2025年初人工智能格局报告：推理模型、主权AI及代理型AI的崛起（英文版）

The AI Landscape of Early 2025: Reasoning, Sovereignty, And the Rise of Agentic AIJeongkyu Shin, Lablup Inc.Table of Contents Table of ContentsChapter 1Chapter 2Chapter 3Chapter 4Chapter 5Chapter 6Chapter 7Chapter 8The New Frontier: Reasoning Models and the Test-Time Compute RevolutionChina's Rise: DeepSeek Shock and China's AI TrajectoryThe AI Infrastructure War: Power, Cooling, and GeopoliticsThe Silicon Engine: Accelerators, NPUs, and the Journey Toward IndependenceThe Rise of Sovereign AI: Age of Competition, National StrategiesInnovators: Startup Approaches and Big Tech Power DynamicsThe H1 2025 Frontier Model LandscapeThe Proliferation of Coding AI: From Assistant to Autonomous Teammate24679111314The AI Landscape of Early 2025: Reasoning, Sovereignty, And the Rise of Agentic AIExecutive SummaryAs we reach mid-2025, the artificial intelligence (AI) ecosystem is undergoing a fundamental paradigm shift amid intensifying geopolitical competition. This analysis examines the major transformations from the second half of 2024 through July 2025, aiming to provide guidance for navigating this rapidly evolving landscape.The most significant change in the first half of this year has been the shift in how AI models improve their performance—from the capital-intensive 'Train-Time Compute' approach to 'Test-Time Compute,' which invests more computation during inference to induce deeper reasoning. This transformation has catalyzed the emergence of 'Reasoning Models' such as OpenAI's o1 and China's DeepSeek R1, creating economic ripple effects that lower the barriers to entry for frontier model development. This became the technical backdrop for the 'DeepSeek Shock' of January 2025, when DeepSeek's claims of achieving frontier performance at remarkably low cost sent shockwaves through global markets. This event elevated the U.S.-China technology competition to a new dimension, demonstrating that Chinese AI capabilities had reached world-class levels in software and algorithmic efficiency despite hardware sanctions.This competition is spreading into an infrastructure war. Data centers are transitioning to liquid cooling systems to handle power densities exceeding 250kW per rack, with nuclear energy being mobilized to secure stable power supplies. This signifies the emergence of energy sovereignty as a new geopolitical variable directly linked to AI sovereignty. In the AI accelerator market, NVIDIA is consolidating its dominance through integrated 'AI Factory' platforms with its Blackwell Ultra and Rubin roadmap, while AMD challenges with cost, availability, and openness as its weapons through the MI400 series and open ROCm software stack.From a geopolitical perspective, 'Sovereign AI' has emerged as a core national security agenda. Major nations including the United States, United Kingdom, France, and Japan have announced massive investment plans to control their own AI models, infrastructure, and data, while South Korea aims to leap into the global AI top