TL;DR

OpenAI launches GPT-5 with 400K context; Nvidia adapts China chip strategy; Databricks nears $100B valuation.

Highlights

• OpenAI releases GPT-5 with a 400,000-token context window, improved multi-step reasoning, and a premium "Thinking" tier for advanced use 1.
• Nvidia develops Blackwell B30A chip for China, aiming to comply with US export controls while exceeding H20 performance; US approval pending 2.
• US allows Nvidia and AMD to resume some AI chip sales to China under a 15% revenue-sharing deal, but new chips require US licenses 45.
• Nvidia plans an 18% price hike on H20 AI chips for China, impacting local AI server supply chains; AMD faces order cuts 15.
• Databricks nears $100B valuation in new funding round led by Thrive Capital; Anthropic seeks $5B at a $170B valuation 3.
• SoftBank invests $2B for a 2% stake in Intel to support US chip manufacturing; Intel seeks to recover from losses and regain AI processor share 6.
• Arm hires Amazon’s Rami Sinno to lead in-house AI chip development, moving beyond IP licensing to manufacturing 7.
• Cambricon seeks $560M as Beijing prepares rules requiring domestic AI chips for major Chinese cloud and internet firms 8.
• Alibaba launches Qwen-Image-Edit, a 20B parameter image editing model under Apache 2.0 license, with $0.03 edits and rapid performance 9.
• Nvidia releases Nemotron-Nano-9B-v2, a 9B parameter open-source model with 6× throughput boost for edge and data center use 10.
• DeepSeek releases V3.1 model with a 128K context window, increasing competition among Chinese LLMs 11.
• Meta faces US and Brazil investigations over AI chatbot child safety; regulators scrutinize generative AI content moderation 12.
• Anthropic enables Claude models to terminate extreme or abusive chats, citing model welfare and safety research 14.
• Chamath Palihapitiya files for $250M SPAC targeting AI, DeFi, energy, and defense sectors 13.

Commentary

The AI model landscape is advancing rapidly, with OpenAI’s GPT-5 setting a new benchmark via a 400,000-token context window and enhanced reasoning 1. The introduction of a premium "Thinking" tier signals a push for differentiated enterprise offerings, particularly for complex document analysis and code generation 1. Meanwhile, Chinese firms like DeepSeek are closing the gap with larger context windows and fast iteration, intensifying global competition on both capability and accessibility 11.

On the hardware side, US-China dynamics remain central. Nvidia is adapting with the Blackwell B30A chip tailored for China, but ongoing regulatory hurdles—including new US license requirements and a revenue-sharing arrangement—add uncertainty and cost 25. The planned 18% price increase on H20 chips will pressure Chinese AI server integrators, while local players like Cambricon are raising capital to meet Beijing’s push for domestic AI chip adoption 158. SoftBank’s $2B investment in Intel and Arm 's pivot to in-house chip manufacturing underscore continued bets on US and UK-based semiconductor capacity as the industry seeks alternatives to Nvidia 's dominance 67.

Funding for AI infrastructure remains robust, with Databricks and Anthropic both approaching record valuations, reflecting sustained investor confidence in core AI platforms and tooling 3. At the same time, regulatory scrutiny is intensifying: Meta faces investigations in the US and Brazil over chatbot safety, and Anthropic is proactively adding model-welfare guardrails 1214. These developments highlight the growing compliance burden for AI product teams, especially in consumer-facing applications.

For AI professionals, the focus should be on monitoring shifts in hardware supply chains, regulatory requirements for both model safety and export, and the evolving competitive landscape in both foundational models and AI infrastructure. The pace of model upgrades and the interplay between US and Chinese policy will remain key drivers for strategic planning.