Core Insights - Tech giants are engaged in a "reality simulator" arms race, with Google and Ant Group simultaneously launching world model technologies that will transform digital interactions [1] Group 1: Company Strategies - Google has introduced a subscription model at $125/month for US adult users, while Ant Group has opted for a fully open-source approach [3] - Both companies have achieved significant breakthroughs, including interaction latency under 1 second, continuous generation for up to 10 minutes, and physical collision calculation accuracy exceeding 92% [3] - Ant's model is trained on 20,000 hours of real robot data covering 9 mainstream robot configurations, while Google's relies on the collaborative computing architecture of Gemini 3 and Nano Banana Pro [3] Group 2: Technical Breakthroughs - The advancements in world models are evident in three areas: upgrading physical collision calculations from traditional video frame interpolation to real-world simulation, real-time conversion of text, images, and operational commands, and allowing users to control virtual perspectives via keyboard [3] - A significant challenge remains the computational power bottleneck, as scene drift occurs when continuous interaction exceeds 10 minutes [3] Group 3: Industry Impact - The impact on the industry is already noticeable, with 3D modeling costs in game development potentially decreasing by 70%, embodied intelligence training efficiency improving by 3 times, and autonomous driving simulation testing costs expected to drop by 85% [5] - The divergence in technical paths has led to regional characteristics in the supply chain, with US companies focusing on commercial API ecosystems and Chinese firms concentrating on vertical scene adaptation [5] Group 4: Market Response and Future Outlook - The capital market has reacted positively, with 23 new physical engine startups emerging globally in early 2026, and Nvidia launching the "Physical AI" architecture [5] - Analysts predict that by the end of the year, investments related to world models will account for 35% of total AI investments [5] - The next technological milestone is to extend continuous interaction time beyond 30 minutes, with Google aiming for this by Q4 2026 and Ant Group seeking to achieve it through distributed computing architecture [5] - As Google and Ant set examples, it is expected that companies like Microsoft and Meta will also launch their world model platforms within the year, marking the beginning of a new era in artificial intelligence focused on environmental cognition [5]