My watchlist
Croatia · Computing infrastructure

BURA Supercomputer, Rijeka

Historical research installation; current production unverified

BURA is a documented research installation at the University of Rijeka. Atos reported performance above initial expectations, but the claim is historical and lacks a published baseline suitable for an independent production-forecast comparison.

Published and evidence reviewed November 9, 2026.

Compare this project · Evidence checkpoints · FJ Intelligence

Investment takeaway

BURA is useful as a historical example of a supplier claiming a better result than initially expected. It does not establish that today’s AI facilities using the same brand will outperform. The baseline, energy boundary and operating period would need to be recovered before calculating an independently verified output beat.

Project at a glance

Location
Rijeka, Croatia
Status
Historical research installation; current production unverified
Capacity scope
Historical 288-node CPU cluster with 6,912 physical cores; separate GPU and large-memory partitions. Current available capacity is not independently established.
Named participants
University of Rijeka; Atos / Bull; SmartIS; MEP; Intel
Map precision
Approximate city/locality marker, not a surveyed facility, building or parcel.

FJI source-qualified: four-document gate met, including project-specific government evidence. Qualification evidence and decisions →

What is documented

The University of Rijeka describes a 288-node CPU cluster, separate GPU nodes and large-memory systems. Croatia’s government export portal recorded the installation’s appearance in the 2015 TOP500 and HPCG rankings. Independent Hina reporting documented the February 2016 full-operation inauguration and EU funding. [1, 2, 3]

The performance claim

An Atos-authored case study says BURA delivered 16% more performance and used 56% less energy than initially expected. It names Atos/Bull, SmartIS and MEP in the system and facility delivery. The document does not publish the original baseline values, full measurement method or sustained production series. These percentages remain supplier-reported historical claims. [4]

Keep computing and energy separate

A faster benchmark, lower cooling overhead and lower total electricity are different measurements. The case study does not establish a common boundary allowing its performance and energy percentages to be combined into a campus-wide productivity multiplier. No current operating load is derived from them.

Commercial relevance and limits

The documented beneficiary is a university research service. There is no reviewed project-level contract allocating extra production, nor a disclosed profit uplift to suppliers. The profile describes participants at the time of delivery; subsequent corporate transactions do not establish current supplier ownership or recurring BURA revenue.

Current status

An accessible institutional hardware description supports the identity of the installation. It is not live metering or independent confirmation of current production. This record is therefore labeled as an existing facility with current readiness unverified, rather than as a newly commissioned AI campus.

What remains unverified

Original baseline, complete test method, sustained output series, current service availability and project economics remain unverified. Historical brand ownership is not treated as current ownership.

Next evidence checkpoint

Recover the original acceptance baseline and test method; verify current configuration and service availability before counting current production.

Related project research

Sources and claim boundaries

  1. University of Rijeka

    Institutional description of CPU, GPU and large-memory partitions.

    Computing resources · Document date: Not established · Read November 9, 2026.

  2. Croatian government export portal

    Historical installation identity and benchmark rankings; government-hosted media account.

    Article body · Document date: 2015-11-30 · Read November 9, 2026.

  3. Hina / tportal

    Independent account of full-operation inauguration and EU-funded acquisition.

    Article body · Document date: 2016-02-26 · Read November 9, 2026.

  4. Atos case study

    Supplier-authored claim of 16% more performance and 56% less energy; hosted by FeaturedCustomers.

    The results; solution sections · Document date: Not established · Read November 9, 2026.

AI-assisted public-source research by Fjord Journal. No site visit, private contract review or independent metering. Institutional, supplier and benchmark claims retain their stated scope.

Company investor information & quarterly statements

Financial-disclosure review remains incomplete for one or more participants. The company-level limitations below remain unresolved; this section does not establish full corporate-disclosure verification.

Corporate disclosures provide company-wide context. They do not establish this project’s revenue, spending, profit, ownership share or operating capacity. Disclosure verification records →