Back to top
Back to top

Sovereign AI Index

Tracking the Global Push for AI Self-Reliance

Explore projects by category and status

The Rise of Sovereign AI

Ninety percent of the computing power needed to develop and deploy frontier AI is in the United States and China.1 Companies in those two countries have developed virtually all of the world’s top-ranked AI foundation models.2 Concerned about this concentration of AI power, governments worldwide have responded with initiatives to strengthen their AI capabilities under the banner of “sovereign AI.”

Although a consensus definition of sovereign AI remains elusive, this index defines a sovereign AI project as a government-backed AI initiative tied explicitly to national strategic interests and backed by material public investment in domestic compute, models, or data ecosystems.3

The drivers of sovereign AI vary widely. For some countries, the imperative is security: protecting sensitive data and ensuring access to advanced capabilities for defense and intelligence. For others, it is the economy: leveraging AI to spur local investment, jobs, productivity, and long-term value. Culture is another driver, with nations seeking AI systems that better reflect local languages and norms. Autonomy also motivates countries that see danger in growing AI dependence on the United States or China. These drivers often overlap.

The result is a surge in sovereign AI activity across the globe.

Sovereign AI Projects Have Accelerated4

Breakdown of Sovereign AI Projects Across the Stack

This index tracks 185 sovereign AI projects by three broad categories that mirror key layers of the AI stack: infrastructure, models, and data.

  • Infrastructure projects, such as AI data centers, supercomputers, graphics processing unit (GPU) clusters, and compute access programs, make up 59 percent of all projects tracked.
  • Model projects, which include government-backed efforts to develop or adapt foundation models, make up 32 percent.
  • Data projects, or initiatives to build national training datasets or data-sharing platforms, make up just 9 percent.
Sovereign AI Projects Concentrate On Compute and Models5

Sovereign AI infrastructure projects are growing in number. Although infrastructure and model projects grew at roughly similar rates through mid-2024, infrastructure projects have accelerated sharply since then. There were more infrastructure projects announced in the first quarter of 2026 than in all of 2024.

Sovereign AI Infrastructure Projects Lead the Way6

Sovereign AI Projects by Region

Governments worldwide have declared AI sovereignty a priority, but rhetoric has not translated evenly into reality. The index reveals a substantial gap between intent and action. Investment remains heavily concentrated in two regions—the Middle East and East Asia—which together account for more than 80 percent of all tracked and publicly disclosed sovereign AI investment worldwide.

Governments worldwide have declared AI sovereignty a priority, but rhetoric has not translated evenly into reality.

The Middle East and East Asia Lead Global Sovereign AI Investment7

Top Sovereign AI Investors

Zooming in from regions to countries sharpens the picture. Within regions, sovereign AI investments are dominated by a few outsized national bets. The 10 largest spenders account for roughly 90 percent of all disclosed sovereign AI investment, and the United Arab Emirates (UAE) and Japan alone account for nearly two-thirds. Outside that top tier, disclosed totals rarely exceed a few hundred million dollars.

The UAE and Japan Lead in Disclosed Sovereign AI Investment8

Who Builds “Sovereign” AI?

In practice, most “sovereign” AI projects rely heavily on foreign—overwhelmingly American—technology providers. These foreign companies remain deeply embedded across more than 170 national infrastructure and model projects tracked by the index. More than three-fifths of tracked projects disclose at least one foreign partner, and four-fifths of these involve a U.S. company.

Since the index’s release in April 2026, projects with no disclosed foreign partner rose from 31 to 37 percent, accounting for roughly half of all new tracked projects since January 2026.

Countries Seeking AI Sovereignty Depend Heavily on American Technology9

U.S. Vendors Are Embedded Across the Sovereign AI Stack

A range of mostly U.S. technology providers underpins the supply chains vital to other countries’ AI ambitions. NVIDIA alone supplies GPUs for 45 percent of all tracked infrastructure projects. U.S.-headquartered companies also appear in the majority of sovereign infrastructure projects across nearly every layer of the stack—from accelerator chips to server systems to cloud platforms.

Foreign technology companies play different roles in sovereign AI infrastructure: accelerator chip designers (NVIDIA, AMD, Intel, Cerebras), server manufacturers (HPE, Dell, Supermicro, Lenovo), cloud and compute platforms (AWS, Oracle, Nebius), and networking providers (Cisco).

This concentration is itself a driver of sovereign AI initiatives. Governments wary of dependence on foreign models and cloud providers are pouring resources into compute that is under their control. But building national data centers to escape reliance on U.S. cloud platforms does not eliminate dependence on American technology. It only shifts exposure from one layer of the U.S. tech stack to another.

In the near term, it is hard to envision any sovereign compute project outside of China wholly independent of the U.S. technology stack.

Building national data centers to escape reliance on U.S. cloud platforms does not eliminate dependence on American technology. It only shifts exposure from one layer of the U.S. tech stack to another.

Sovereign Compute Relies Heavily on U.S. Tech10

Open-Weight Models and Sovereign AI

Only a handful of countries have the compute, data, talent, and capital to develop competitive frontier models. Most countries pursuing model projects instead opt to fine-tune open-weight models on local data to embed national languages, cultural context, and domain-specific knowledge at a fraction of the cost. Many also train sub-frontier models from scratch. As of June 2026, 21 such projects exist—more than double the number of projects at the end of 2024.

At the model layer, most countries conceptualize sovereignty as less about exclusive development than about exercising the ability to run, modify, and govern a model domestically without dependence on a foreign firm’s application programming interface (API) or licensing terms.

Among countries building on open-weight models, Meta’s Llama family is the clear favorite, appearing in 14 of the 25 model projects that disclose their base, followed by Mistral (France), Google (United States), and Alibaba (China). Even at the model layer, realizing sovereign AI ambitions still often relies on U.S. foundations.

Even at the model layer, realizing sovereign AI ambitions still often relies on U.S. foundations.

Most Sovereign AI Base Models are American11

The Sovereign AI Paradox

The path to sovereign AI is full of tensions. Countries across the globe are pursuing sovereign AI, but actual sovereign AI spending remains concentrated in just a few. For all the rhetoric of technology independence, most sovereign AI projects still depend heavily on foreign—mostly U.S.—technology across the stack. For most of the world, the gap between sovereign aspiration and capacity remains vast.

No country—not even the United States—can achieve full control over the complex and varied inputs that power frontier AI systems. Even the most advanced economies, for instance, rely on chips and chipmaking equipment produced overseas. For most economies, the index shows that sovereign AI means managing rather than eliminating dependencies—choosing which layers of the stack matter most and accepting reliance on foreign providers for the rest. Non-U.S. alternatives exist at specific layers—France’s Mistral, China’s open-weight models—but none offer an alternative to the Unites States that is as comprehensive across the stack.

The question facing policymakers worldwide is not whether full sovereign AI is achievable (it is not), but whether they can craft partnerships, deploy technologies, and implement governance frameworks to give countries meaningful agency over how AI is deployed within their borders.

Data is current as of June 2026.

For most economies, the index shows that sovereign AI means managing rather than eliminating dependencies—choosing which layers of the stack matter most and accepting reliance on foreign providers for the rest.

References
  1. Konstantin F. Pilz et al., “The US Hosts the Majority of GPU Cluster Performance, Followed by China,” Epoch AI, June 5, 2025, https://epoch.ai/data-insights/ai-supercomputers-performance-share-by-country ↩︎
  2. “Leaderboard,” LMArena, accessed July 30, 2026, https://lmarena.ai/leaderboard/text. ↩︎
  3. Additional information on definitions and inclusion criteria can be found in the index’s methodology memo. ↩︎
  4. Quarter reflects announcement date. Pablo Chavez, Vivek Chilukuri, and Ruby Scanlon, “Sovereign AI Index,” Center for a New American Security, published April 20, 2026, updated August 18, 2026, https://interactives.cnas.org/reports/sovereign-ai-index ↩︎
  5. Projects spanning multiple categories are counted in each. Percentages reflect share of 185 total projects. Illustrative examples: Infrastructure: Germany’s JUPITER Exascale Supercomputer, Japan’s METI Cloud Program; Models: Switzerland’s Apertus, Chile’s Latam-GPT; Data: EU Common European Language Data Space, India’s AIKosha. Chavez, Chilukuri, and Scanlon, “Sovereign AI Index.” ↩︎
  6. Based on announcement quarter. Projects spanning multiple categories are counted in each. Chavez, Chilukuri, and Scanlon, “Sovereign AI Index.” ↩︎
  7. Figures reflect publicly disclosed sovereign AI investment aggregated at the country and regional level. Disclosed totals encompass a range of funding stages—from announced roadmaps and signaled government intent to signed contracts to appropriated program budgets to operational spending already deployed. Source documents do not consistently distinguish between these stages and totals should be read as an aggregate of publicly disclosed investment across the pipeline rather than as a direct measure of deployed capital. Chavez, Chilukuri, and Scanlon, “Sovereign AI Index.” ↩︎
  8. Figures reflect publicly disclosed sovereign AI investment aggregated at the country and regional level. Disclosed totals encompass a range of funding stages—from announced roadmaps and signaled government intent to signed contracts, to appropriated program budgets, to operational spending already deployed. Source documents do not consistently distinguish between these stages, and totals should be read as an aggregate of publicly disclosed investment across the pipeline rather than as a direct measure of deployed capital. Chavez, Chilukuri, and Scanlon, “Sovereign AI Index.” ↩︎
  9. Covers national infrastructure and model projects only. “U.S. Partner” means at least one American company is involved in the project. Chavez, Chilukuri, and Scanlon, “Sovereign AI Index.” ↩︎
  10. Companies shown play different roles in sovereign AI infrastructure: accelerator chip designers (NVIDIA, AMD, Intel, Cerebras), server and systems original equipment manufacturers (HPE, Dell, Supermicro, Lenovo), cloud platforms (AWS, Oracle, Nebius), networking (Cisco), systems integrators (Eviden), and national high-performance computing coordinators (GENCI). Counts reflect project-level appearances and are comparable within a role but not across roles. A single project may involve multiple partners. Only partners appearing in two-plus projects are shown. Chavez, Chilukuri, and Scanlon, “Sovereign AI Index.” ↩︎
  11. Projects may use multiple base models and are counted under each, including co-developed foundations. Only projects with identified base models are included: 25 sovereign AI model projects out of approximately 63 tracked. The remainder are trained from scratch or use undisclosed bases. Chavez, Chilukuri, and Scanlon, “Sovereign AI Index.” ↩︎

By Pablo Chavez, Vivek Chilukuri, and Ruby Scanlon

Last Updated August 2026

Explore All Projects

Region:
Country:
Category:
Status:
Sort:

  • Infrastructure

    1HealthAI

    Spain (EU)

    Europe In Development

    Infrastructure | System type: European High Performance Computing (EuroHPC) AI Factory (1HealthAI; planned; CESGA AI-specific supercomputer + experimental AI-optimized platform) | Accelerators: Not specified | Interconnect: Not specified | Performance: Not specified | Quantified power/cooling: Not specified

    1HealthAI

    Spain (EU)

    Spain (EU)

    Infrastructure

    In Development (2025)

    Budget

    $97.1 million total

    Access & Licensing

    AI Factory services + access pathways to EuroHPC resources; sector scope includes climate and health interactions, genomics, personalized medicine, sustainable agrifood systems, blue biotechnology, pharmaceuticals, and environmental health

    Purpose

    An EU-selected AI Factory delivering health-focused AI capability across human, animal, and environmental health domains, including a dedicated AI supercomputer, experimental AI platform, and free support services for companies and research centers

    Lead Sponsor

    Ministry of Science, Innovation and Universities | EuroHPC Joint Undertaking (JU) | Xunta de Galicia

    Location

    Centro de Supercomputacion de Galicia (CESGA) | Santiago de Compostela, Galicia, Spain

    Technical Specifications

    Infrastructure | System type: European High Performance Computing (EuroHPC) AI Factory (1HealthAI; planned; CESGA AI-specific supercomputer + experimental AI-optimized platform) | Accelerators: Not specified | Interconnect: Not specified | Performance: Not specified | Quantified power/cooling: Not specified

    Partners

    CESGA (host/operator; project lead) (ES) Consejo Superior de Investigaciones Cientificas (CESGA coowner) (ES) | EuroHPC JU (lead sponsor; funder) (EU) | European Digital Innovation Hub DATAlife (ecosystem partner) (ES) | Gradiant Technology Center (ecosystem partner) (ES) | Ministry of Science, Innovation and Universities (lead sponsor; funder) (ES) | University of A Coruna (ecosystem partner; Galician University System Network of Research Centres [CIGUS] Network) (ES) | University of Santiago de Compostela (ecosystem partner; CIGUS Network) (ES) | University of Vigo (ecosystem partner; CIGUS Network) (ES) | Xunta de Galicia (lead sponsor; funder; CESGA coowner) (ES)

    Sovereign Profile

    “Spain is leading Europe in advanced technological capabilities, democratising access to AI innovation, thanks to the intense work that this government is carrying out to boost the technological independence of Spain and Europe.” —Oscar Lopez, minister for digital transformation and public function, quoted by La Moncloa.

    More Information

  • Model

    A.X K1 (SKT-led Sovereign Foundation Model)

    South Korea

    Asia-Pacific Operational

    Model | Build type: South Korean sovereign foundation model, trained from scratch under MSIT’s Sovereign AI Foundation Model Project (Data Track) | Architecture: Mixture-of-Experts, 519B total parameters/33B active, 61 layers, Multihead Latent Attention, 192 routed experts + 1 shared, 160K vocabulary (BBPE) | Context length: 128K tokens | Primary language: Korean (also English, with Chinese, Japanese, and Spanish vocabulary coverage) | Training data: ~10T-token curated corpus (web, code, STEM, books, parsed Korean PDFs, synthetic data) | Training compute: NVIDIA H200 GPUs (140GB), scaled from 1,024 to 1,536 GPUs, ~73 days of pretraining within a ~4-month project, FP8 precision; GPUs self-procured by SKT | Posttraining: “Think-Fusion” recipe (dual-track SFT, model merging, and on-policy RL) enabling user-controllable thinking/nonthinking modes | Modalities: text only (multimodal stated as future work)

    A.X K1 (SKT-led Sovereign Foundation Model)

    South Korea

    South Korea

    Model

    Operational

    Budget

    Public-private; amount not disclosed. Government-funded under MSIT’s Sovereign AI Foundation Model Project (grants: MSIT/ National IT Industry Promotion Agency (NIPA) PJT-25-080042 and MSIT/ National Information Society Agency (NIA) 2025-AIData-WII43); SKT self-procured the H200 GPUs

    Access & Licensing

    Open weights on Hugging Face (skt/A.X-K1) under the Apache 2.0 license; free for commercial use, modification, and redistribution

    Purpose

    A 519-billion-parameter Korean sovereign foundation model developed by a SK Telecom (SKT)-led consortium under the Ministry of Science and ICT’s (MSIT’s) Sovereign AI Foundation Model Project, designed for Korean linguistic and cultural context

    Lead Sponsor

    MSIT—Sovereign AI Foundation Model Project (program authority and funder, via NIPA and NIA) | SKT(consortium lead and developer)

    Location

    SKT-led consortium/MSIT Sovereign AI Foundation Model Project | South Korea

    Technical Specifications

    Model | Build type: South Korean sovereign foundation model, trained from scratch under MSIT’s Sovereign AI Foundation Model Project (Data Track) | Architecture: Mixture-of-Experts, 519B total parameters/33B active, 61 layers, Multihead Latent Attention, 192 routed experts + 1 shared, 160K vocabulary (BBPE) | Context length: 128K tokens | Primary language: Korean (also English, with Chinese, Japanese, and Spanish vocabulary coverage) | Training data: ~10T-token curated corpus (web, code, STEM, books, parsed Korean PDFs, synthetic data) | Training compute: NVIDIA H200 GPUs (140GB), scaled from 1,024 to 1,536 GPUs, ~73 days of pretraining within a ~4-month project, FP8 precision; GPUs self-procured by SKT | Posttraining: “Think-Fusion” recipe (dual-track SFT, model merging, and on-policy RL) enabling user-controllable thinking/nonthinking modes | Modalities: text only (multimodal stated as future work)

    Partners

    SKT (consortium lead/developer) (KR) Krafton (consortium partner; codeveloped the Think-Fusion SFT recipe) (KR) | 42dot (consortium partner) (KR) | Rebellions (AI chip consortium partner) (KR) | Liner (consortium partner) (KR) | SelectStar (data consortium partner) (KR) | Seoul National University (consortium partner) (KR) | Korea Advanced Institute of Science and Technology (consortium partner) (KR) | MSIT (program sponsor/funder) (KR) | NIPA (grant agency) (KR) | NIA (grant agency/national benchmark evaluator) (KR)

    Sovereign Profile

    Developed under MSIT’s government-funded Sovereign AI Foundation Model Project, explicitly framed around South Korean foundation models built from scratch with open-source commercial usability, in service of South Korea’s goal of becoming a top three AI power. The technical report states the model “represents a strategic effort to establish a robust Sovereign AI ecosystem” and to “reduce reliance on foreign proprietary models” by deeply modeling South Korea’s linguistic and cultural characteristics for local industry, government, and academia

    More Information

  • Infrastructure

    Abu Dhabi Sovereign AI Cloud

    United Arab Emirates

    Middle East Operational

    Infrastructure | System type: OCI Supercluster deployment in the Oracle Cloud Abu Dhabi Region | Accelerators: More than 4,000 x NVIDIA Blackwell GPUs | Interconnect: Not specified | Performance: Not specified | Quantified power/cooling: Not specified

    Abu Dhabi Sovereign AI Cloud

    United Arab Emirates

    United Arab Emirates

    Infrastructure

    Operational

    Budget

    $3.54 billion

    Access & Licensing

    Cloud service access (OCI); specific public sector/regulatory use cases referenced; detailed access model not specified

    Purpose

    Sovereign AI training and inference capacity in Abu Dhabi via an Oracle Cloud supercluster, positioned to support government and regulated-industry AI workloads while meeting data sovereignty requirements

    Lead Sponsor

    Department of Government Enablement (DGE)

    Location

    Oracle Cloud Infrastructure (OCI) Dedicated Region/Oracle Cloud Abu Dhabi Region | Abu Dhabi, United Arab Emirates (in-country sovereign boundaries; Core42/G42 infrastructure referenced)

    Technical Specifications

    Infrastructure | System type: OCI Supercluster deployment in the Oracle Cloud Abu Dhabi Region | Accelerators: More than 4,000 x NVIDIA Blackwell GPUs | Interconnect: Not specified | Performance: Not specified | Quantified power/cooling: Not specified

    Partners

    DGE (lead sponsor; government owner) (AE) OCI (cloud provider/operator) (U.S.) | Core42/G42 (infrastructure foundation) (AE) | Deloitte (implementation/orchestration) (U.S.) | NVIDIA (accelerator platform) (U.S.)

    Sovereign Profile

    “OCI’s sovereign AI infrastructure will directly support Abu Dhabi’s goals of becoming the world’s first fully AI-native government by 2027.” —Oracle.

    “. . . ensuring their sensitive government data does not leave the emirate . . .” —NVIDIA blog, quoting Deloitte.

    More Information

  • Infrastructure

    Adastra2 (MI300A partition at CINES)

    France (EU)

    Europe Operational

    Infrastructure | Hewlett Packard Enterprise (HPE) Cray EX255a | Accelerators: 112 x AMD Instinct MI300A APUs | Interconnect: Slingshot-11 | Performance: HPL (FP64) Rmax: 2.53 PFLOP/s; Rpeak: 4.06 PFLOP/s | Quantified power/cooling: Power: 36.60 kW; Cooling: 100% fanless direct liquid cooling (97% of heat cooled via warm water); PUE: Not specified

    Adastra2 (MI300A partition at CINES)

    France (EU)

    France (EU)

    Infrastructure

    Operational (2024)

    Budget

    EUR 25 million ($26.45 million)

    Access & Licensing

    Open to academic and industry researchers via France's national computing allocation framework

    Purpose

    A new AMD-based computing partition integrated into France's national Adastra supercomputer, combining traditional high-performance computing with AI capability

    Lead Sponsor

    Grand Equipement National de Calcul Intensif (GENCI) | CINES

    Location

    Centre Informatique National de l'Enseignement Superieur (CINES) | Montpellier, Herault, France

    Technical Specifications

    Infrastructure | Hewlett Packard Enterprise (HPE) Cray EX255a | Accelerators: 112 x AMD Instinct MI300A APUs | Interconnect: Slingshot-11 | Performance: HPL (FP64) Rmax: 2.53 PFLOP/s; Rpeak: 4.06 PFLOP/s | Quantified power/cooling: Power: 36.60 kW; Cooling: 100% fanless direct liquid cooling (97% of heat cooled via warm water); PUE: Not specified

    Partners

    GENCI (owner/coordinator) (FR) CINES (host/operator) (FR) | AMD (hardware provider) (U.S.) | HPE (system vendor) (U.S.)

    Sovereign Profile

    “France continues to strengthen its strategic autonomy with three major supercomputers: Jean Zay, Adastra, and Alice Recoque.” —Elysee document (Make France an AI Powerhouse).

    More Information

  • Infrastructure + access program

    African Compute Initiative at the University of Cape Town

    South Africa

    Africa In Development

    Infrastructure | Modern GPU servers; accelerator type and count not disclosed | Multipetabyte secure storage | High-speed networking | OpenStack and Ceph environment | Approximately 180 kWp solar contribution | Targets: About 100 users in year one and 300 users across at least five institutions by year three

    African Compute Initiative at the University of Cape Town

    South Africa

    South Africa

    Infrastructure + access program

    In Development

    Budget

    ZAR 71 million (~$4.2 million) infrastructure grant reported by UCT Council. The initiative also sits within the broader GBP 58 million ($78.1 million) AI4D partnership, but that full partnership amount is not the project budget.

    Access & Licensing

    Shared higher education and research access, initially through UCT and expanding to African partner institutions; detailed allocation and pricing rules are not yet published

    Purpose

    A University of Cape Town (UCT)-led shared AI compute and data infrastructure initiative intended to give African universities and researchers locally governed capacity for model training, fine-tuning, inference, and data-intensive research

    Lead Sponsor

    University of Cape Town

    Location

    UCT High Performance Computing Data Centre | Cape Town, Western Cape, South Africa

    Technical Specifications

    Infrastructure | Modern GPU servers; accelerator type and count not disclosed | Multipetabyte secure storage | High-speed networking | OpenStack and Ceph environment | Approximately 180 kWp solar contribution | Targets: About 100 users in year one and 300 users across at least five institutions by year three

    Partners

    Foreign, Commonwealth & Development Office (public funder) (GB) International Development Research Centre (public funder) (CA) | AI4D partnership (program framework) (INT) | UCT and participating African universities (hosts/users) (ZA/AFR)

    Sovereign Profile

    “African researchers have the ideas and the talent, but they have been held back by a lack of access to the computing power that AI development demands. The African Compute Initiative changes that. It means researchers and students across Africa can work at the frontier of AI, not just consume what is built elsewhere.” —Associate Professor Jonathan Shock, Interim Director of the UCT AI Initiative.

    More Information

  • Infrastructure

    AGH Cyfronet Helios

    Poland (EU)

    Europe Operational

    Infrastructure | System type: Hewlett Packard Enterprise (HPE) Cray EX254n (Helios GPU partition; Cyfronet) | Accelerators: 440 x NVIDIA Grace Hopper GH200 Superchips + 24 x NVIDIA H100 GPUs | Interconnect: Slingshot-11 | Performance: HPL (FP64) Rmax: 19.14 PFLOP/s; Rpeak: 30.44 PFLOP/s | Quantified power/cooling: Power: 316.88 kW; Cooling: Not specified; PUE: Not specified

    AGH Cyfronet Helios

    Poland (EU)

    Poland (EU)

    Infrastructure

    Operational (2024)

    Budget

    Not specified

    Access & Licensing

    Research access via ACC Cyfronet AGH allocation processes (details not specified)

    Purpose

    A national high-performance computing system supporting large-scale scientific simulation and AI workloads for Poland's research community

    Lead Sponsor

    Ministry of Development Funds and Regional Policy | European Union

    Location

    Academic Computer Centre (ACC) Cyfronet, AGH University of Krakow | Krakow, Lesser Poland Voivodeship, Poland

    Technical Specifications

    Infrastructure | System type: Hewlett Packard Enterprise (HPE) Cray EX254n (Helios GPU partition; Cyfronet) | Accelerators: 440 x NVIDIA Grace Hopper GH200 Superchips + 24 x NVIDIA H100 GPUs | Interconnect: Slingshot-11 | Performance: HPL (FP64) Rmax: 19.14 PFLOP/s; Rpeak: 30.44 PFLOP/s | Quantified power/cooling: Power: 316.88 kW; Cooling: Not specified; PUE: Not specified

    Partners

    Ministry of Development Funds and Regional Policy (lead sponsor) (PL) European Union (lead sponsor) (EU) | ACC Cyfronet AGH (operator) (PL) | AGH University of Krakow (host institution) (PL) | HPE (system vendor) (U.S.) | NVIDIA (GPU platform supplier) (U.S.) | AMD (CPU platform supplier) (U.S.)

    Sovereign Profile

    “Wspierajac rozwoj AI na rodzimych zasobach, dbamy o cyfrowa suwerennosc oraz budujemy fundament pod inteligentne uslugi publiczne i biznesowe [By supporting AI development on domestic resources, we safeguard digital sovereignty and build a foundation for intelligent public and business services].” —EuroCC (European High Performance Computing National Competence Center) Poland, post about Polish Large Language Model training moving to Helios at Cyfronet AGH).

    More Information

Authors

  • Pablo Chavez

    Adjunct Senior Fellow, Technology and National Security

    Pablo Chavez is an Adjunct Senior Fellow with CNAS's Technology and National Security Program. His research focuses on the national security, economic, and geopolitical implications of cloud and AI policy.

  • Vivek Chilukuri

    Senior Fellow and Director, Technology and National Security Program

    Vivek Chilukuri is senior fellow and program director of the Technology and National Security Program at the Center for a New American Security (CNAS), where he studies how artificial intelligence (AI) and emerging technologies are reshaping national security and great power competition.

  • Ruby Scanlon

    Research Associate, Technology and National Security Program

    Ruby Scanlon is a research associate for the Technology and National Security Program at the Center for a New American Security (CNAS), supporting the Center’s research on US-China technology competition, China's innovation ecosystem, and artificial intelligence (AI) policy.

More On This Report

This project was made possible with general support for the CNAS Technology and National Security Program, along with corporate support for the index.

As a research and policy institution committed to the highest standards of organizational, intellectual, and personal integrity, CNAS maintains strict intellectual independence and sole editorial direction and control over its ideas, projects, publications, events, and other research activities. CNAS does not take institutional positions on policy issues, and the content of CNAS publications reflects the views of their authors alone. In keeping with its mission and values, CNAS does not engage in lobbying activity and complies fully with all applicable federal, state, and local laws. CNAS will not engage in any representational activities or advocacy on behalf of any entities or interests and, to the extent that the Center accepts funding from non-U.S. sources, its activities will be limited to bona fide scholastic, academic, and research-related activities, consistent with applicable federal law. The Center publicly acknowledges on its website annually all donors who contribute.