smartcybers786@gmail.com 
AI Artificial intelligence words, miniature of robot and EU flag are seen in this illustration taken December 21, 2023. —Reuters
#Europe #builds #gigafactories #industry
Amsterdam: The European Commission has collected $ 20 billion to build four “AI giga factories” as part of European strategy to meet the United States and China on artificial intelligence, but some industry experts have asked if they have no meaning in their construction.
Last month, the European Commission’s President Ursola Van Deer Leene will face challenges from the acquisition of chips to the appropriate locations and the search for electricity.
Even if we will build such a large computing factory in Europe, and even if we will train a model on this infrastructure, once it is ready, what will we do with it? “The Economic Think Tank said Britain Martin of Berjeel.
This is a chicken and egg problem. Hopefully, new local firms such as France’s NVIDIA -backed Mistal Startup will increase and will be used to create AI model that works according to the rules of EU AI safety and data protection, which are tightly compared to the United States or China. But making hardware on this scale is a dangerous plan in the absence of large European cloud services business, such as Google and Amazon, or millions of paying users, such as a Chattagpat Maker.
European response to Star Gate
The Giga Factory Plan is part of Europe’s response to the competitive report, which suggests bold investment and more active industrial policy. Van Der Leene released details for the first time at the February 11 AI summit in Paris as part of the Investment, responding to Europe’s 200 billion euros ($ 216.92 billion) $ 500 billion in response to the US Star Gate project.
They will enable Giga factories “public private partnerships … (which) to create all our scientists and companies – not only the largest – Europe to create the latest model needed to make the AI continent.”
They should be financed through the new 20 billion euros funds, which have been drawn from existing EU programs and from member states. The European Investment Bank will participate.
Giga factories will have 100,000 “modern” chips, Van Dare Leean said. The US chipmaker sells GPU chips needed to train AII for about $ 40,000-which means that a gigfactory per gigfactory is tagged by several billion euros.
Although this is a big deal, it still trails the plans announced by US firms. Facebook owners are spending $ 10 billion to facilitate 1.3 million GPUs in Meta Louisiana, with 1.5 GW of electric power.
Giga factories will face the problems facing private projects in Europe: NVIDIA chips need difficulty in getting NVIDIA chips and lack of electricity, Data Center Consultancy Consultancy CBRE’s Kevin Restio said.
Obstacles to access to chips?
Former President Joe Biden, the US government, has access to AI chips to avoid the construction of Giga factories in many European countries, though it is unclear whether the Trump administration will support it.
“There is nothing to say that the government cannot lay hands on these chips or … great plans cannot come from it, but it is unlikely to happen in the short term,” said Restio.
Burgail’s Martins said spending public money in the race to spend AI has no meaning. “The life of such factories, before you write it down and buy new Nvidia chips, is almost … a year and a half,” he said.
Meanwhile, the progress of the Chinese AI Model DiPsic raised questions about whether AI models can be trained with less computing power, and instead, the costs should be focused on applications, which require a variety of chips.
Europe’s former main support plan for technology infrastructure, the 2023 chips act, failed to meet the targets of modern chip manufacturing in Europe or reaching 20 % of the global production, though it led to investment in new factories needed to make automotive chips.
Along with the Giga factory plan, the Commission is also upgrading 12 scientific supercomputer centers to convert them to AI factories.
It is not yet clear how AIGGFFCs will be different in size, said Kimo Koski, Managing Director of Finland’s Lumi Super Computer.
“In my understanding, it is related to the use of the industry,” he said. It will be “an innovation in Europe, a very happy event.”
He said that supercomputers are already used for machine learning projects, as well as scientific use, such as climate modeling. He pointed to a Finnish firm Silo Ai, who used Lumi last year to help produce a large language AI model before speaking US chipmaker AMD for $ 665 million.
The potential beneficiaries of the supercompoting extension include European chipmakers who make non -GPU chips, which are still useful in data centers, including Germany’s Infinone and France’s ST micro electronics, as well as the French separable and the Netherlands, including the startups.
