Nvidia AI Day Seoul opened with a wave of physical AI partnerships with Samsung, SK, and LG, while Anthropic reportedly took over a $3.4 billion gas-generator order to secure power for its data centers. The U.S. Department of Defense signed contracts with seven companies, including OpenAI, Google, and Nvidia, to use AI on classified networks, and Apple's new CEO pushed to accelerate AI development beyond Siri. Meanwhile, a real-world coding-agent test showed even the best model completing only 26.4% of tasks perfectly, exposing the gap between benchmarks and reality.
🇰🇷 Nvidia AI Day Seoul Opens… 'Physical AI' Alliance with Samsung, SK, and LG
Nvidia opened "Nvidia AI Day Seoul 2026" at COEX in Seoul on the 9th, unveiling physical AI collaboration results with Korean companies. Samsung Electronics said it used Nvidia PhysicsNeMo to make semiconductor design simulations about 1,600 times faster than experimental workflows, and LG Electronics said its Isaac GR00T-based CLOiD robot performs assembly and manufacturing work in real factory environments. SK Hynix, Naver Cloud, KakaoMobility, and Hyundai Motor Group also shared collaboration cases in memory simulation, open-model training, Level 4 autonomous driving, and enterprise AI agents, respectively.
⚡ Anthropic Takes Over $3.4 Billion Gas Generator Order to Secure Data Center Power
Anthropic took over a 1.7-gigawatt gas-generator order worth about $3.4 billion to secure power for future AI data centers, the Financial Times reported. The deal was struck in September but only became known on November 8; Anthropic took over the existing order from Texas-based OnePWR Solutions without acquiring the company itself. With data center power shortages becoming a bottleneck for AI expansion, the move is seen as a way to bypass the years-long wait for power equipment, though neither Anthropic nor OnePWR officially confirmed the deal at the time of the report.
🪖 Pentagon Signs Classified-Network AI Deals with OpenAI, Google, Nvidia, and Four Others
The U.S. Department of Defense announced contracts with seven tech companies — OpenAI, Alphabet (Google), Nvidia, SpaceX, Microsoft, Amazon, and Reflection — to use each company's AI tools on classified military networks for "lawful operational purposes." The department described it as an acceleration of turning the U.S. military into an "AI-first fighting force." The deals came amid a conflict with Anthropic, which refused the Pentagon's demand for unrestricted model access, was then designated a supply-chain risk company and barred from federal agency use, and filed a lawsuit in March in response.
🍎 Apple's New CEO Presses to Accelerate AI Development… Agentic Siri Could Come as Early as 2027
Apple's new CEO has reportedly pushed hard to accelerate AI development beyond Siri. According to Bloomberg, the CEO pressed software chief Federighi to speed up AI development, while a separate report predicted Siri could evolve into a more agentic form as early as 2027. Both reports cover plans and forecasts rather than released products, but they are read as signs that Apple's AI strategy has entered a serious execution phase.
🤖 Alibaba's Real-World Coding Agent Test: Even the Best Model Completed Only 26.4% Perfectly
In a coding-agent test run by Alibaba under conditions close to real work environments, even the best-performing model completed only 26.4% of the selected tasks perfectly. That stands in stark contrast to some existing benchmarks, where similar models score over 99%. Critics say benchmark scores may be overstating agents' abilities in actual development work, fueling debate over the need for new evaluation standards to gauge coding agents' real-world performance.
🇨🇳 DeepSeek's Agent Platform Tops 1.54 Million Daily Users
DeepSeek's AI agent platform reached 1.54 million daily active users, Bloomberg reported. The figure shows DeepSeek expanding beyond standalone models into an agent platform, in line with intensifying agent competition among Chinese AI companies. DeepSeek has drawn attention for its low cost relative to model performance, and it is now rapidly widening its user base in the agent space as well.
⚖️ Trump: "AI Guardrails Are Fine"… Signals Willingness to Work with Democrats, But Not at the Cost of Falling Behind China
President Trump said he has no objection to AI guardrails and signaled he could work with Democrats on them. But he also warned that regulation must not put the U.S. behind China. With bipartisan cooperation on AI regulation now being discussed, the debate centers on how to put safety measures in place while maintaining America's technological competitiveness.