AI hardware's spearhead and the most valuable company in the world, Nvidia, has reported another quarter of record sales and sustained growth, with a total of $46.7 billion in revenue. Compared to the same period last year, Nvidia's revenue is up 56%, while its net income spiked 59% to $26.4 billion. Although Nvidia's Blackwell architecture continues to lead the company's charge while CEO Jensen Huang preaches about booming demand, its reported zero sales of H20's to China for that quarter may have caused some temporary uncertainty among investors, causing Nvidia's share price to slip in after-hours trading.
The U.S. government's recent deal to acquire a near-10% stake in Intel also turns out to include provisions to discourage the company from spinning off its struggling foundry unit. The Financial Times reported that Intel CFO David Zinsner stated that the deal carries a five-year warrant allowing the government to claim an additional 5% stake if Intel reduces its ownership of the unit below 51%. While this does align with U.S. efforts to reinforce domestic chip production, it forces Intel to retain a business that lost $3.1 billion operationally just last quarter.
Nano Banana, a then-unknown AI image editor that went viral on social media, debuted this week by Google under a new name - Gemini 2.5 Flash Image. The update rolled out to all Gemini users as well as to developers via the Gemini API, Google AI Studio, and Vertex AI platforms. The new AI model was designed with precision in edits in mind, allowing users to change subtle details on images without changing anything not explicitly mentioned. Coupled with that, Google also released Google Vids a day later, an AI video creation tool for the workplace.
China is pushing to triple its AI chip output by 2026 in a bid to reduce the mainland's reliance on Nvidia, the Financial Times reported. To this end, Huawei plans to start production at a new chip plant by the end of the year, with another two facilities targeted for 2026. Meanwhile, SMIC is looking to double its 7nm chip production next year to support Huawei, its biggest customer. The news comes not long after Beijing began heavily scrutinizing Nvidia's chips as officials urged companies to move onto domestic tech.
Anthropic has announced that it will implement a new opt-in policy to begin training its AI models on both new and "resumed chats." Effective September 28, the policy also extends data retention to five years for those who consent. With the exception of Claude Gov and Claude for Work, the policy update applies to all individual user subscription tiers, as a pop-up will prompt users to make a choice, with the default toggle set to allow training. Although users can change their preference at any time, past training data cannot be withdrawn.