The European supercomputer JUPITER has achieved something never seen before: recreating step by step how a real quantum processor works. The achievement comes from Jülich Research Centerone of the largest research centers in Europe, located in Germany, and specialized in supercomputing, energy, neuroscience and quantum technology. From there, its scientists have managed to reproduce the internal physics of a quantum chip with an unprecedented level of detail, a milestone that opens up new possibilities for research across the continent.
What makes this simulation special is the amount of information it moves. According to the Forschungszentrum Jülich itself, each operation of the simulated quantum processor manipulates more than two quadrillion complex numerical values. These are figures that are difficult to imagine: a quadrillion is a ‘1’ followed by fifteen zeros. That is, a single operation manages more data than a normal computer could handle in its entire useful life.
To understand why it is so complicated, just remember that quantum computers work differently from current ones. Instead of working only with zeros and ones, its basic units—the qubits— can be in several states at the same time. This characteristic makes them very powerful, but also extremely difficult to imitate with classic technology.
What JUPITER has achieved
The simulation carried out on JUPITER reproduces a quantum processor of 50 qubitsand doing so accurately requires an almost inconceivable memory: around 2 petabyteswhich is equivalent to two million gigabytes. To put it in perspective, it would take about 2,000 hard drives from a terabyte —the typical ones sold in stores— to store that amount.
This memory requirement is the main reason why simulating a full quantum computer has, until now, been difficult. a practically unapproachable challenge even for the most powerful supercomputers in the world.
The result, according to the European research center, “exceeds the previous world record of 48 qubitsestablished by Jülich researchers in 2019 on the K supercomputer in Japan.” “This demonstrates the immense computational power of JUPITER and opens new horizons for the development and testing of quantum algorithms,” they add.
However, this milestone does not mean that a practical 50-bit quantum computer already exists. qubits fully functional — the technology still has to overcome challenges such as bug fixes or stability of the qubits—but it does allow test algorithms, anticipate problems and move faster towards those future systems.
For Europe, it represents a strategic boost: it strengthens its position in advanced supercomputing and reduces dependence on external infrastructures in a key field for science, industry and digital security.
How has this progress been possible?
This achievement is not only because JUPITER is very powerful, but because it combines several new ideas to make better use of its memory and ability. hardware.
The first key is in the superchips NVIDIA GH200which combine a CPU and a GPU in the same piece. This allows, when the GPU memory runs out, some of the data is automatically moved to the CPU memory without performance sinking. Thanks to this hybrid system, JUPITER can handle more information than a GPU alone would normally allow.
The second key has been the work of the Jülich and NVIDIA team to improve their quantum simulator. They have created a new version, called JUQCS-50which is designed to work just with this mix of memories. Thus, even though the data is distributed between the CPU and GPU, the simulation continues to run smoothly.
In addition, they have added two important improvements:
- A compression method which reduces the memory required up to eightfold.
- A system that optimizes how data moves between the chips.
Those responsible for the project explain that this tool, JUQCS-50, will also be available to other institutions through Jülich’s quantum infrastructure, called JUNIQwhich will allow more researchers to use it in the future.
JUPITER, the machine that allows the impossible
At first glance, JUPITER might seem like ‘just’ another supercomputer in that long list of giant machines that Europe has been accumulating in recent decades. But you only have to enter the enormous building of the Forschungszentrum Jülich to realize that something different is happening here.
The hallway opens to a room as large as four tennis courtsilluminated by the constant glow of thousands of indicators. Under the technical floor, a network of 260 kilometers of wiring It connects each module of the system as if they were neurons in a huge brain. And, in a way, they are: this is the heart of European supercomputing.
JUPITER is not built like traditional supercomputers. Its modular architecture – something like a technological city built by neighborhoods – allows it to be expanded or updated without ever being torn down. Inside their black closets they beat 24.000 superchips NVIDIA GH200a combination of processors and GPU designed to move amounts of data that no ‘normal’ computer can even dream of.
Each of these chips works hand in hand thanks to an ultra-fast interconnection network. The result is a computing capacity that exceeds the quintillion operations per seconda figure that is difficult to even pronounce. In practical terms, it’s as if the entire world’s population did math non-stop for thousands of years… and still wouldn’t come close to what JUPITER solves in the blink of an eye.
And yet, the most surprising thing about this machine is not its raw power, but how to use it. Its direct liquid cooling system keeps its thousands of components stable while reusing heat to heat buildings on campus. Not only does it break performance records: it has also come to dominate the Green500 ranking, which classifies the most efficient supercomputers on the planet.
JUPITER is already working on ultra-precise climate models, biomedical simulations, new sustainable materials and energies, and now, something even more ambitious: exact reproduction of the behavior of a real quantum processor.
