November 22, 2026

Dell Overtakes HPE in Shifting HPC and AI-Sci Market as Global Spending Eyes $135 Billion Milestone

The annual rhythm of the supercomputing community is defined by two major milestones: the Supercomputing conference in the United States and the International Supercomputing conference in Europe. For those embedded in the industry, these events are more than just showcases for the latest silicon and interconnects; they are endurance tests of networking and knowledge. One of the most storied traditions of these gatherings is the early Tuesday morning trek to the breakfast briefing hosted by Hyperion Research, formerly the IDC high-performance computing group. It is a ritual where attendees shake off the social fatigue of the previous evening, fueled by strong coffee and a traditional breakfast, to absorb the most authoritative data on the state of the High-Performance Computing (HPC) and AI for Science (AI-Sci) markets.

While physical attendance at this year’s SC26 and ISC26 was not possible for everyone, the data presented by industry stalwarts Earl Joseph, Bob Sorensen, and Mark Nossokoff provides a comprehensive window into the trajectory of the market through the first half of 2026. The findings reveal a sector in the midst of a profound transformation, where traditional simulation and modeling are increasingly intertwined with—and sometimes overshadowed by—the massive infrastructure requirements of modern artificial intelligence. The numbers suggest a market that is not only expanding in terms of dollar value but is also undergoing a significant realignment of vendor leadership and a strategic shift in how compute capacity is consumed.

Casing The HPC And AI-Sci Market With Hyperion Research

According to the Hyperion Research data, the first half of 2026 was a period of robust financial growth for on-premises HPC and AI-Sci systems. Total revenues for this period reached $16.48 billion, representing a significant 19.9 percent increase compared to the same period in 2025. However, this double-digit revenue growth masks a more complex reality regarding the volume of hardware actually moving into data centers. While the money flowing into the sector surged, worldwide unit shipments of these systems only grew by a modest 4.4 percent. This stark divergence between revenue and units highlights a fundamental change in the composition and cost of modern supercomputing infrastructure.

Several factors contribute to this gap. Modern systems are increasingly dense, packing more nodes into each rack and a higher number of CPUs and accelerators into each node. Furthermore, the industry has been grappling with sustained price increases for critical components. The cost of high-end CPUs, sophisticated GPUs, and specialized memory like HBM (High Bandwidth Memory), along with flash storage, has risen steadily. In many ways, it is surprising that the gap between revenue and units isn’t even wider. The data suggests that many HPC centers are adopting a more conservative procurement strategy, perhaps hanging back one or two generations on hardware to manage capital outlays. This trend is visible in recent Top500 rankings, where many new installations feature NVIDIA Grace-Hopper or Intel Xeon-Hopper configurations, despite the fact that the newer Blackwell architectures have been available for some time. For many scientific research centers, the most cutting-edge GPUs remain both too scarce and too expensive to deploy at scale immediately.

Perhaps the most striking revelation from the mid-2026 data is the shift in vendor leadership. For years, Hewlett Packard Enterprise (HPE) has maintained a firm grip on the top spot for on-premises HPC and AI-Sci installations, a position fortified by its strategic acquisitions of SGI and Cray. However, the first half of 2026 saw Dell emerge as the revenue leader in this space. Dell’s technical computing revenues skyrocketed by 67.7 percent to reach $3.93 billion, while HPE saw a much flatter growth rate of just 1.1 percent, bringing in $3.44 billion.