在欧美关系出现新的不确定性之际，欧洲与加拿大正在进一步加强人工智能和高性能计算领域的合作。欧盟委员会主席乌尔苏拉·冯德莱恩近日确认，加拿大将加入欧洲高性能计算合作网络EuroHPC Joint Undertaking。这意味着加拿大将获得欧洲大型超级计算设施的优先使用权，同时欧洲也将借助加拿大的科技产业，进一步扩大双方共同的人工智能基础设施。
根据合作安排，加拿大将成为欧洲公共与私人合作高性能计算网络的积极参与者。加拿大企业将能够优先使用欧洲的超级计算资源，其初创企业和科研机构也有机会参与相关项目。对于欧盟而言，这项合作则有助于引入加拿大具有竞争力的科技企业和技术资源，扩大欧洲人工智能生态系统的规模。
EuroHPC目前拥有相当庞大的计算基础设施网络，包括14台世界级超级计算机、10台量子计算机以及19座人工智能工厂。整个网络覆盖38个参与国家，支持67个研究项目，并设有36个国家级能力中心。
该计划目前的总预算约为82亿欧元。此外，欧洲还计划追加约300亿欧元投资，用于建设最多7座新的人工智能数据中心，相关设施预计将从2027年开始动工。这些数据中心将进一步增强欧洲在人工智能训练、科学计算和复杂工程模拟等领域的计算能力。
此次合作并不只是简单地开放计算资源。随着人工智能模型规模不断扩大，训练和运行先进模型所需的算力、能源和数据中心投资已经成为国家科技竞争力的重要组成部分。对于加拿大而言，加入EuroHPC意味着其科研机构和企业可以获得更广泛的欧洲计算资源，而不必完全依赖本国有限的超级计算设施。
欧洲则希望借助加拿大的科技产业优势，扩大自身人工智能产业的合作范围。双方可以在高性能计算、量子计算、人工智能研究以及相关产业应用方面共享资源，形成一个跨大西洋的技术合作网络。
这一合作还有明显的地缘政治背景。近年来，加拿大与美国之间的关系逐渐趋于紧张，双方在贸易、产业政策和国家战略等问题上的分歧不断增加。欧洲显然希望利用这一机会，加强与加拿大的关系，将其发展为更加紧密的科技合作伙伴。
加拿大总理马克·卡尼与冯德莱恩此前已经共同启动名为“未来联盟”（Alliance for the Future）的合作倡议，为理念相近的国家提供新的合作渠道。冯德莱恩还曾提出，考虑让加拿大成为欧盟第一个准成员国。共享超级计算资源只是这一更广泛合作框架中的一部分。
不过，这项联盟能否真正改变全球高性能计算领域的力量格局，仍然存在很大疑问。
卡内基国际和平基金会人工智能专家安东·莱希特认为，加拿大与欧盟这样的“中等强国”加强合作，是建立共同人工智能基础设施最直接的短期举措之一。但他同时指出，如果没有大规模新增计算设施，这种合作很难对美国在超级计算领域的主导地位产生实质性影响。
美国目前正在投入巨额资金建设新一代高性能计算集群和人工智能数据中心。莱希特别强调，只要美国仍在建设以吉瓦级能源规模为基础的大型计算集群，其他国家即使加强资源共享，也难以单纯依靠合作改变竞争格局。
这实际上指出了欧洲与加拿大面临的核心问题：共享现有算力固然能够提高资源利用效率，但人工智能竞争最终仍然取决于计算设施的绝对规模、先进芯片供应能力、能源保障以及持续投入的资金。
因此，欧洲与加拿大此次合作更像是在美国主导的人工智能基础设施体系之外，建立一个更具自主性的跨国合作网络，而不是立即形成能够与美国抗衡的独立技术阵营。双方能否将现有合作转化为真正具有竞争力的计算能力，还要取决于未来数百亿欧元投资能否按计划落地。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
At a time of new uncertainty in European-American relations, cooperation between Europe and Canada in the area of artificial intelligence and high performance computing is being further strengthened. The President of the Council of the European Union, Ursula von DeLean, recently confirmed that Canada would join EuroHPC Jordan Union, a European network for cooperation in high-performance computing. This means that Canada will receive preferential access to large European supercalculator facilities, while Europe will further expand its common artificial intelligence infrastructure through the Canadian science and technology industry.
Under cooperative arrangements, Canada will be an active participant in the European Network for High Performance Calculations of Public-Private Cooperation. Canadian enterprises will be able to prioritize the use of European supercalculative resources and their start-ups and scientific institutions will have the opportunity to participate in related projects. For the EU, this cooperation has contributed to the introduction of Canada ' s competitive scientific and technological resources and to the expansion of the European artificially intelligent ecosystem.
EuroHPC currently has a considerable computing infrastructure network, including 14 world class supercomputers, 10 quantum computers and 19 artificial intelligence plants. The network covers 38 participating countries, supports 67 research projects and has 36 national capacity centres.
The current total budget of the plan is approximately Euro8.2 billion. In addition, Europe plans to invest about Euro30 billion in the construction of up to seven new artificial intelligence data centres, which are expected to commence in 2027. These data centres will further enhance European computing capabilities in areas such as artificial intelligence training, scientific computing and complex engineering simulations.
This cooperation is not simply an open calculation of resources. With the expansion of artificial intelligence models, investments in computing, energy and data centres for the training and operation of advanced models have become important components of national science and technology competitiveness. For Canada, joining EuroHPC means that its research institutions and enterprises have access to a broader European computing resource, without having to rely entirely on its limited super computing facilities.
For its part, Europe wants to build on Canada ' s scientific and technological industry advantage and expand its cooperation with the artificial intelligence industry. They can share resources in high-performance computing, quantum computing, artificial intelligence research and related industry applications, forming a transatlantic technical cooperation network.
This cooperation also has a clear geopolitical context. Relations between Canada and the United States have become increasingly tense in recent years, with growing differences on trade, industrial policy and national strategies. Europe clearly wishes to take advantage of this opportunity to strengthen its relationship with Canada and to develop it as a closer science and technology partner.
Canada's Prime Minister, Mark Kany, and von DeLean have previously launched a cooperative initiative called Alliance for the Future to provide new avenues of cooperation for like-minded countries. von DeLean also suggested that Canada be considered as the first associate member of the EU. Sharing supercalculative resources is only part of this broader cooperation framework.
However, there remains considerable doubt as to whether the alliance will truly change the dynamics of the global field of high performance computing.
Anton Lecht, an artificial intelligence expert of the Carnegie Endowment for International Peace, believes that Canada's enhanced cooperation with a “middle power” such as the European Union is one of the most immediate short-term initiatives to build a common artificial intelligence infrastructure. At the same time, however, he noted that without large-scale new computing facilities, it would be difficult for such cooperation to have a substantial impact on the dominance of the United States in the area of supercalculations.
The United States is currently investing heavily in a new generation of high-performance computing clusters and artificial intelligence data centres. Lech stressed, in particular, that as long as the United States was still building large computing clusters based on the scale of Giva energy, it would be difficult for other countries to rely solely on cooperation to change the pattern of competition, even if resources were to be shared.
This in fact points to the core problems facing Europe and Canada: while sharing existing computing power can improve resource efficiency, competition for artificial intelligence ultimately depends on the absolute size of computing facilities, the availability of advanced chips, energy security and sustained investment.
Thus, this European-Canadian cooperation is more like a more autonomous network of transnational cooperation outside the United States-led system of artificial intelligence infrastructure than an immediate stand-alone technological camp that can counterbalance with the United States. The ability of both sides to translate existing cooperation into a truly competitive computing capability will also depend on the planned landing of hundreds of billion euros in future investments.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.