TL;DR

OpenAI launches o3 Pro model; AMD secures major AI chip wins; Nvidia drops China from forecasts.

Highlights

• OpenAI released the “o3 Pro” model for ChatGPT, citing major reasoning and performance gains; Mistral, Meta, and ByteDance also announced new AGI and reasoning initiatives1.
• AMD launched MI350/MI355 AI chips, claiming better price/performance than Nvidia Blackwell ; OpenAI, xAI, Meta , and Oracle are early adopters2.
• Oracle will offer clusters of up to 131,000 AMD MI355X chips for large-scale AI workloads; Seekr is among the first customers5.
• Nvidia removed China from its sales forecasts due to U.S. export restrictions; Chinese AI firms are seeking offshore hardware access3.
• Scale AI will appoint Jason Droege as CEO after Meta acquires a 49% stake ($15B valuation), raising client data privacy concerns4.
• Major Google Cloud and Cloudflare outages disrupted global web traffic and AI platforms for several hours, highlighting infrastructure concentration risks6.
• Apple delayed its rebuilt Siri AI assistant to spring 2026, extending its timeline for generative AI features7.
• Microsoft released Copilot Vision, a free AI assistant for Windows 10/11 that analyzes and interacts with the full desktop8.
• Meta’s AI app exposed users’ private chats in a public feed due to interface design issues, prompting privacy concerns9.
• Meta sued a Hong Kong company over 5,000 deepfake nude ads, part of a broader crackdown on AI-driven exploitation10.
• Google DeepMind launched Weather Lab, an AI platform for hurricane forecasting up to 15 days ahead, now under evaluation by U.S. and UK agencies11.
• U.S. electricity demand is projected to hit record highs in 2025–26 due to AI/data center growth; NERC warns of grid reliability risks and highlights nuclear energy’s role12.
• AstraZeneca and CSPC announced a $5.2B AI-driven drug discovery partnership, leveraging CSPC’s AI platform for chronic and immune-related diseases13.
• Adobe beat revenue estimates and raised its outlook on strong demand for AI-powered cloud design tools14.
• Harvard and Boston libraries are opening book and document collections for AI training, expanding data sources beyond internet content15.

Commentary

OpenAI’s o3 Pro model rollout, alongside new AGI initiatives from Mistral, Meta , and ByteDance, demonstrates the ongoing competition in foundation models and reasoning capabilities1. OpenAI is focusing on expanding its user base and infrastructure, while competitors are building dedicated teams and releasing new benchmark-topping models. This competitive pressure is likely to accelerate both technical progress and ecosystem lock-in, with switching costs and network effects becoming more pronounced1.

AMD’s MI350/MI355 launch, with public adoption by major AI players and Oracle’s massive cluster offering, signals growing momentum for alternatives to Nvidia in the AI hardware market25. AMD’s claims of better price/performance and its annual product cadence are likely to influence procurement strategies, especially as hyperscalers and enterprises seek cost-effective, multi-vendor options2. Nvidia’s removal of China from its forecasts, due to U.S. export controls, underscores ongoing geopolitical risks and the potential for further fragmentation in global AI hardware supply chains3.

Cloud infrastructure reliability remains a critical concern, as simultaneous outages at Google Cloud and Cloudflare disrupted numerous AI platforms and services6. The incident highlights the operational risks of concentrated cloud dependencies, reinforcing the need for robust multi-cloud and disaster recovery strategies, particularly as AI workloads scale6.

On the enterprise side, Meta’s increased stake in Scale AI is raising trust and data privacy concerns among clients, reflecting broader anxieties about data governance when tech giants expand through acquisition4. Meanwhile, Meta’s privacy lapse in its AI app and legal action against deepfake vendors illustrate the growing regulatory and reputational risks tied to AI deployment910. Apple’s delay on Siri and Microsoft’s broader Copilot Vision rollout show divergent speeds in consumer AI feature delivery78.

Finally, the surge in U.S. electricity demand—driven by AI and data center growth—puts energy infrastructure and sustainability in the spotlight, with nuclear power identified as a key enabler12. The inclusion of library archives for AI training and the expansion of AI in drug discovery and weather forecasting further demonstrate the sector’s broadening impact and the importance of high-quality, diverse data sources111315.