Recent developments indicate a shift toward frontier model development by European cloud infrastructure providers and significant capital infusion into world-model simulation startups.

近期發展顯示，歐洲雲端基礎設施供應商正轉向開發前沿模型，而世界模型模擬新創公司亦獲得了大量資金注入。

The European cloud services sector is witnessing a strategic pivot toward the internal development of frontier artificial intelligence models. OVHcloud, the largest cloud provider in Europe, has announced its intention to engineer large-scale systems from inception. This initiative is framed by CEO Octave Klaba as a necessity for institutional longevity and a response to the volatility of reliance on non-European AI architectures. Klaba posits that the fiscal requirements for such development have diminished, suggesting that advancements in synthetic data and hardware efficiency have reduced estimated costs from one billion euros to a range between 150 million and 200 million euros. The firm intends to deploy a diversified family of specialized models, utilizing the Jupiter supercomputer and the acquired assets of DragonLLM, with a long-term objective of open-sourcing these systems upon the attainment of requisite performance benchmarks.

歐洲雲端服務部門正經歷一次策略轉向，目標是內部開發前沿人工智能模型。歐洲最大的雲端供應商 OVHcloud 宣布，其計劃從零開始設計大規模系統。執行長 Octave Klaba 將此舉定義為維持機構長久生存的必要條件，以及對依賴非歐洲 AI 架構之不穩定性的回應。Klaba 主張，此類開發的財政需求已經降低，並指出合成數據與硬體效率的進步，將預計成本從 10 億歐元降低至 1.5 億至 2 億歐元之間。該公司打算部署一系列多元化的專業模型，利用 Jupiter 超級電腦及收購的 DragonLLM 資產，長期目標是在達到所需性能基準後將這些系統開源。

Simultaneously, the venture capital landscape is prioritizing the transition from text-based large language models to world models capable of simulating physical environments. Odyssey, a startup founded by former autonomous vehicle specialists Oliver Cameron and Jeff Hawke, has secured $310 million in Series B funding, resulting in a valuation of $1.45 billion. The funding round, led by Natural Capital and supported by entities including Amazon and AMD Ventures, facilitates the development of systems that predict and interact with physical reality. Odyssey's methodology involves the collection of empirical physical data via human-operated cameras to ensure physics accuracy. Consequently, the organization has established a strategic partnership with AWS, optimizing its models for Trainium hardware to reduce dependence on Nvidia architecture.

與此同時，風險投資領域正將優先權從基於文本的大語言模型，轉向能夠模擬物理環境的世界模型。由前自動駕駛車專家 Oliver Cameron 與 Jeff Hawke 創立的新創公司 Odyssey，已在 B 輪融資中獲得 3.1 億美元，使其估值達到 14.5 億美元。本輪融資由 Natural Capital 領投，Amazon 與 AMD Ventures 等實體支持，旨在推動開發能預測並與物理現實互動的系統。Odyssey 的方法涉及透過人工操作的攝影機收集實證物理數據，以確保物理準確性。因此，該組織與 AWS 建立了策略合作夥伴關係，針對 Trainium 硬體優化其模型，以減少對 Nvidia 架構的依賴。

The AI industry is currently characterized by the pursuit of sovereign European capabilities and the scaling of physics-based simulation models through substantial private investment.

AI 產業目前的特點是追求歐洲自主能力，以及透過大量私人投資來擴大基於物理的模擬模型。