Quantum computing is still very green—a bet, not a business.
FSAS ganó contrato con Telefónica para desarrollar primera de 17 factorías de IA financiadas por UE, ampliando superordenador MareNostrum 5 en Barcelona. IA es tecnología rentable con mercado establecido, mientras computación cuántica sigue siendo inversión prospectiva sin retorno monetario inmediato, limitada a sector público.
- FSAS and Telefónica won contract for first of 17 EU-funded AI factories at Barcelona Supercomputing Center
- MareNostrum 5 currently ranks 14th globally among supercomputers
- Quantum computing has zero monetary return; only public institutions can fund it
- Liquid cooling remains rare and expensive despite being essential for AI infrastructure
Fujitsu España (FSAS) se posiciona como líder en infraestructura de supercomputación, IA y tecnología cuántica, ganando licitación europea para primera factoría de IA en Barcelona Supercomputing Center.
Artificial intelligence is everywhere now—even the subway conversations have shifted toward machine learning and neural networks. Quantum computing, by contrast, remains the province of specialists and theoretical physicists. The public imagination struggles with it: something about cats being alive and dead simultaneously, a paradox that refuses to resolve into common sense. Yet a persistent misconception has taken hold: that quantum computing is simply the next phase of AI, arriving soon to reshape how ordinary people work and live.
Gonzalo Romeo, the general director of FSAS Technologies Spain (the Spanish subsidiary of Japanese tech giant Fujitsu), has spent considerable time trying to untangle this confusion. His distinction is sharp and unsentimental. Artificial intelligence, he explains, already works. It has a functioning market. Companies invest in it because it returns money—though not everyone approves of how or where those returns accumulate. Quantum computing, by contrast, remains a speculative investment. Its monetary return is currently zero. Only public institutions can afford to fund it as pure research, as a bet on a future that may or may not arrive.
The relationship between supercomputing, artificial intelligence, and quantum technology is not a simple sequence, one leading inevitably to the next. Fujitsu created FSAS in 2024 and began expanding into Europe a year later, starting with Spain. The company had already built a track record selling supercomputers to Spanish institutions, and that experience became the foundation for a new strategy: FSAS would concentrate its resources on what the industry calls critical data center environments—designing and assembling the infrastructure that integrates supercomputing, artificial intelligence, and quantum systems, wrapped in hybrid cloud solutions that blend remote and local processing.
The conversation began with a contract award. A consortium led by FSAS (holding 51 percent) and partnered with Telefónica (49 percent) won the right to develop the first of seventeen AI factories that the European Union is promoting and financing. The facility will operate within the Barcelona Supercomputing Center and will partly expand the physical footprint of MareNostrum 5, a supercomputer currently ranked 14th globally. Building such an infrastructure demands choices that are far from obvious. Liquid cooling—essential for managing heat—remains rare and expensive. Storage systems and network architecture must be selected with extreme care, often requiring equipment from multiple vendors. The Barcelona center has become, over the years, a living museum of competing technologies.
The European project incorporates servers from Supermicro, Nvidia chips and software, storage equipment from IBM, and other components—but notably no hardware from Fujitsu itself. Romeo explains that FSAS has a different role to play. The company sees itself as part of a broader movement in which supercomputing centers are evolving toward mixed environments where artificial intelligence carries substantial weight. But this evolution demands diverse architectures, not a single unified approach. Quantum computing fits into this landscape as a separate track. The Barcelona center has already acquired a quantum machine from Quilimanjaro, which it plans to couple with MareNostrum 5.
A parallel effort is underway at Cesga, the supercomputing center in Galicia. FSAS is upgrading the current Finisterrae IV system (built by Fujitsu) to a new version that will run hybrid AI algorithms and connect to a quantum machine. Other centers equipped by Fujitsu may pursue similar strategies. Romeo acknowledges the immediate challenge: adding artificial intelligence to a supercomputing center means absorbing a steep increase in energy consumption. Yet the momentum behind AI is overwhelming—it is reshaping entire sectors, making the investment unavoidable. Quantum computing, he notes, remains much earlier in its development. When quantum machines become reliable, they will excel at specific computational problems where classical systems struggle. For now, they fail frequently enough that researchers must resort to quantum emulation—essentially simulating quantum behavior to verify whether the machines are working as intended. The future may belong to quantum systems, but that future is not yet here.
Notable Quotes
AI functions, has a market, and is a profitable investment. Quantum technology investment is speculative spending with zero monetary return, so only public institutions can afford it.— Gonzalo Romeo, general director of FSAS Technologies Spain
When quantum machines become reliable, they will be unbeatable for certain computational loads, but today they fail frequently enough that we must resort to quantum emulation to verify their behavior.— Gonzalo Romeo
The Hearth Conversation Another angle on the story
Why does Fujitsu need to position itself across all three technologies—supercomputing, AI, and quantum—rather than specializing in one?
Because the infrastructure itself is changing. Data centers can't stay single-purpose anymore. A customer might need massive computational power for AI training today, but also want to experiment with quantum algorithms for optimization problems. Fujitsu is betting that the centers that survive will be the ones that can integrate all three.
But Romeo said quantum computing has zero monetary return right now. Why would a company invest in something that doesn't make money?
Because they're not selling quantum machines—they're selling the infrastructure that will eventually use them. It's a long-term positioning play. If you're the company that knows how to wire quantum into a supercomputing center, you're valuable when quantum finally matures.
The Barcelona project is interesting because it's partly an expansion of an existing supercomputer. Why not just build something new from scratch?
Efficiency. MareNostrum 5 is already ranked 14th in the world. You don't abandon that. You grow it. And you learn from what's already there—all those different vendors, all those different technologies already coexisting. That's your blueprint.
Energy consumption seems like the real constraint here. How do you cool a system that's running AI workloads at scale?
Liquid cooling is the answer, but it's still rare and expensive. Most data centers use air cooling, which doesn't cut it for this kind of density. So you're choosing between vendors carefully, making sure every component works together thermally. It's not just about the chips anymore—it's about the whole ecosystem.
What happens to quantum computing if it keeps failing as often as Romeo suggests?
It stays in research mode. Researchers will keep using emulation—basically running quantum algorithms on classical computers to test them—until the hardware gets reliable enough to matter. That could take years. But the centers that are ready when it does happen will have a real advantage.