Nvidia plots China comeback with a tailored AI chip
Two quiet-but-significant product moves today show the AI business is as much about distribution and defaults as it is about raw model power.
Nvidia is building a China-specific AI chip it plans to start shipping before the end of the year. According to The Information, the part is an inference processor designed to comply with the US export-control rules that have kept Nvidia's most powerful silicon out of the mainland since 2023. The move is a direct attempt to claw back share in a market that has been rapidly filling with domestic alternatives as Beijing pushes self-reliance in AI hardware. The reporting is corroborated by Wccftech, CNBC, and the Financial Times, which separately note that China has already let ByteDance and Tencent import thousands of Nvidia's H200 chips — a sign both sides are testing how much commerce the rules actually allow.
Google is turning Discover into a chatbot you talk to. In the coming days the Google app will let you open a chat-style interface from the three-dot menu on your feed, describe what you want to see in plain language, and have AI reshape the recommendations — the assistant confirms your preferences, remembers them for future visits, and offers a "Refresh your feed" button when it's ready. It's part of a broader personalization push that also adds topic controls and custom audio briefings to Google News, and follows YouTube, Instagram, Bluesky, and X in handing users an AI-tuned feed. The bigger shift is who controls the algorithm: when you can simply tell Google what you like, the feed stops being a mystery and starts being a negotiation.
Should export-compliant AI chips for China be designed at all, or does any foothold just buy Beijing time to catch up? Tell us in the comments.
