Blog

8 trends shaping the future of high-performance computing in 2026

Written by Alicia Hagan | Wed, Jul 8, 2026

High-performance computing (HPC) is changing faster than its benchmarks can measure.  

The industry was driven primarily by hardware progress and raw compute. It is now becoming a complex ecosystem where multiple forces intersect: artificial intelligence, heterogeneous architectures, sustainability constraints, evolving software stacks and increasingly distributed innovation models.

This convergence is changing how systems are built, used and orchestrated. As workloads become more diverse and interconnected, HPC is shifting toward a more integrated paradigm where compute, data and AI capabilities must operate seamlessly together.

The following eight trends, drawn from discussions at ISC High Performance Conference 2026, illustrate how this transformation is redefining the foundations of high-performance computing. 

Summary:

  1. HPC is moving from performance to impact

  2. Heterogeneous computing is becoming the new standard

  3. AI augmented computing will complement HPC

  4. Hybrid infrastructures are paving the way for quantum computing

  5. Energy effiency is becoming a design principle

  6. Software, co-design and collaboration are becoming strategic foundations

  7. AI factories are redefining HPC infrastructure

  8. Trust and sovereignty are becoming critical enablers

 

1. HPC is moving from performance to impact

For decades, HPC progress meant one thing: building increasingly powerful systems. The industry organised itself around that objective through hardware roadmaps, benchmarks and procurement decisions.

That logic is changing. In ISC26 closing keynote, Jack Dongarra, Research Professor Emeritus, University of Tennessee, Oak Ridge National Laboratory, shared that the world's top supercomputers routinely achieve less than 1% of their theoretical peak performance in practice. The reason is that processors spend most of their time waiting for data to move between memory, storage and compute. If a machine running at full theoretical speed still delivers less than 1% of that in practice, peak FLOPS is not a perfect measure. This is why benchmarks such as high performance conjugate gradients (HPCG) have been created as they better reflect the memory and data movement challenges of real applications. 

What matters is what the system actually produces: a validated simulation, a trained model, a scientific result that can be trusted and reproduced.

Raw performance remains important, but it is no longer sufficient to evaluate success. The metrics that matter now are around time-to-solution, energy per trusted result, workflow efficiency and reproducibility.

The question is shifting from “How powerful is this system?” to “What does it enable?”.  This shift also redirects attention from isolated systems to full workflows, where simulation, data and AI must be orchestrated end to end.

Bull's perspective

  • The shift from FLOPS to outcomes is precisely why Bull has historically prioritised system-level efficiency over raw benchmark performance.

  • Our advantage lies in end-to-end system ownership: when we control the full stack from interconnect to application layer, we can optimise for actual scientific output, not just peak capability.

 

2. Heterogeneous computing is becoming the new standard

HPC progress used to follow Moore’s LawMoore's Law is the observation made by Gordon Moore in 1965 that the number of transistors on a microchip would roughly double every two years, enabling a steady increase in computing performance. While it has driven decades of progress in computing, physical, energy and economic constraints have significantly slowed this pace in recent years., with each generation delivering predictable performance gains. As these gains slow, the industry is moving toward heterogeneous architectures, combining multiple computing architectures, each optimised for different types of workloads.

CPUs and GPUs remain central, but they are now complemented by AI accelerators AI accelerators are specialised hardware designed to efficiently run AI workloads such as training and inference, optimised for parallel processing. , quantum processors, and other domain-specific technologies. Rather than replacing each other, these architectures are designed to coexist and specialise. 

As these architectures converge, software becomes the critical enabler. The main bottleneck is data movement, scheduling and dependency management across distributed environments.

Orchestrating classical and quantum resources, ensuring interoperability and supporting hybrid workflows will be key to unlocking the next generation of HPC systems.

Bull's perspective

  • We have engaged early partnerships with EU accelerator makers (VSORA, Axelera.AI, SemiDynamics).

  • Our ambition is to become pivotal for heterogeneous systems.

 

3. AI augmented computing will complement HPC

AI is reshaping HPC by extending its simulation capabilities.

Hybrid workflows are becoming increasingly common, where physics-based simulations are combined with AI models. AI can accelerate calculations, explore large design spaces and reduce computation time by learning from simulation data. AI acts as an assistant to simplify and accelerate specific stages of the computation process. This approach is already transforming fields such as scientific research, climate modelling, drug discovery, advanced manufacturing and digital twins.

Rather than competing, AI and HPC are becoming interdependent. Modern infrastructures must now support simulation, AI training, inference and analytics within a unified environment.


4. Hybrid infrastructures are paving the way for quantum computing

The conversation about quantum computing has evolved.

Rather than waiting for fault-tolerant quantum computers, organisations are already preparing for this quantum era through hybrid infrastructures that combine classical HPC, AI and quantum technologies.

The focus is also about developing relevant use cases designed to accelerate specific workloads such as optimisation, chemistry and materials simulation. 

Bull's perspective

  • Our quantum investment strategy was initiated 10 years ago. We are working with partners such as Equal1 building a hybrid quantum – HPC compute node. This hybrid approach is accelerating innovation while ensuring organisations are ready to leverage quantum computing as the technology matures.

  • Quantum emulation also plays a key role, enabling researchers and developers to design, validate and optimise quantum algorithms on existing HPC systems before running them on quantum hardware.

 

5. Energy effiency is becoming a design principle

As systems scale toward exascale, energy consumption, cooling and infrastructure costs have become limiting factors. Performance alone is no longer viable if it comes with unsustainable energy use.

Energy efficiency is now a key element into HPC architecture and software decisions. Software is increasingly optimised for energy-to-solution, while hardware design focuses on thermal efficiency, power management and data movement reduction.

This shift is reinforced by benchmarking frameworks such as the Green500. In recent rankings, Bull-built systems have consistently demonstrated leadership, securing the top position for five consecutive editions and occupying the entire podium in 2 successive rankings.

Bull's perspective

  • Bull took the turn toward energy efficiency back in 2007 with the objective of reaching a PUE Power Usage Effectiveness (PUE) is a standard metric for assessing data center energy efficiency. It compares the total energy consumed by the facility to the energy used by its IT equipment. A PUE value closer to 1 indicates a highly efficient infrastructure, where most of the energy consumed is used directly for computing. close to 1.

  • Hardware and software solutions make the direct liquid cooling (DLC) Direct Liquid Cooling (DLC) is a technology that cools servers by circulating liquid directly to the hottest components, enabling heat to be captured and dissipated far more efficiently than with conventional air cooling, offering the most efficient on the market.


6. Software, co-design and collaboration are becoming strategic foundations

The complexity of HPC systems (heterogeneous hardware, hybrid workflow, quantum integration, software…) requires collaboration across a broad ecosystem of actors: researchers, industry, software developers, hardware manufacturers and public initiatives. 

Indeed, progress in HPC increasingly relies on a tight integration between hardware and software through co-design, where architectures and applications are developed in close alignment to maximise efficiency and scalability.

Shared development models, including open-source contributions and large-scale programs such as EuroHPC Joint Undertaking, reflect this shift toward collective innovation, where complexity can only be addressed through coordinated effort.

 

Bull's perspective

  • Europe's competitiveness in HPC will not be built by a single organisation. Success depends on the ability to federate a diverse ecosystem of research institutions, industrial partners, technology providers and public initiatives around shared ambitions.

  • Having contributed to major European programmes for many years, such as EPI and, MontBlanc, or with partners such as EuroHPC JU, CEA and GENCI, Bull believes that open-source contributions and collaboration are essential to accelerate innovation and strategic autonomy.

 

7. AI factories are redefining HPC infrastructure

AI factories are emerging as integrated ecosystems that combine compute, data, AI capabilities and workflow orchestration into unified environments.

Rather than acting as traditional data centres, they operate as continuous production systems for AI, enabling large-scale training, fine-tuning and deployment of models.

In Europe, initiatives such as EuroHPC Joint Undertaking are positioning these platforms as strategic assets for competitiveness and sovereignty across sectors such as healthcare, energy and manufacturing.

Within this landscape, Bull has been involved in several AI-Gigafactory initiatives such as AION consortium contributing to the development of integrated HPC-AI platforms for large-scale model training and deployment.

More broadly, AI factories reflect a shift from infrastructure as a resource to infrastructure as a strategic innovation platform.

Bull's perspective

  • At Bull, we see AI Factories as the foundation for scaling AI, combining the right technology, skills and operational framework to transform experiments into production-ready business outcomes.

 

8. Trust and sovereignty are becoming critical enablers

The fishbowl panel at ISC26, titled "What is truth, anyway?", addressed the fact that outputs of AI and simulation systems are not inherently trustworthy, and the gap between what systems produce and what users can verify is widening.

The panel, led by Addison Snell of Intersect360 Research and Martin Schulz of TU Munich, covered AI's probabilistic outputs and their limitations in scientific contexts, the difficulty of communicating uncertainty to non-specialist audiences, quantum computing's inference-based nature (phenomena are inferred, not directly observed), and the tension between open scientific collaboration and national sovereignty over foundational technologies.

Trust is built through validation, explainability, reproducibility and accessibility across complex workflows that combine simulation, AI and data-driven methods for both scientific validity and industrial adoption.

These developments are also tied to questions of digital sovereignty. High-performance computing infrastructures are strategic assets for nations and regions, making control over data, systems and compute capabilities a key dimension of competitiveness and resilience.

Conclusion

The trends shaping the future of HPC all point in the same direction: success will no longer be defined solely by computational performance, but by the ability to deliver meaningful scientific and industrial outcomes.

HPC is becoming an integrated ecosystem where compute, AI, quantum, data and software must work together across scales. The challenge is now to design coherent systems capable of delivering end-to-end value.

Within this transformation, Bull’s positioning reflects a broader European contribution to HPC innovation, spanning energy-efficient architectures, co-designed systems, AI factory initiatives, quantum / HPC hybridisation and emulation and participation in collaborative frameworks such as EuroHPC.

Ultimately, the future of HPC will belong to systems and ecosystems capable of turning computational power into trusted, usable and sustained impact.