Amazon, Microsoft, and Alphabet's clouds grew fast on AI demand, but huge infrastructure spending squeezed cash flow, and Samsung Electronics and SK Hynix shares fell in the aftermath. Here's a roundup of OpenAI's model behavior guidelines overhaul and stronger agent monitoring, Google's Gemini 4 Carbon topping the benchmark, Mistral's trillion-parameter open-weights release, the Canadian government's equity investment in Cohere, Samsung's first AI glasses launch schedule, and FuriosaAI's $620 million raise.
📊 Big Tech Earnings: Cloud Growing Fast, AI Spending Squeezing Cash Flow… Korean Chip Stocks Also Fall
Microsoft reported revenue of $92.3 billion for the quarter ending in September (up 19% year over year) with Azure up 48%, and Alphabet posted revenue of $127.58 billion (up 25%) with Google Cloud revenue doubling to $30.7 billion. Amazon also reported revenue of $201.9 billion (up 12%) with AWS revenue up 39% to $46 billion, its fastest growth in about five years — but Alphabet's quarterly capital expenditure of $51.4 billion turned free cash flow negative, sending its shares down sharply, and on October 30 Samsung Electronics and SK Hynix shares fell about 4% and 7%, respectively.
🛡️ OpenAI Overhauls Model Behavior Guidelines Around Five Values… Also Reveals Agent Escape Monitoring System
On October 29, OpenAI overhauled the Model Spec, the public guidelines that define how its models should behave, restructuring it around five core values: fidelity to human intent, minimizing harm, honesty and transparency, human agency, and beneficence. The company explained that models should grasp what users, developers, and OpenAI intend rather than follow instructions literally, and said current models do not yet fully reflect the new guidelines. Separately, OpenAI said it has expanded a monitoring system that detects unauthorized agent behavior; when past Hugging Face-related incident records were replayed through the new system, it reportedly sounded an alarm about 30 minutes after the agent escaped its restricted environment.
🏆 Google's Gemini 4 Carbon Tops New AI Intelligence Index… Ahead of Claude and GPT-6
In the Intelligence Index v5 published by Artificial Analysis on October 27, Google's Gemini 4 Carbon ranked first with 59 points, followed by Anthropic's Claude Opus 5.5 at 58 and OpenAI's GPT-6 Astra at 57. This revision is notable for moving 60% of the evaluation to a private test set, reducing reliance on data scraped from the web, and combines nine evaluations covering knowledge work, coding, and science. The index also showed that the gap between open-weight and proprietary models has widened sharply.
French AI company Mistral on October 29 released the weights for Mistral Large 4, a mixture-of-experts model with about one trillion parameters, on Hugging Face. The model takes text and images as input and outputs text and code, handles a 1-million-token context, and can be used, modified, and commercialized freely under a modified MIT-style license — though businesses reselling the model as a service need a separate commercial license. Mistral positions the model for cyber defense, saying it refused 96.6% of malicious cyber requests, and said a version for penetration testing that could sharpen offensive capabilities will be offered only to verified partners.
🇨🇦 Canada Takes Up to C$1.6 Billion Equity Stake in Cohere… Building 'Sovereign AI'
The Canadian government announced on October 29 it will invest up to C$1.6 billion (about US$1.1 billion) in stages to take an equity stake in AI company Cohere, Prime Minister Mark Carney announced in Montreal. The investment is part of Cohere's Series E, which aims to raise up to C$7 billion at a valuation of about C$35.7 billion, led by the Schwartz Group with Nvidia, Bell, and others participating. In return, Cohere agreed to keep its headquarters and corporate registration in Canada, invest at least C$1.7 billion in Canadian computing infrastructure by 2030, and create more than 1,000 technology jobs.
👓 Samsung to Launch First AI Glasses 'Intelligent Eyewear' in Korea and the U.S. in Mid-December
Samsung Electronics will sell its first AI smart glasses, 'Intelligent Eyewear', starting in mid-December in Korea and the United States, industry sources were cited as saying on October 30. The product is based on Google's Android XR platform and Gemini AI, runs on a Qualcomm Snapdragon AR1 Gen 1 chip, and lasts up to 9 hours, with design by Gentle Monster and Warby Parker. The price has not yet been set, and other major markets, including Europe, China, and Japan, are excluded from the first launch.
💰 FuriosaAI Raises $620 Million in Series D… Valuation About $2.9 Billion
Korean AI semiconductor company FuriosaAI announced on October 29 that it closed a Series D raise of 820 billion won (about $620 million). The valuation was about 3.8 trillion won (about $2.9 billion), with the National Growth Fund the largest investor at 370 billion won and overseas investors adding $110 million. The funds will go toward mass production and expanded supply of its second-generation AI inference chip, RNGD, strengthening its software ecosystem, and developing a third-generation accelerator with Broadcom. The company plans to produce 20,000 RNGD units this year and raise that to more than 40,000 next year.