Gefion is Denmark’s national AI supercomputer, not Europe’s sole or formally designated AI engine. Operated by the Danish Centre for AI Innovation (DCAI), it gives Danish researchers, companies and public bodies access to large-scale computing in Denmark. Its importance is real; the continent-wide superlative is promotional shorthand, not a verified performance ranking.
What is the Gefion supercomputer?
Named for a goddess in Danish mythology, Gefion is a shared supercomputer built to train and run AI models and support other demanding research. It is an “AI factory”: compute, storage, software and user support organized for AI projects, rather than a chatbot or an ordinary cloud virtual machine.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
NVD RTX PRO 6000 Blackwell Professional Workstation Edition Graphics Card for AI, Design,... | $19,999.99 | Buy on Amazon |
| 2 |
|
NVIDIA RTX PRO 4000 Blackwell Graphics Card - 24GB GDDR7 ECC Memory, PCIe 5.0 x16, 4X DisplayPort... | $3,134.14 | Buy on Amazon |
| 3 |
|
PNY NVIDIA RTX A6000 | $5,981.00 | Buy on Amazon |
The Danish Centre for AI Innovation operates the facility in the Copenhagen area. The centre was established with funding from the Novo Nordisk Foundation and Denmark’s Export and Investment Fund; NVIDIA is a strategic technology partner and supplies core technology, not the Danish operator. The system was inaugurated on October 23, 2024, with NVIDIA CEO Jensen Huang and King Frederik X taking part. DCAI explains its role and access model, while the Novo Nordisk Foundation outlines the centre’s establishment.
Gefion is intended for universities and research institutions, startups and scale-ups, public bodies, life-sciences companies and larger businesses. Its national role is to make substantial AI capacity available to organizations that may not be able to build or routinely rent a large cluster themselves.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware match#1 Best Overall
- PLEASE NOTE: Exporting an NVIDIA RTX Pro 6000 GPU outside the US requires strict adherence to the U.S. Export Administration Regulations (EAR) and issuance of an export license from the Bureau of Industry and Security (BIS). Compliance and Know Your Customer (KYC) screening may be required as a condition of order acceptance. [NVIDIA Blackwell Streaming Multiprocessor] The new SM features increased processing throughput, and new neural shaders that integrate neural networks inside of programmable shaders | DLSS 4: Multi Frame Generation ensures ultra-smooth frame pacing for lifelike simulations.
- [Double-Flow-Through Design] The RTX PRO 6000 Blackwell features a double-flow-through cooling design, optimizing efficiency and airflow to sustain peak performance under 600W power loads. | [5th Gen Tensor Cores] Deliver up to 3X the performance of the previous generation and support for FP4 precision for faster AI model processing times with reduced memory usage, enabling local fine-tuning of LLMs and generative AI | [4th Gen Ray Tracing Cores] Double the ray-triangle intersection rate of the previous generation to create photoreal, physically accurate scenes and immersive 3D designs with RTX Mega Geometry, which enables up to 100X more ray-traced triangles.
- [PCIe Gen 5] Support for PCIe Gen 5 provides double the bandwidth of PCIe Gen 4, improving data-transfer speeds from CPU memory and unlocking faster performance for data-intensive tasks like AI, data science, and 3D modeling. | [GDDR7 Memory] With 96 GB of GPU memory and 1.8 TB ps bandwidth, it can tackle massive 3D and AI projects, fine-tune AI models locally, explore large-scale VR environments, and drive larger multi-app workflows.
- [DisplayPort 2.1] Achieve unparalleled visual clarity and performance, driving high resolution displays at up to 8K at 240 Hz and 16K at 60 Hz. Increased bandwidth enables seamless multi-monitor setups while HDR and higher color depth support ensures superior color accuracy for precision work, such as video editing, 3D design, and live broadcasting.
- [Universal MIG] Divide a single RTX PRO 6000 Blackwell into multiple isolated instances, each with dedicated resources, allowing for concurrent execution of multiple workloads, optimized GPU utilization, and secure isolation of different applications or users. [WARRANTY] 3 YR Manufacturer's Warranty. Bulk OEM Packaging. Retail Packaging is NOT included.
What hardware does Gefion use?
The launch system and DCAI’s current description refer to different configurations, so their figures should not be blended into one snapshot.
| Specification | What the source says |
|---|---|
| Launch configuration, October 2024 | 1,528 NVIDIA H100 Tensor Core GPUs in an NVIDIA DGX SuperPOD, connected by NVIDIA Quantum-2 InfiniBand. NVIDIA’s launch announcement. |
| Later configuration described by DCAI | More than 1,540 GPUs, combining NVIDIA DGX H100 and B300 systems, plus 110 petabytes of WEKA high-performance storage. DCAI does not state the exact B300 count on its page. DCAI’s Gefion specifications. |
| Software highlighted by DCAI | NVIDIA BioNeMo for life-sciences research and CUDA Quantum for hybrid CPU, GPU and quantum-processing workflows. DCAI’s Gefion page. |
DCAI’s current description does not establish the precise deployment status of each B300 system or give a new benchmark for the expanded configuration. The publicly listed TOP500 result therefore should be read as a record for the benchmarked H100-era system, not assumed to represent every later production change.
How powerful is Gefion?
In the June 2026 TOP500 list, Gefion ranked No. 43 worldwide on the HPL benchmark, which measures conventional high-performance computing performance. The system record gives 66.59 petaflops of HPL performance, a theoretical peak of 100.63 petaflops, HPCG performance of 749.786 teraflops and reported power consumption of 1,753.20 kW. Its June 2026 Green500 position was No. 69, at 44.832 gigaflops per watt. These are dated records, not an all-purpose measurement of AI capability. TOP500’s Gefion record and the June 2026 Green500 list provide the underlying figures.
HPL is not the same as a benchmark of training a particular AI model. AI work often uses lower-precision arithmetic such as FP8, FP16, BF16 or INT8; advertised AI-performance figures may use different precisions and workloads from TOP500’s HPL test. GPU count and theoretical peak alone also do not show how efficiently a real job scales across the network, memory and storage.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsSo “Europe’s fastest” cannot be inferred from Gefion’s launch or its TOP500 place. Any such comparison needs a named benchmark, precision, date and comparison set—especially now that DCAI describes a later mixed-GPU configuration.
Why did Denmark build it?
Gefion is part of an effort to give Denmark domestic access to advanced computing, not simply a showcase for NVIDIA hardware. NVIDIA’s account of the project points to recurring barriers: scarcity and cost of advanced GPUs, delays in getting access, and a lack of technical support for organizations trying to use large-scale systems. NVIDIA’s 2025 GTC session discusses those motivations.
Rank #2
- Professional GPU with Blackwell Architecture
- Blackwell Architecture
- 24GB GDDR7 with PCIe 5.0 & Ray Tracing
- AI Workstation
A shared national facility can also help keep sensitive project data and workloads under Danish operational and legal arrangements, while connecting organizations to infrastructure and expertise they could not readily assemble alone. The intended fields include life sciences, healthcare, weather and climate, green-transition technology and quantum computing. Its value will depend not just on hardware, but on whether researchers and companies can gain useful access and turn that capacity into results.
What does “sovereign AI” mean here?
In Gefion’s context, sovereignty is about where data is stored and workloads run, which jurisdiction applies, who administers the infrastructure, and how Danish institutions and companies obtain access. DCAI says data and workloads remain under Danish sovereignty and describes its platform as designed to meet GDPR, NIS2 and ISO 27001 requirements. Those are DCAI’s own operational and compliance statements; they should not be read as an independent finding about every workload or as proof of complete technological independence. DCAI describes its security and sovereignty approach.
There is a strategic trade-off: Denmark can exercise local operational control while relying heavily on NVIDIA GPUs, networking and software. Sovereignty can reduce reliance on foreign cloud hosting for a project; it does not remove dependence on a foreign technology supplier or its ecosystem.
What work is being done on Gefion?
Drug discovery and life sciences
NVIDIA announced a collaboration with Novo Nordisk and DCAI to use Gefion for drug-discovery and agentic-AI work. The announcement also described a venture-backed company using the system to investigate oral alternatives to biologic medicines and difficult-to-drug proteins. These are announced projects and research aims, not evidence by themselves that a medicine or commercial breakthrough has resulted. The collaboration announcement describes the work.
Weather and climate
The Danish Meteorological Institute is developing an AI weather model on Gefion. That is an active development project; it should not be conflated with a claim that the system has already replaced operational forecasting. DMI describes the project.
Quantum computing and other targets
DCAI lists CUDA Quantum as a supported platform for hybrid workflows involving CPUs, GPUs and quantum processing units. Healthcare, green transition and fault-tolerant quantum computing are among the centre’s broader target areas. A stated target or available software platform signals intended capability, not a completed scientific result. DCAI’s overview sets out those wider aims.
Rank #3
- NVIDIA Ampere Architecture-based CUDA Cores - Double-speed processing for single-precision floating point (FP32) operations and improved power efficiency provide significant performance improvements for graphics and simulation workflows, such as complex 3D computer-aided design (CAD) and computer-aided engineering (CAE), on the desktop.
- Second-Generation RT Cores - With up to 2X the throughput over the previous generation and the ability to concurrently run ray tracing with either shading or denoising capabilities, second-generation RT Cores deliver massive speedups for workloads like photorealistic rendering of movie content, architectural design evaluations, and virtual prototyping of product designs. This technology also speeds up the rendering of ray-traced motion blur for faster results with greater visual accuracy.
- Third-Generation Tensor Cores - New Tensor Float 32 (TF32) precision provides up to 5X the training throughput over the previous generation to accelerate AI and data science model training without requiring any code changes. Hardware support for structural sparsity doubles the throughput for inferencing. Tensor Cores also bring AI to graphics with capabilities like DLSS, AI denoising, and enhanced editing for select applications.
- Third-Generation NVIDIA NVLink - Increased GPU-to-GPU interconnect bandwidth provides a single scalable memory to accelerate graphics and compute workloads and tackle larger datasets.
- 48 Gigabytes (GB) of GPU Memory - Ultra-fast GDDR6 memory, scalable up to 96 GB with NVLink, gives data scientists, engineers, and creative professionals the large memory necessary to work with massive datasets and workloads like data science and simulation.
Who can use Gefion, and is it free?
DCAI says it serves public and private organizations, including businesses, startups and academia. In practice, access may come through a direct commercial relationship, a research grant, a partnership or an allocated access programme; it is not simply a public, instant-signup GPU service. DCAI describes a GPU-based fee model but has not broadly published final pricing, so prospective users should confirm current terms directly. DCAI’s access information is the starting point.
Eligible researchers affiliated with Danish universities, hospitals or nonprofit research institutions may apply for Novo Nordisk Foundation grants supporting Gefion access. Such funding is a research-grant route, not a general subsidy for commercial users. The foundation lists its Gefion access grants.
EuroHPC AI Factory calls offer access time free of charge under programme conditions, including eligibility and review. That does not make all DCAI access free or guarantee an immediate or unlimited allocation. The EuroHPC FAQ explains the conditions; its Fast Lane and Large Scale routes have separate application processes and project requirements.
Before pursuing access, a team should check whether its workload can use multiple GPUs efficiently; its memory, interconnect and storage needs; data-transfer and classification requirements; and compatibility with its containers, CUDA, NCCL, MPI and machine-learning framework versions. Training, fine-tuning, inference and simulation can have very different resource profiles. Teams should also confirm support arrangements, allocation availability and the technical readiness expected by their chosen route.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →How does Gefion fit into Europe’s AI infrastructure?
Gefion is one national facility in a wider European build-out. NVIDIA said in June 2026 that 35 NVIDIA AI supercomputers were in development across 23 European countries, including projects such as Barcelona Supercomputing Center’s MareNostrum 5 AI upgrade, BavariaAI’s Blue Swan, Italy’s IT4LIA, Germany’s HammerHAI and Sweden’s Mimer AI Factory. That is NVIDIA’s account of systems in development, not a like-for-like performance ranking. The announcement lists the build-out.
| Resource | What it is | How to compare it with Gefion |
|---|---|---|
| Gefion | Denmark-focused AI facility operated by DCAI; launched with H100 GPUs, with a later H100/B300 configuration described by DCAI. | Its TOP500 HPL record applies to a specific benchmarked system configuration, not automatically to all AI workloads or the later expansion. |
| Isambard-AI | UK AI research system based on 5,448 NVIDIA Grace Hopper GPUs. | Its research paper reports more than 21 AI exaflops at 8-bit precision. That figure is not directly comparable with Gefion’s HPL petaflops because the precision and measurement differ. Isambard-AI’s research paper. |
| EuroHPC AI Factories | A European network of AI infrastructure and support, rather than one physical supercomputer. EuroHPC describes 19 AI Factories and 13 AI Factory Antennas. | It is an access and support framework; an applicant’s route and eligibility matter as much as the name of any individual machine. EuroHPC’s network overview. |
For a prospective user, the relevant comparison is not a continent-wide league table. It is whether a particular access route provides the required GPU generation, data jurisdiction, support, schedule and price for the workload.
When might Gefion be the right fit?
- Consider it if your organization needs substantial multi-GPU capacity, Danish operational control, or a research or industry partnership in fields such as life sciences, climate or healthcare.
- Check access and costs first if your project depends on predictable budget, a quick start or a guaranteed allocation. DCAI’s final pricing is not broadly published, and grant or EuroHPC access requires the relevant application and eligibility.
- Compare other options for small, occasional inference jobs or teams that need instant, elastic, globally distributed cloud capacity; a hyperscaler may be operationally simpler, though it will have different jurisdiction and governance implications.
- Plan for the software stack if your code depends on AMD, Google TPU or non-CUDA systems. NVIDIA-centric infrastructure can mean migration work, and a sovereign hosting location does not remove that technology dependency.
Gefion is best understood as Denmark’s sovereign AI engine and a notable European AI factory: a shared national capability with a real place in Europe’s expanding infrastructure. Calling it “Europe’s new AI engine” is defensible as a metaphor for that growth, but not as a claim that it is Europe’s only, largest or clearly dominant AI supercomputer.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Recommended Free Tools
