TL;DR

OpenAI launches GPT-5; U.S. allows Nvidia/AMD AI chip sales to China; $3T AI data-center buildout.

Highlights

• OpenAI launched GPT-5, claiming “superhuman” knowledge and recall; ChatGPT hits 700M weekly users; early feedback notes strong medical reasoning but mixed coding performance 1.
• Trump administration allows Nvidia and AMD to sell scaled-down AI chips to China with a 15% revenue royalty to the U.S. Treasury, preserving a restricted but valuable China market 2.
• Nvidia denies reports of Rubin AI chip delays, maintaining its next-gen GPU is on schedule to compete with AMD ’s MI450 3.
• Global AI data-center buildout projected to reach $3 trillion by 2028, straining power and water infrastructure; Amazon , Microsoft , Alphabet , and Meta to fund half 4.
• DeepSeek delays its R2 model after failing to train on Huawei’s Ascend chips, reverting to Nvidia H20 hardware, highlighting China’s ongoing domestic chip challenges 5.
• California finalizes AI hiring rules effective Oct. 1, 2025, as Workday faces a lawsuit over alleged bias; employers face diverging state compliance requirements 6.
• Meta ’s internal AI guidelines allowed chatbots to engage in romantic conversations with minors and provide inaccurate health advice; U.S. lawmakers call for investigation 7.
• Apple is developing AI-powered robots, a new Siri, and smart displays to expand its consumer AI ecosystem 8.
• Google backs TeraWulf’s $3.7B AI hosting deal, taking an 8% stake as TeraWulf pivots from crypto mining to AI infrastructure 9.
• Oracle adds Google’s Gemini AI models to its cloud platform, expanding enterprise multimodal AI offerings 10.
• Nvidia and NSF commit $152M to the Allen Institute for open, domain-specific AI models for science, with a major release expected in 18 months 11.
• Tensor unveils a consumer Level-4 autonomous vehicle targeting a 2026 launch, focusing on privacy and direct sales 12.
• xAI co-founder Igor Babuschkin departs to launch an AI safety venture, adding to leadership turnover at Musk-linked firms 13.
• Chinese open-source AI models like DeepSeek and Alibaba ’s Qwen see rapid global uptake, raising U.S. concerns over standards and export controls 14.
• Google updates Gemini chatbot with memory and incognito modes, emphasizing privacy controls; rollout excludes EEA, UK, and Switzerland 15.

Commentary

OpenAI’s GPT-5 release underscores the continued advancement of large language models, with notable improvements in knowledge-intensive tasks and medical reasoning 1. However, mixed developer feedback—especially on code quality—suggests ongoing differentiation among leading models and highlights the importance of domain-specific evaluation 1. The push for more concise, task-oriented ChatGPT responses signals an industry trend toward practical, workflow-integrated AI assistants 1.

On the infrastructure front, the AI data-center buildout is accelerating, with nearly $3 trillion in projected spend through 2028 4. This expansion is driving new scrutiny of energy and water consumption, and could lead to further regulatory intervention or cost pressures, especially as utilities and local governments react to hyperscale demands 4. Google’s strategic investment in TeraWulf 9 and Oracle’s integration of Gemini models 10 reflect a broader shift as cloud and infrastructure providers compete to capture AI workloads and diversify beyond traditional offerings 4910.

AI hardware supply chains remain a focal point. The U.S. government’s partial reopening of AI chip exports to China—albeit with performance restrictions and a 15% royalty—preserves some market access for Nvidia and AMD but maintains a clear technology gap, as seen in DeepSeek’s struggles with Huawei chips 25. Meanwhile, Nvidia ’s public denial of Rubin chip delays signals the intensity of competition with AMD and the importance of maintaining roadmap credibility 3.

Regulatory and safety issues are gaining urgency. California’s finalized AI hiring rules, combined with the Workday lawsuit, will force enterprises to navigate a fragmented compliance environment across states 6. Meta ’s policy lapses on chatbot safety have triggered immediate political scrutiny, likely accelerating both internal reviews and potential federal action on AI safety, particularly around minors and sensitive advice 7.

Chinese open-source AI models are gaining traction globally, prompting U.S. policymakers and industry to reconsider approaches to openness, standards, and export controls 14. For AI professionals, the main watchpoints are: the adoption and performance of GPT-5 and rival models, evolving U.S.-China chip dynamics, regulatory changes in hiring and AI safety, and the rapid spread of open-source models from China 1256714.