Last week I moderated Panel 19 at AI Summit Barcelona, “From BSC to Brussels: Europe’s Supercomputing Bet on AI”, and opened it with a question to the audience: “Who thinks ‘supercomputer’ is the name of a Marvel character out to save the world?” A few kind attendees raised their hands, and I jokingly invited everyone else to leave the room!
Earlier this week the White House gave the whole AI field a name that would fit on a comic-book cover or a film poster: President Donald Trump signed an executive order renaming artificial intelligence to “Super Intelligence”. A couple of months ago we dedicated a whole issue of The Compute to the definitions of AI, and fortunately it does not need to be updated, because the new term means exactly what the old one did. The order defines “Super Intelligence” as the technologies already covered by the definition of artificial intelligence in US federal law, so every chatbot, image generator and recommendation engine that was called AI on Monday became SI on Tuesday without any change to what it does or could do.
The supercomputers on my panel had six decades to earn their prefix, measured against clear, published benchmarks and in competition with every other machine in the world.
This issue follows the prefix from a New York newspaper in 1929 to Washington and Barcelona in 2026, and looks at what each kind of “super” actually delivers.
Origin story
The New York World was the first to use the term “super computing”, in 1929, describing the large custom-built tabulators (punched-card machines that sort and total columns of figures) that IBM had made for Columbia University. It benchmarked them against humans, stating that the machines had “the mental power of 100 skilled mathematicians”.
By then the prefix was already in fashion. In 1899 the Daily Mail printed the earliest use of “superhero” that the Oxford English Dictionary has found, and in 1903 George Bernard Shaw’s play Man and Superman popularised “superman” in English as a rendering of Friedrich Nietzsche’s Übermensch, a being beyond ordinary humans. Comic books gave both words their capes when Superman appeared in Action Comics in 1938.
Two decades later, Seymour Cray made the prefix measurable. Around 1960 he set out at Control Data Corporation to build the fastest computer in the world by a large margin, and the result, the CDC 6600, released in 1964, is generally considered the first supercomputer. Since then a machine earns the “supercomputer” label only when it ranks among the fastest on a published benchmark, and it loses the label as soon as faster ones are switched on. The TOP500 list keeps the score on that. Europe’s current champion is JUPITER in Germany: built around some 24,000 Nvidia chips, it was the first machine in Europe to pass one exaflop (a quintillion calculations a second), and it ranked fourth in the world on the list in June 2025.
A rose by any other name
Mr Trump signed the order, titled “Inaugurating the Era of Super Intelligence”, after a White House lunch with technology leaders, and said the government was “renaming artificial intelligence because it’s not artificial”. The order directs the executive branch to use “Super Intelligence” and “SI” in place of “artificial intelligence” and “AI”, to the maximum extent the law allows, and gives the president’s science adviser 60 days to propose legislative language for a new federal definition.
At the lunch, six of the guests signed the White House Accord on Super Intelligence alongside Mr Trump: Sundar Pichai of Google, Dario Amodei of Anthropic, Mark Zuckerberg of Meta, Jensen Huang of Nvidia, Greg Brockman of OpenAI and Elon Musk of xAI. The text runs to just over 300 words and commits the companies to four layers of AI (or should I say SI) checks: internal controls, an internal oversight team, external audits and an independent oversight board. Nothing in it can be enforced, since the companies signed voluntarily, and Mr Trump called it “morally binding”.
The signatures hide a real disagreement about the speed of AI development. In mid-September Mr Amodei called on frontier labs to slow the pace at which they improve their models, and within days Mr Huang and Mr Zuckerberg rejected a coordinated slowdown, arguing that competition and legal liability already push companies to make their models safe. All three signed the accord anyway.
No cape required
Back on my panel, Diego Perino, director of the AI Institute at the Barcelona Supercomputing Center (BSC), and Raphaëlle Achach, AI project manager for AI Factory France at GENCI, France’s national agency for high-performance computing, described the bet that Europe is placing on AI.
Unlikely to outspend the US companies racing to the frontier, Europe puts its public supercomputers, and the engineers who run them, at the disposal of the people building AI in Europe. Through the European High Performance Computing Joint Undertaking (EuroHPC), the EU body that buys the machines and allocates their time, 19 AI Factories give that compute time free of charge to researchers, startups and small and medium-sized enterprises (SMEs).
Both panellists stressed that the compute comes with people: the centre’s engineers help a team get its model running. The free access covers training and developing models, and excludes inference, that is, running a finished model in production to serve customers. The conditions are not difficult to meet, the main ones being civilian use, compliance with the EU AI Act and a final report on the project, and the machines, both said, are already running at capacity.
What Europe hopes to get back is a generation of European researchers and companies who know how to build AI on these machines.
Coming soon to a gigafactory near you
The next step up is bigger and more powerful machines: the EU is choosing up to seven sites for AI Gigafactories, backed by up to €10bn of EU and national funds, with the winners due to be picked early next year.
The public money comes in the form of guaranteed orders: EuroHPC and 18 member states will buy part of each gigafactory’s computing time, and private investors are expected to fund the rest, an additional €20bn or more according to the Commission’s estimate. Unlike the AI Factories, the gigafactories are designed to provide AI inference as well as to train the models, and the first is expected to go live around the second half of 2028.
In 1929 the New York World measured its super machine at 100 skilled mathematicians. The tender measures gigafactories in processors: each must hold three to four times as many of the most advanced AI chips as Europe’s most powerful AI factory, which, on our arithmetic against JUPITER’s 24,000, means roughly 70,000 to 100,000. That is about the size of the first phase of Stargate UAE in Abu Dhabi, which G42 and OpenAI are building with an estimated 100,000 Nvidia chips and expect to bring online by the end of this year.
Note to readers
Following the success of the session we held during AI Week, we are planning a second session of “Follow the Token: from budgeting to legal compliance” in November… stay tuned for more!
Wirelessly yours,
Ziad Matar
Co-founder, Tokenando
Editor-in-chief, The Compute
After Hours
On the silver screen, I watched Task and Monster: The Lizzie Borden Story, and finished season one of The Gentlemen, so I am now ready to start season two.
Reading and socialising took a back seat in the past week, and so did my foodie habits, although today’s lunch centred on a takeaway plate of sacchetti with ricotta and pear from one of the eateries in the office neighbourhood, which never fails to delight… so that is definitely what fuelled the writing of this issue, followed by a good cup of Americano!
Music this week stayed on the focus theme, with the discovery of a good album, Jardins Migrateurs by Constantinople and Ablaye Cissoko, Stay by Joris Delacroix, and recurrent guest Ólafur Arnalds with Happiness Does Not Wait.
References
The Compute, issue on AI definitions: https://tokenandoai.substack.com/p/standard-definition-5-ai-words-we
Columbia University Computing History, “The Columbia Difference Tabulator 1931” (reproduces the New York World article “Super Computing Machines Shown”): https://www.columbia.edu/cu/computinghistory/packard.html
Oxford English Dictionary, “superhero, n.”: https://www.oed.com/dictionary/superhero_n
Encyclopaedia Britannica, “Man and Superman”: https://britannica.com/topic/Man-and-Superman
The American Heritage Dictionary of the English Language, “superman”, word history
Forschungszentrum Jülich, “JUPITER”: https://www.fz-juelich.de/en/jupiter
Forschungszentrum Jülich, “Europe’s First Supercomputer Reaches 1 ExaFLOP/s”: https://www.fz-juelich.de/en/news/archive/press-release/2025/europes-first-supercomputer-reaches-1-exaflop-s
The White House, “Inaugurating the Era of Super Intelligence”, executive order, September 29, 2026: https://www.whitehouse.gov/presidential-actions/2026/09/inaugurating-the-era-of-super-intelligence/
The Hill, “Trump to sign order renaming AI as ‘superintelligence’”, September 29, 2026: https://thehill.com/policy/technology/6118532-trump-to-rename-ai-superintelligence/
Washington Examiner, “READ IN FULL: Trump and tech leaders’ White House Accord on Super Intelligence”, September 29, 2026: https://www.washingtonexaminer.com/news/white-house/4747747/full-trump-white-house-accord-ai-super-intelligence/
Forbes, “White House Releases ‘Accord’ Between Billionaire AI Execs: Here’s What It Says”, September 29, 2026: https://www.forbes.com/sites/saradorn/2026/09/29/white-house-releases-accord-between-billionaire-ai-execs-heres-what-it-says/
CNBC, “Trump tries to rename AI ‘super intelligence’ as polls show him sinking on key issue”, September 29, 2026: https://www.cnbc.com/2026/09/29/trump-ai-super-intelligence.html
Fortune, “Mark Zuckerberg says AI doesn’t need an industry-wide slowdown”, September 16, 2026: https://fortune.com/2026/09/16/mark-zuckerberg-meta-ai-safety-jensen-huang-dario-amodei/
EuroHPC Joint Undertaking, “AI Factories access modes”: https://eurohpc-ju.europa.eu/ai-factories/ai-factories-access-modes_en
EuroHPC Joint Undertaking, “The EuroHPC Joint Undertaking launches the AI Gigafactories Call”, July 30, 2026: https://www.eurohpc-ju.europa.eu/eurohpc-joint-undertaking-launches-ai-gigafactories-call-2026-07-30_en
EuroHPC Joint Undertaking, “Call for tenders for the selection of Artificial Intelligence Gigafactory Consortia and the establishment of AI Gigafactories”: https://www.eurohpc-ju.europa.eu/call-tenders-selection-artificial-intelligence-gigafactory-consortia-and-establishment-ai_en
European Commission, “EU launches AI Gigafactories call to boost Europe’s computing capacity”, July 30, 2026: https://ec.europa.eu/commission/presscorner/api/files/document/print/en/ip_26_1708/IP_26_1708_EN.pdf
EuroHPC Joint Undertaking, presentation to the ECMWF high-performance computing workshop: https://events.ecmwf.int/event/460/contributions/5527/attachments/3208/5351/HPC-WS_Wood.pdf
Enterprise, “G42 to develop Stargate UAE alongside US AI powerhouses”, May 23, 2025: https://enterpriseam.com/uae/2025/05/23/g42-to-develop-stargate-uae-alongside-us-ai-powerhouses/
Epoch AI, “OpenAI Stargate UAE”: https://epoch.ai/data/ai-data-centers/directory/openai-stargate-uae