Europe orders its €387.8M AI supercomputer from its own champions

LUMI-AI will rise in Finland: Bull architecture, AMD accelerators, Nokia networking and IBM storage, delivering 10x the AI capacity of today's LUMI. This is the European link in the sovereignty chain.

Paylaş
Europe orders its €387.8M AI supercomputer from its own champions

The AI sovereignty wave we tracked this week from India to Britain has gained its European link: the €387.8 million LUMI-AI supercomputer has been ordered. The system will be installed at CSC's data center in Kajaani, Finland, delivering a tenfold increase in AI capacity over today's LUMI and nearly double its high-performance computing muscle. Deployment is scheduled for the second half of 2027.

The parts list is deliberately European: prime contractor Bull brings its BullSequana XH3500 architecture, the France-developed BXI interconnect and its patented warm-water direct liquid cooling; Nokia builds the data center network. The compute comes from AMD: Instinct MI430X accelerators and sixth-generation EPYC processors with 256 cores. IBM lays the storage foundation with IBM Storage Scale, announcing it will "help power one of Europe's most advanced AI and HPC environments".

LUMI-AI will serve as the backbone of Europe's AI Factory program: industrial innovation, scientific research and advanced AI applications across the continent will run on this machine. The message reads straight off the hardware list: Europe wants its critical AI infrastructure built, as far as possible, from its own supply chain.

The bigger picture: Bedrock's in-country inferencing in India, Oracle's dedicated cloud regions, Britain's state-as-customer programme, and now Europe's own pile of iron. The geography of AI is being drawn fast, and everyone is wiring a socket into their own continent.

Devamını oku

Liquid AI تطلق أكبر نماذج LFM2.5 المفتوحة لتشغيل الوكلاء على الأجهزة

Liquid AI تطلق أكبر نماذج LFM2.5 المفتوحة لتشغيل الوكلاء على الأجهزة

أطلقت Liquid AI نموذج LFM2.5-24B-A2B كنموذج مفتوح، ويضم 24 مليار معامل إجمالا لكنه يشغل نحو 2.5 مليار منها في كل خطوة. وبحسب النتائج التي نشرتها الشركة، يولد النموذج المدرب للوكلاء الذين يستخدمون الأدوات 184 رمزا في الثانية على Apple M5 Max.

GlobalFeed Editor tarafından