--- type: "Topics" locale: "zh-HK" url: "https://longbridge.com/zh-HK/topics/39719049.md" description: "🔥🚀 Europe's computing power gap is being rewritten: $Nebius(NBIS.US) is not just a follower, but is defining "Sovereign AI infrastructure".After reading this logic, I've become clearer about one thing:Europe's problem has never been "whether it has AI", but rather — whether it has its own computing power.The US has hyperscalers, China has national-level computing power initiatives.What Europe lacks is "control".This is the so-called Sovereign AI.Not a stronger model, but:Data must stay localComputing power must be controlled locallyInfrastructure cannot rely on external sourcesOnce this premise is established, demand is no longer optional, but a necessity.This is also why I'm starting to re-evaluate $Nebius(NBIS.US)'s positioning.It's not an ordinary GPU rental provider.It's closer to an "infrastructure layer".The timeline is crucial.On March 31st, $Nebius(NBIS.US) announced the construction of a 310MW AI data center in Lappeenranta, Finland.This is not an ordinary project.This is directly on par with the level of "Europe's largest AI computing power base".And this is not an isolated event.Its facility in Mäntsälä has already expanded to 75MW and has deployed NVIDIA's latest architecture.What does this indicate?Not the planning stage, but already in the execution stage.I'm more focused on its expansion pace.240MW near Lille, France1GW project in Missouri, USAPlus the 310MW in FinlandThe goal is to secure over 3GW of power contracts by 2026.This is essentially not a "growth plan".But rather locking in future computing power supply in advance.Because in the AI era, what's truly scarce isn't GPUs, but:Power + Data Centers + Compliance CapabilityThe combination of these three is the real barrier to entry.So I'm now more inclined to view $Nebius(NBIS.US) as:Europe's version of a "computing power plant".You're not buying GPUs, you're buying "permitted computing power".This is completely different from traditional cloud.Looking at the revenue expectations.$3–3.4 billion in 2026.The key isn't the number, but the structure:Every new construction project is a "physical realization" of future revenue.Unlike software where you can tell stories.Every dollar here has to be built with power, racks, and cooling systems.This is also why I think it's still in the "construction phase".Not the "harvesting phase".The real explosion won't happen during factory construction, but at the moment when:Demand explodes + Capacity is just right.So the question becomes even more interesting:When AI enters the "sovereignization" stage, are models still the most valuable asset?Or is it these seemingly heavier, slower, but harder-to-replace infrastructures?Are you more inclined to bet on "model companies", or this type of computing power player that is locking in power and land?" datetime: "2026-04-03T14:50:58.000Z" locales: - [en](https://longbridge.com/en/topics/39719049.md) - [zh-CN](https://longbridge.com/zh-CN/topics/39719049.md) - [zh-HK](https://longbridge.com/zh-HK/topics/39719049.md) author: "[辰逸](https://longbridge.com/zh-HK/profiles/16318663.md)" --- > 支持的語言: [English](https://longbridge.com/en/topics/39719049.md) | [简体中文](https://longbridge.com/zh-CN/topics/39719049.md) # 🔥🚀 Europe's computing power gap is being rewritt… ### 相關股票 - [Nebius (NBIS.US)](https://longbridge.com/zh-HK/quote/NBIS.US.md) - [NVIDIA (NVDA.US)](https://longbridge.com/zh-HK/quote/NVDA.US.md)