Daily Highlights
- Nvidia's DSX POD cuts AI factory deployment to 14 days
- Meta posts $64.6B Q3 revenue, up 26% year over year
- Google Cloud makes Vera Rubin infrastructure generally available
- Fireworks AI seeks funding at a $30-40B valuation
- China's top offices mandate full rollout of "AI Plus" action
TODAY'S HEADLINES
[At GTC Berlin, Jensen Huang unveils DSX POD that cuts AI factory deployment to 14 days, backed by nearly $10B for Europe] According to Cailianshe, speaking at the GTC Berlin conference on Tuesday, Nvidia CEO Jensen Huang introduced the DSX POD reference design built on the Vera Rubin platform, which is expected to shrink AI factory deployment from about six months to as little as 14 days once on-site utilities are ready. Huang said the computing industry stands at the start of a new industrial revolution, and Nvidia plans to invest nearly $10 billion in the European technology ecosystem; next-generation vehicles from Volkswagen Commercial Vehicles, Jaguar Land Rover, Volvo and Mercedes-Benz will adopt the Hyperion autonomous-driving platform. (Source: Cailianshe)
[Meta posts $64.6B Q3 revenue beats as a $9.9B legal settlement charge drags down profit] After the close on Oct 21, Meta reported third-quarter revenue of $64.6 billion, up 26% year over year and ahead of expectations; a $9.9 billion charge from its August settlement with a coalition of state attorneys general pushed GAAP operating income down 37% to $12.99 billion. Capital expenditures reached $34.9 billion in the quarter, free cash flow came in at negative $2.1 billion, and full-year capex guidance held at $130-145 billion. The stock climbed nearly 4% after hours as investors embraced the pace of its AI spending. (Source: Cailianshe)
NEW PRODUCTS & TECH
[Google Cloud and Nvidia declare Vera Rubin infrastructure generally available, with up to 960,000 GPUs planned] Cailianshe reports that Google Cloud and Nvidia announced the general availability of Vera Rubin-based infrastructure, to be expanded across multiple sites with as many as 960,000 Rubin GPUs planned. Vera Rubin is Nvidia's new full-stack computing platform for agentic AI, spanning CPU, GPU, networking and storage, and its commercial debut signals that large-scale training and inference clusters are now ramping fast. (Source: Cailianshe)
[Nvidia reveals Agile ONE SN humanoid robot reference platform, expected in 2027] At GTC Berlin, Nvidia introduced the Agile ONE SN reference humanoid robot platform built on Isaac GR00T, featuring 31 degrees of freedom and combining open models, simulation training tools and a runtime system to lower the barrier to robot development; it is expected to launch in 2027. Nvidia also upgraded its Isaac robotics development platform with new humanoid reference frameworks, simulation workflows and data tools, with companies including Agility Robotics, Figure and Kuka already adopting the technology. (Source: Cailianshe)
IPO & FUNDINGS
[Fireworks AI seeks new funding at a $30-40 billion valuation] According to Business Insider, AI inference infrastructure startup Fireworks AI is in early talks for a new funding round at a valuation between $30 billion and $40 billion, a dramatic jump from the $17.5 billion post-money valuation of its Series D just three months ago. Its annualized revenue has surpassed $1 billion, roughly five times the year-ago level, and the platform processes more than 40 trillion tokens a day for customers including Cursor, Uber and Shopify. (Source: Business Insider)
[SK Hynix completes a 40 trillion won buyback, the largest share cancellation by a Korean listed company] According to Cailianshe, SK Hynix has completed a 40 trillion won share buyback, retiring shares equal to about 3.3% of its outstanding stock - the largest share-cancellation operation ever by a Korean listed company. HBM memory remains in tight supply on surging AI training demand, and the move steps up returns to shareholders. (Source: Cailianshe)
[Nvidia expands supply-chain capacity, targeting a 2.5x boost in substrate output within three years] Cailianshe reports that Nvidia announced plans with supply-chain partners including Samsung Electro-Mechanics and Unimicron to expand substrate capacity, aiming to lift related substrate output to 2.5 times current levels within three years as it works to relieve the packaging bottleneck for AI chips. (Source: Cailianshe)
OPEN SOURCE & RESEARCH
[Nvidia spotlights its Nemotron, Cosmos and GR00T open model families and unveils the OpenShell secure agent architecture] At GTC Berlin, Nvidia showcased open models spanning multiple domains: Nemotron for language tasks, Cosmos as a world foundation model, GR00T for humanoids and Alpamayo for autonomous driving. Addressing the risk of runaway agent permissions, Nvidia also introduced OpenShell, a secure agent reference architecture that uses an open-source runtime, kernel-level sandboxing and policy controls to restrict which resources an agent can reach and monitor its behavior; Huang said agent permissions should adjust dynamically to the task at hand. (Source: Cailianshe)
COMPANY & PEOPLE
[Unitree's Wang Xingxing: leading LLM companies have largely achieved model self-evolution] At the 2026 World Physical AI Conference and Unitree partner ecosystem summit held in Hangzhou on Oct 22, Unitree chairman and CTO Wang Xingxing said in a keynote that leading large-model companies have internally achieved basic model self-evolution - models write their own programs and collect their own data while humans merely guide the process - though it still takes five to ten years for breakthroughs to spread to everyday users. He argued physical AI could hit its breakthrough moment within the next one to three years, and that the core challenge on the road to an embodied-intelligence "ChatGPT moment" is aligning model inputs and outputs with real robots. (Source: Securities Times)
[Tesla posts $28.05B Q3 revenue, down slightly, as operating profit falls 56%] According to Cailianshe, Tesla reported third-quarter revenue of $28.05 billion, down 0.2% year over year, with automotive revenue of $20.56 billion, down 3%, and operating profit of $718 million, down 56%. The company is navigating a transition in which its core EV business is under pressure while it pours investment into autonomous driving and robotics. (Source: Cailianshe)
[Zhipu wins Xiamen healthcare large-model project worth 10.58 million yuan] ZhiDongxi reports that the Xiamen Health Medical Big Data Center awarded its "AI medical large-model foundation and application engineering phase one" project to Beijing Zhipu Huazhang for 10.58 million yuan against a budget of 10.8 million yuan; the build covers a healthcare LLM foundation platform, data-security collaboration with federated learning, and intelligent mental-health service applications. (Source: ZhiDongxi)
POLICIES & DATA
[China's CPC and State Council general offices mandate full implementation of the "AI Plus" action] According to Cailianshe, the general offices of the CPC Central Committee and the State Council released a policy document on deepening the integration of scientific-technological and industrial innovation, calling for full implementation of the "AI Plus" action to bring AI across every industry, alongside decisive breakthroughs in core technologies including integrated circuits, industrial machine tools and high-end instruments. (Source: Cailianshe)
[OpenAI updates its usage policy, effective November 20] Cailianshe reports that OpenAI has updated its usage policy, effective November 20, 2026, to clarify rules for safe model development and sensitive uses of AI. The update lands as regulatory frameworks multiply worldwide and safety disputes swirl around the leading labs. (Source: Cailianshe)
Powered by AOYII.com
