Europe Enters the Exascale Era with JUPITER Supercomputer

Europe has officially entered the exascale computing era with the launch of JUPITER, the continent’s first supercomputer capable of exceeding a billion-billion calculations per second. The system is now live and was unveiled at an inauguration ceremony with European and German leaders, marking a milestone for science, industry, and the region’s digital sovereignty.

What “Exascale” Means — in Plain Language

Computers measure speed by how many arithmetic operations they can do each second. Exascale means at least one quintillion (10¹⁸) operations every second. If that number feels abstract, think of it this way: a single exascale system can do in one second what a powerful laptop would need many years to complete. The United States crossed the threshold first in 2022 with the Frontier supercomputer, which delivered more than one exaflop on the widely used LINPACK benchmark. JUPITER brings that level of capability to Europe for the first time and places European researchers and companies in the top tier of global computing power.

The Machine: How JUPITER Reaches Exascale

JUPITER combines several cutting-edge technologies to hit exascale performance while remaining unusually energy-efficient for its class:

Processor platform. The system is powered by NVIDIA Grace Hopper (GH200) Superchips, which pair CPU and GPU on one module to accelerate both traditional simulations and modern AI.System architecture. Built on Eviden’s BullSequana XH3000 liquid-cooled cabinets, JUPITER is engineered for dense performance with direct warm-water cooling that carries heat away efficiently.Scale and interconnect. Roughly 24,000 GH200 Superchips are tied together over NVIDIA Quantum-2 InfiniBand with around 51,000 high-speed links, enabling the machine to move data at extraordinary rates and keep all those processors busy.Storage and data center. The system integrates nearly an exabyte of storage and sits inside a modular data-center complex of about 50 specialized containers, a design that sped up deployment and simplifies future upgrades.

In scientific double-precision math (FP64), JUPITER can execute about one quintillion operations per second. For AI workloads that use lower-precision math, it’s expected to reach up to about 90 “AI exaflops” of peak throughput — making it one of the world’s most capable AI supercomputers as well.

Who Built It — and Why It Matters Politically

JUPITER is a EuroHPC project, financed and delivered through a coalition that includes the European High-Performance Computing Joint Undertaking, the German federal government, North Rhine-Westphalia, and an industrial team led by Eviden (Atos) and ParTec, with NVIDIA providing the accelerated-computing platform. The result is Europe’s first exascale-class system and, as of its launch, the fastest in Europe and among the very fastest worldwide. Beyond prestige, it strengthens Europe’s tech sovereignty by giving researchers and companies access to world-class compute on European soil rather than relying on overseas infrastructure.

Politically, that matters. High-performance computing underpins advancements in artificial intelligence, security, climate policy, automotive design, pharmaceuticals, and more. Countries with exascale capacity can iterate faster, keep sensitive data closer to home, and build ecosystems of talent and industry around their facilities. JUPITER’s debut signals that Europe intends to be a producer — not merely a consumer — of cutting-edge compute.

What JUPITER Will Be Used For

The machine is already earmarked for a broad portfolio of projects across science and industry:

Climate and weather. Teams such as the Max Planck Institute for Meteorology are running kilometer-scale climate simulations that can better capture violent thunderstorms, heavy rainfall, and other extreme events — a leap in resolution that was impractical before. Better resolution can translate into more reliable predictions and policy-relevant climate insights.Energy and materials. Exascale simulations help design next-generation batteries, catalysts, semiconductors, and renewable-energy systems, letting researchers test ideas virtually before they ever build a prototype. That accelerates R&D cycles and lowers the cost of discovery.Artificial intelligence. JUPITER doubles as Europe’s most advanced AI supercomputer, built to train large language models (LLMs) across European languages and to enable foundation models for imaging, video, and other data-intensive domains. By keeping training on European infrastructure, projects can adhere more easily to local regulations and data-sovereignty requirements.Medicine and neuroscience. Researchers will use tools like detailed neuron simulators to model brain activity down to sub-cellular behavior, relevant to conditions such as Alzheimer’s and epilepsy. Molecular-dynamics campaigns will tackle massive biomolecular assemblies — pushing toward credible digital twins of organs for testing drugs and treatments in silico.Quantum computing R&D. With vast memory and bandwidth, JUPITER is poised to break records in simulating quantum circuits, potentially surpassing previous limits on qubit-level simulations. That allows scientists to test quantum algorithms and hardware ideas before physical devices reach that scale.Energy Efficiency: Big Power, Smaller Footprint

Supercomputers can consume double-digit megawatts of power, so efficiency is not a side note — it’s central. JUPITER’s design emphasizes performance per watt from the ground up.

Direct warm-water cooling draws heat from CPUs and GPUs more efficiently than air. Because the water leaves the racks warm rather than cold, that energy can be reused.Heat reuse is built into the campus plan: waste heat from JUPITER will warm nearby buildings, turning a by-product into a resource and reducing overall emissions.Efficient silicon matters, too. The Grace Hopper architecture is optimized for high performance-per-watt in both simulation and AI, helping the system deliver more results per unit of energy.Prior to the full deployment, a pilot rack using the same technology topped the Green500 energy-efficiency list, and the finished system is cited as the most energy-efficient among the world’s top-five fastest. Pairing top-tier speed with best-in-class efficiency is rare and important as data-center electricity use climbs worldwide.

All of this adds up to a design that tackles the environmental critique head-on: yes, exascale is powerful, but it doesn’t have to be wasteful.

Why This Is a Big Deal for Europe’s Economy

Leaders across Europe frame JUPITER as an economic as well as scientific platform. With AI and advanced simulation now vital to sectors from biotech and automotive to finance and energy, having home-region exascale capacity is a competitive lever:

It lowers barriers for European startups and SMEs to train and deploy state-of-the-art AI models and run massive simulations without exporting data or depending on non-EU providers.A forthcoming JUPITER AI Factory concept aims to make access feel more like a cloud service — crucial for companies that need bursts of huge compute but don’t own their own supercomputers.By aligning public science missions with industry access, Europe can translate HPC breakthroughs into products faster, from safer materials and lighter vehicles to cleaner energy systems and medical advances.

There’s also a talent dimension. Facilities like JUPITER attract and retain engineers, mathematicians, chemists, and computer scientists. They create training pipelines with universities, fund doctoral programs tied to real-world problems, and help anchor regional innovation clusters. Over time, that feeds a virtuous cycle: better tools bring better people; better people build better tools.

How It Came Together So Quickly

JUPITER’s deployment was accelerated by a modular data-center approach — essentially prefabricated, high-tech “containers” that interlock into a full facility. This shortened construction, reduced on-site disruption, and eases future expansions as chip generations evolve. The site integrates power distribution, networking, and liquid-cooling infrastructure in a layout that prioritizes serviceability and uptime. It’s a template Europe can reuse for the next wave of machines, improving consistency and reducing project risk.

Public Access, Governance, and Data Stewardship

A supercomputer of this scale invites questions about who gets to use it and under what conditions. JUPITER’s roadmap foresees a mix of academic, public-interest, and industrial workloads, allocated through peer-reviewed proposals, strategic initiatives, and commercial channels. Clear governance is crucial: transparent allocation policies; strong privacy and security for sensitive datasets; and adherence to European norms around data protection, safety, and accountability in AI. With those guardrails, JUPITER can expand access without compromising trust.

Challenges to Watch

Even with the triumphant launch, several challenges deserve attention:

Software readiness. Reaching peak performance requires code that’s optimized for GPUs, high-bandwidth memory, and intricate interconnects. Many scientific codes still need extensive modernization.Fair scheduling. Balancing blockbuster projects (like continent-scale climate simulations) with smaller but promising proposals from young teams or startups requires careful policy and communication.Lifecycle and upgrades. Exascale hardware evolves quickly. Planning for incremental upgrades, spare-parts logistics, and compatibility with future processors will keep the system competitive.Energy markets. While highly efficient, JUPITER still draws significant power. Long-term contracts for green electricity and continued investment in heat-reuse infrastructure will keep operating costs and emissions in check.A New Dawn for European Supercomputing

JUPITER isn’t just a bigger computer; it’s a platform for discovery and competitiveness. For scientists, it unlocks simulations with finer resolution and longer timescales than ever before — from thunderstorm-resolving climate models to molecular systems that rival the complexity of life. For industry, it enables faster product cycles, safer designs, and more capable AI trained on European languages and data. For policymakers, it demonstrates that pan-European cooperation can deliver infrastructure at the frontiers of what’s technically possible — efficiently and on schedule.

The exascale race isn’t over; machines will keep getting faster, more specialized, and more efficient. But with JUPITER switched on and already at work, Europe has taken a defining step — from consumer to shaper of the world’s most advanced computing.