

### vClusters: Carved up in software

Instead of one fixed system, Alps is split into virtual clusters, each with its own software stack for a given community.

## Spec sheet

| Spec | Alps | Provenance |
| --- | --- | --- |
| NVIDIA GH200 superchips | 10,752 | reported |
| Linpack performance (Rmax) | 434.9 PFlop/s | reported |
| Theoretical peak (Rpeak) | 574.8 PFlop/s | reported |
| Power during Linpack | 7.1 MW | reported |
| Total cores (TOP500) | 2,121,600 | reported |
| TOP500 rank (Jun 2026) | 10 | reported |

Provenance: reported = stated by the company; estimated = our estimate; sample = a sample figure.

## About Swiss National Supercomputing Centre (CSCS)

CSCS, part of ETH Zurich, runs Switzerland's national supercomputers in Lugano. Its Alps machine, built on NVIDIA Grace Hopper chips, serves weather forecasting, science and the training of Apertus, Switzerland's fully open language model.

- Alps: 10,752 GH200 superchips, 434.9 PFlop/s, No. 10 on the June 2026 TOP500.
- Trained Apertus, a fully open 8B/70B model, on about 40% of Alps for three to four months.
- Cooled with Lake Lugano water; waste heat feeds district heating.

- Headquarters: CH
- Status: government (ETH Zurich)
- People: Thomas Schulthess (Director, CSCS)
- Partners, customers, investors: HPE (supplier); NVIDIA (supplier); MeteoSwiss (customer)

Company page: https://technooptimist.io/companies/cscs.md

## Related programs

Everything in Computing: https://technooptimist.io/computing.md

---

The Techno Optimist tracks every frontier of technology, program by program: https://technooptimist.io/ (a Markdown version of any page: add .md to its URL; site map for agents: https://technooptimist.io/llms.txt).
Data API (JSON, realtime stream): https://technooptimist.io/developers. People read 5 pages free, then Reader $10/month; these Markdown pages are open to agents.