Procurement of GPU computing resources for AI-based climate and weather modelling
Data og analyse
Danmarks Meteorologiske Institut needs services for the procurement of GPU computing resources for AI-based climate and weather modelling through a framework agreement over 4 years, with a maximum total value of 50,000,000 DKK. The resources will provide scalable computing capacity for large-scale model training, inference, simulations, and other computational workloads, supporting modern forecasting and climate analysis systems. Award criteria are Price (70%) and Quality (30%). Qualification requirements include a minimum equity of 250,000,000 DKK in the most recently completed financial year, and 1 reference of delivery of GPU computing resources with 8× NVIDIA H100 GPUs per node or equivalent for AI-based modelling, completed within the last 2 years.
About the bid
- Contact name
- Stine Schou Kristiansen
- Contact phone
- 72545986
- Contact email
- koncernindkoeb@kefm.dk
- Address
- Sankt Kjelds Plads 11, 2100 København Ø
Recent advances in weather and climate modelling, combined with machine learning methods, have significantly increased the computational requirements of the Contracting Authority associated with modern forecasting and climate analysis systems. These activities rely on the processing of large datasets and computationally intensive workflows. The procured GPU computing resources will provide the scalable computing capacity required to support large-scale model training, inference, simulations, and other computational workloads across the Contracting Authority’s research and operational environments.
Custom Assessment Criteria
Start a free trial to ask your own questions about the tender and get AI-generated answers.
Award Criteria
Description
See Tender Conditions section 6.1.1.
Description
The sub-criteria Quality consists of seven sub-sub-criterion: ER 1: High-speed compute fabric (10 %). ER2: GPU Scalability 2 (20 %). ER3: GPU Scalability 3 (20 %). ER4: Storage 4 (10%). ER5 Storage 5 (10%). ER6: Dependency installation (10 %). ER7: Remote access (20 %). See Tender Conditions section 6.2.1.
Sub-criteria
Description
High-speed compute fabric.
Description
GPU Scalability 2.
Description
GPU Scalability 3.
Description
Storage 4.
Description
Storage 5.
Description
Dependency installation.
Description
Remote access.
Qualification requirements
Description
The tenderer must have a minimum equity of 250.000.000 DKK in the most recently completed financial year
Documentation Requirement
The tenderer’s most recently approved annual report, or other documentation evidencing equity, such as an auditor-certified statement.
Description
The tenderer must provide 1 reference of delivery of GPU computing resources with 8× NVIDIA H100 GPUs per node or equivalent for AI-based modelling. The reference must have been completed within the last 2 years, counted from the deadline for submission of tenders. The reference must include date, recipients and a brief description. The reference is considered completed when the GPU computing resources have been made available to the contracting party.
Documentation Requirement
The tenderer's ESPD or the attached separate document is considered as documentation of technical and professional ability. However, the Contracting Authority reserves the right to verify information, for example, by contacting the listed contacts of the references.
Previous Contract Winners
Start a free trial to see who has won similar tenders for this buyer.
See the full notice
- Contracting authority
- Danish Meteorological Institute (DMI)
- Title
- Procurement of GPU computing resources for AI-based climate and weather modelling
- Type
- Main code
- Supplemental codes
- Estimated value
- DKK 50,000,000.00
- Duration
- 4 years