TotalEnergies unveils Pangea 5 supercomputer, increases computing sixfold

TotalEnergies announced it has contracted Dell Technologies, NVIDIA, and storage vendor DDN to design and install Pangea 5, a next-generation supercomputer at the Jean Feger Scientific and Technical Center (CSTJF) in Pau, France, according to the company's newsroom release and corroborating coverage from HPCwire and DataCenterDynamics. The system is expected to enter service in 2027, deliver roughly 75 petaflops of compute, according to TotalEnergies materials cited by Interesting Engineering and DataCenterDynamics, and multiply the company's computing power sixfold versus its current Pangea infrastructure. The project represents an investment of more than 100 million euros (about $117 million), and will replace the existing system to run seismic imaging for hydrocarbon exploration, geological modeling, energy simulation, and AI-driven research. "By increasing our computing power sixfold, we are strengthening our leadership in high-performance computing, ensuring that our expert teams continue to have the means to push the envelope to support the development of our activities and meet the growing global demand for energy," said Namita Shah, President of OneTech at TotalEnergies.

According to TotalEnergies and its vendor partners, the platform will rely on NVIDIA GPUs, CPUs, and InfiniBand networking for massively parallel processing, with Dell handling system design and installation and DDN providing storage. TotalEnergies states Pangea 5 will use roughly 40% less energy than its predecessor at equivalent performance and cut cooling-system energy use by about fivefold, with waste heat piped into the CSTJF campus's district heating system. Dell Technologies EMEA president Adrian McDonald said the system is built "to accelerate discovery, increase efficiency, and drive the energy transition forward," while NVIDIA's VP of HPC and AI, John Josephakis, described the architecture as built to meet "the most demanding industrial and energy challenges."

The announcement continues a broader pattern of energy majors pairing large-scale HPC with AI workflows to accelerate seismic imaging, reservoir simulation, and grid optimization; TotalEnergies separately expanded an AI-platform partnership with Cognite across its upstream asset base last year. Pangea 5 is notable as a large, industry-funded HPC investment explicitly framed around both traditional scientific simulation and AI research, and its efficiency targets (energy and cooling reductions plus heat recycling) reflect growing operational attention to total cost of ownership and sustainability in data-center design across the sector.

Teams evaluating similar industrial-AI infrastructure investments should note that Pangea 5's stated design goals - GPU-accelerated parallel compute paired with a roughly 40% efficiency gain and heat reuse - signal that vendor ecosystems led by NVIDIA and integrators like Dell are increasingly selling efficiency and sustainability metrics alongside raw compute. The mixed workload profile (seismic imaging plus AI research on shared infrastructure) is also a pattern worth watching for organizations weighing dedicated AI clusters against converged HPC/AI systems.

Watch for the release of detailed technical specifications (node counts, exact GPU models, interconnect topology, and storage configuration), which will clarify how well-suited the system is for large transformer-based training versus tightly coupled simulation workloads; for vendor-published performance benchmarks once the system is commissioned in 2027; and for follow-on disclosures about the campus heat-reuse implementation, which could serve as a reference case for sustainable HPC deployment.