Artificial intelligence has reached a critical strategic inflection point. For the past several years, the prevailing enterprise narrative dictated that cutting-edge AI belonged naturally in the public cloud. However, new data reveals a massive shift in how the world’s most critical infrastructure sectors are approaching accelerated computing. The era of blind cloud dependency is giving way to a new paradigm: the sovereign AI factory.

According to Intersect360 Research, national sovereign AI data centers have become the fastest-growing segment of accelerated computing infrastructure globally. Spending exploded from $1.6 billion in 2023 to $7.2 billion in 2024—a staggering 349% year-over-year increase that outpaces hyperscale AI and AI-focused cloud providers alike. Looking ahead, this market is projected to reach $122.8 billion by 2026, driven by a 90.4% compound annual growth rate (CAGR).

For C-level executives, this trend is not just about expanding server footprints. It represents a fundamental structural realignment of data sovereignty, national security, and strategic autonomy.

The convergence of HPC and AI

Drawing on market data and insights from the HPC-AI Leadership Organization (HALO) uncovers a crucial architectural truth: sovereign AI installations are fundamentally high-performance computing (HPC) centers operating under intense AI pressure.

Rather than building isolated silos or throwing data into standard cloud instances, organizations are consolidating their infrastructure. Intersect360 Research found that 62.5% of surveyed organizations run their traditional HPC and AI workloads on the exact same systems. Similarly, 67.1% leverage the same shared storage layer for both workloads.

This shared infrastructure model is born out of necessity. Traditional scientific modeling and simulation continue to drive baseline technical requirements, while AI training workloads are layered directly on top of them. This reality favors a technology stack that builds upon proven, highly optimized HPC designs rather than purely cloud-native or enterprise-only architectures.

The strict requirement for on-premises control

For sovereign AI initiatives—such as Japan’s AI Bridging Cloud Infrastructure or the G42 cloud in the UAE—keeping critical workloads off the public cloud is a mandatory constraint. The data confirms that 56.3% of sites keep active HPC-AI compute entirely on-premises, and 68.2% do the same for their storage.

When asked to rate the importance of system features for their next infrastructure purchase, buyers rated cloud integration lowest, scoring it a mere 2.25 out of 5. In fact, 36% of respondents classified cloud integration as “not important at all.”

Because data and workloads cannot be offloaded to external providers due to regulatory, national jurisdiction, and security boundaries, on-premises systems must be engineered for flawless execution without a cloud safety net.

Uncovering the industry’s technical bottlenecks

The true value of the data lies in its “satisfaction gap” analysis, highlighting exactly where current market offerings fail to meet the rigorous demands of sovereign environments.

Sovereign AI buyers are optimizing their configurations around fast data movement, not just raw processing numbers (or peak FLOPS). Buyers placed the highest importance on memory or system bandwidth (scoring 4.57 out of 5) and GPU performance (4.53). However, memory bandwidth also yielded the largest satisfaction gap in the industry at 0.86. If data cannot move to the processors fast enough, training speeds stall, introducing severe strategic bottlenecks.

Similar infrastructure pain points appear across fabrics and storage:

• Networking and interconnects: Interconnect topology is viewed as highly strategic. While architectures have largely standardized on InfiniBand (76.5% as primary) and high-speed Ethernet (68.1% as secondary) scaling from 100 to 400Gbps, users still report significant industry shortfalls in latency, congestion management, and container orchestration.
• Storage hurdles: Data storage is viewed as strategically vital by 93% of respondents. Parallel file systems (81.2% importance) and tight data locality (83.5% importance) function as core sovereignty enforcement mechanisms. Yet, the industry’s single largest satisfaction gap sits at a glaring 0.98 for storage performance when handling millions of small files—a hallmark of AI training data sets.

To resolve these gaps, market leaders like HPE, utilizing high-bandwidth AMD CPUs and GPUs, are designing configurable, memory-rich on-premises architectures. These systems are purpose-built to balance both large checkpoint-style HPC files and small AI objects within a unified sovereign boundary.

3 practical tips for the C-suite

As your organization navigates the shift toward localized, high-performance AI deployment, keep these data-driven principles in mind:

1. Treat storage locality as a security strategy: Do not view storage as a generic commodity. Leverage parallel file systems to maintain strict physical control over your training data and model weights, ensuring they remain firmly under your legal jurisdiction.
2. Prioritize node bandwidth over raw FLOPS: When evaluating vendor hardware, optimize your system configurations around memory bandwidth and fabric predictability rather than simple processor speeds. Bandwidth is the ultimate governor of real-world AI training performance.
3. Seek co-design partners, not product vendors: Sovereign AI requires balanced, highly specialized, and supportable architecture. Partner with infrastructure ecosystems—like HPE and AMD—that can co-design flexible compute nodes, high-speed networking, and custom services tailored to mixed HPC-AI environments.

Ready to architect your own sovereign AI strategy? Read the full Intersect360 Research report to explore the complete data set and blueprint for the future of on-premises sovereign AI factories.

*************

HPE and AMD are partnering to deliver the next generation of AI and supercomputing infrastructure. Combining HPE’s deep expertise in large-scale system design with AMD EPYC™ processors and AMD Instinct™ accelerators, the two companies help customers build platforms capable of supporting the most demanding AI, modeling, simulation, and analytics workloads. Whether accelerating scientific research, enabling digital engineering, or training advanced AI models, HPE and AMD focus on what matters most: faster insights, greater efficiency, and the flexibility to scale. Together, they provide the technology foundation organizations need to solve complex problems, drive discovery, and create lasting business and societal impact.