The Dispatch

Samsung will begin mass production of HBM4 memory, crucial for AI, featuring double the bandwidth and 40% less power consumption than HBM3. This gives it a temporary advantage over SK Hynix and native support in 4nm process, positioning it as a leader in AI hardware. https://tinyurl.com/53cfjuum

OpenAI and Anthropic unveil GPT-5.3-Codex and Claude Opus 4.6 for advanced programming. Codex excels in rapid generation and multitasking; Opus enhances autonomy, extensive context handling, and adaptability, according to benchmarks and specialization https://tinyurl.com/4r5269nn

Between 2025-2026, OpenAI loses 24% of mobile market share in the U.S. to Gemini and Grok, which grow by 10.4% and 13.6%. Google and Anthropic are advancing in enterprise AI; Claude Code leads. https://tinyurl.com/bde68tat

Alibaba is advancing the Qwen series, a family of open and modular AI models that has already surpassed 700M downloads and 100K variants, fostering a collaborative ecosystem in China to reduce the digital divide and enhance robotics, medicine, and industry. https://tinyurl.com/3mfkwvu6

Alibaba's open Qwen family surpasses 700M downloads on Hugging Face, dominating the market and displacing Llama. It features lightweight models optimized for common devices and a wide variety of specialized versions for multimodal AI. https://tinyurl.com/2m5w7szh

NVIDIA demands full upfront payment for H200 chips destined for China, citing regulatory uncertainty and import risks. With high demand and limited stock, it plans to increase production with TSMC in 2026 to serve the Chinese market. https://tinyurl.com/3kwcrxpn