meta, scale ai and google Updates

Meta Platforms Inc. is significantly overhauling its artificial intelligence strategy, moving away from its long-standing open-source approach. CEO Mark Zuckerberg initiated this shift after the Llama 4 open-source model did not meet performance expectations. The company's new proprietary AI model, codenamed Avocado, is now slated for an early 2026 launch, rather than late 2025, and will operate as a closed system. Alexandr Wang, the 28-year-old founder of Scale AI, now leads Meta's AI division as Chief AI Officer and supports this new direction. The TBD Lab group, led by Zuckerberg, is training Avocado using models from Google, OpenAI, and Alibaba. This strategic pivot includes a substantial investment in AI infrastructure, with Meta committing $600 billion for US infrastructure. The company has also secured cloud deals with CoreWeave Inc and Oracle Corp and is constructing a large Hyperion facility in Louisiana. To prioritize AI-driven devices and data centers, Meta cut its metaverse budget by 30% and acquired Limitless for AI wearables, signaling a clear focus on tangible AI applications and hardware. Meanwhile, Microsoft is making its largest investment in Asia, pledging $17.5 billion in India over four years, from 2026 to 2029. This funding will significantly boost India's cloud and artificial intelligence infrastructure, building on a previous $3 billion commitment. A new India South Central cloud region in Hyderabad is set to open in mid-2026, alongside expansions to three existing data centers. Microsoft also plans to train 20 million Indians with essential AI skills by 2030, doubling its earlier goal. Microsoft's AI assistant, Copilot, demonstrates versatile usage, according to its "It's About Time: The Copilot Usage Report 2025." Analyzing 37.5 million conversations, researchers found Copilot serves different roles based on device and time. On desktops, users primarily ask work and technical questions during business hours. However, on mobile devices, "Health and Fitness" combined with information-seeking is the most common topic, positioning Copilot as a private advisor for personal queries. The broader AI landscape sees significant developments in infrastructure and investment. AI hardware, such as GPUs and TPUs, generates intense heat, pushing data centers, especially at edge locations, towards liquid cooling solutions that can handle thermal loads exceeding 30 or even 50 kilowatts. Economically, Asian AI stocks currently offer cheaper investment options compared to their American counterparts, with US markets, including Nvidia and Microsoft, showing very high valuations. Innovation in AI hardware continues with Pebble launching a new $75 AI ring. This device focuses on core functions, sending audio to the Pebble App for local AI models to set reminders or perform tasks, avoiding cloud processing and privacy concerns. In the financial sector, Wall Street investment banks like Moelis and Houlihan Lokey are building extensive AI deal databases, integrating years of data to gain an advantage, offer better advice, and speed up transactions. Educational institutions are also adapting, with Fairfax County Public Schools introducing new courses in artificial intelligence, personal finance, and cybersecurity for middle and high school students starting in the 2026-27 school year. These AI courses will cover core concepts, ethical implications, and generative AI prompt creation. On a cautionary note, a TikTok creator in Florida faces accusations of using AI to create fake, frightening videos, highlighting the growing realism of AI-generated content and the need for viewer discretion. Finally, Channel3 secured $6 million in seed funding to develop a neutral, universal product catalog and API, aiming to make every product sold online discoverable by AI shopping agents like ChatGPT and Gemini.

Key Takeaways

• Meta is shifting its AI strategy from open-source to proprietary models, with the new 'Avocado' model launching as a closed system in early 2026.
• Alexandr Wang, founder of Scale AI, is now Meta's Chief AI Officer, leading the company's new closed-model AI approach.
• Meta is cutting its metaverse budget by 30% and investing $600 billion in US AI infrastructure, including new data centers and cloud deals with CoreWeave Inc and Oracle Corp.
• Microsoft is making its largest investment in Asia, pledging $17.5 billion in India's cloud and AI infrastructure over four years (2026-2029), and plans to train 20 million Indians in AI skills by 2030.
• Microsoft's Copilot AI assistant adapts its role based on device, serving as a work assistant on desktop and a personal confidant for health and fitness queries on mobile.
• AI hardware, such as GPUs and TPUs, demands advanced liquid cooling solutions for data centers due to significantly higher heat output compared to traditional processors.
• Asian AI stocks are currently more affordable than US counterparts like Nvidia and Microsoft, which are experiencing high valuations.
• Wall Street investment banks, including Moelis and Houlihan Lokey, are building AI deal databases to leverage years of transaction data for improved advice and faster deal execution.
• Fairfax County Public Schools will introduce new courses in artificial intelligence, personal finance, and cybersecurity for middle and high school students starting in the 2026-27 school year.
• Channel3 raised $6 million in seed funding to create a universal product catalog and API, enabling AI shopping agents like ChatGPT and Gemini to discover products from any online store.

Meta Overhauls AI and Hardware Plans

Meta Platforms Inc. is changing its artificial intelligence strategy and hardware plans. CEO Mark Zuckerberg is moving away from open-source AI, especially after Llama 4 did not perform well. Meta is developing a new AI model called Avocado, which will be proprietary and launch in early 2026 instead of late 2025. Alexandr Wang, 28-year-old founder of Scale AI, now leads Meta's AI division as chief AI officer. The company also bought Limitless for AI wearables and cut its metaverse budget by 30% to focus on AI-driven devices and data centers. Meta made cloud deals with CoreWeave Inc and Oracle Corp and is building a large Hyperion facility in Louisiana.

Meta Shifts to Closed AI Models for Profit

Meta Platforms Inc. is changing its AI strategy to focus on "closed" models that it can control and sell access to. This marks a big change from its long-standing open-source approach, and new Chief AI Officer Alexandr Wang supports this move. The company's new model, codenamed Avocado, will likely launch next spring as a closed system. CEO Mark Zuckerberg shifted the strategy after the Llama 4 open-source model did not meet expectations. Meta is investing heavily in AI, including $600 billion for US infrastructure, and cutting virtual reality and metaverse spending to prioritize AI glasses and hardware. The TBD Lab group, led by Zuckerberg, is training Avocado using models from Google, OpenAI, and Alibaba.

Meta Pivots to Profitable Closed AI Models

Meta Platforms Inc. is changing its AI strategy to focus on "closed" models that it can control and sell access to. This marks a big change from its long-standing open-source approach, and new Chief AI Officer Alexandr Wang supports this move. The company's new model, codenamed Avocado, will likely launch next spring as a closed system. CEO Mark Zuckerberg shifted the strategy after the Llama 4 open-source model did not meet expectations. Meta is investing heavily in AI, including $600 billion for US infrastructure, and cutting virtual reality and metaverse spending to prioritize AI glasses and hardware. The TBD Lab group, led by Zuckerberg, is training Avocado using models from Google, OpenAI, and Alibaba.

Microsoft Invests $17.5 Billion in India AI

Microsoft announced its largest investment in Asia, pledging $17.5 billion in India over four years from 2026 to 2029. This money will boost India's cloud and artificial intelligence infrastructure, improve skills, and support ongoing operations. The investment builds on a previous $3 billion commitment and aims to create secure, large-scale infrastructure for AI adoption. A major part of this plan is the new India South Central cloud region in Hyderabad, set to open in mid-2026. Microsoft will also expand its three current data centers in Chennai, Hyderabad, and Pune. The company plans to train 20 million Indians with essential AI skills by 2030, doubling its earlier goal.

Microsoft Pledges $17.5 Billion for India AI

Microsoft announced a massive $17.5 billion investment in India's cloud and artificial intelligence infrastructure. This investment will take place over the next four years and marks Microsoft's largest investment in Asia to date. The funds will help advance India's technology capabilities and support its growing AI sector.

TikTok Creator Uses AI to Scare Strangers

A TikTok content creator in Florida faces accusations of using artificial intelligence to make fake videos that scared strangers. West Palm Beach police say this is the fourth time the 22-year-old man created frightening AI scenarios. In one incident, he showed a gas station clerk an AI video of someone being dragged away, leading to an argument and his arrest for resisting an officer. He also showed Melanie Valentine a fake AI video of her truck being stolen and recorded a shopper, creating a fake video of him kissing another woman to show his wife. The man admitted he creates these videos for social media content, believing people will like them more if he gets arrested. While creating AI videos in public is not illegal without criminal intent, police warn that these videos are becoming very realistic and viewers should be careful.

Asian AI Stocks Offer Cheaper Investment Options

Asian artificial intelligence stocks are currently much cheaper than their American counterparts. This difference should concern American investors, as the tech market in Japan, South Korea, and Taiwan looks very different. While US markets are surging, stocks like Nvidia and Microsoft, along with the broader S&P 500 index, are very expensive. A popular measure, the cyclically adjusted price-earnings ratio, shows that the S&P 500 has only been this expensive once before, during the dotcom boom.

Fairfax County Schools Add AI and Tech Courses

Fairfax County Public Schools will offer new courses in artificial intelligence, personal finance, and cybersecurity starting in the 2026-27 school year. The Fairfax County School Board unanimously approved these changes for middle and high school students. New AI courses will cover core concepts, including ethical implications and how to create effective prompts for generative AI. Other new classes include Advanced Placement Business with Personal Finance, AP Cybersecurity, Advanced Topics in Chemistry, and Geology. The district also expanded its World Language offerings to include Arabic, Chinese, and American Sign Language, among others. Most new classes are optional, allowing schools to decide based on student interest, and virtual options may be explored for wider access.

Liquid Cooling Becomes Key for AI Data Centers

Artificial intelligence is creating new cooling challenges for critical digital infrastructure, especially at edge locations. AI hardware, like GPUs and TPUs, uses much more power and produces significantly more heat per rack than traditional processors. Existing air-based cooling systems often cannot handle these high thermal loads, which can exceed 30 or even 50 kilowatts in edge computing. This intense heat risks hardware failures, data loss, and higher operating costs, as air systems struggle to respond to unpredictable AI workloads. Many operators are now turning to liquid cooling, which is more efficient at removing heat directly from the source. Liquid cooling methods, including direct-to-chip and immersive solutions, are becoming more practical for smaller facilities due to advances in reliability and installation.

Microsoft Copilot Adapts to User Needs

Microsoft's new report, "It's About Time: The Copilot Usage Report 2025," shows how people use its AI assistant, Copilot. Researchers analyzed 37.5 million conversations from January to September, finding that Copilot serves different roles based on the device and time of day. On desktop computers, users mainly ask work and technical questions during business hours. However, on mobile devices, "Health and Fitness" combined with information-seeking is the most common topic, showing Copilot acts as a private advisor for personal questions. This suggests users see Copilot as both a colleague at their desk and a confidant in their pocket. The study also noted a rise in advice-seeking, highlighting the growing importance of accuracy and trust in AI tools.

Pebble Launches Simple and Smart AI Ring

A new $75 AI ring from Pebble is praised for its simple and effective hardware design. Unlike other AI gadgets that try to do too much, this ring focuses on core functions. When a user presses the button, the audio goes to the Pebble App, where local AI models can set reminders or perform other tasks. This design avoids cloud processing, subscriptions, and privacy concerns associated with "always on" listening devices. Eric Migicovsky, who founded Pebble, has made the system open for developers to add new features. The ring offers a convenient way to capture thoughts without the friction of unlocking a phone, though it faces competition from other smart rings like Sandbar.

Wall Street Banks Build AI Deal Databases

Investment banks on Wall Street are gathering years of deal data to gain an advantage using artificial intelligence. Firms like Moelis and Houlihan Lokey are integrating vast datasets into AI tools to win business, offer better advice, and speed up transactions. Moelis CEO Navid Mahmoodzadegan aims to make 18 years of data usable for AI, while Houlihan Lokey CEO Scott Adelson highlights their edge from large datasets. Houlihan Lokey even launched a private credit data platform using over 60,000 loan valuations. This shift raises questions about the future need for junior bankers, though firms like Moelis are still hiring large analyst classes. Leaders believe AI will greatly increase transaction speed and banker productivity, leading to better client service.

Channel3 Raises $6M for AI Shopping Platform

Channel3 raised $6 million in seed funding to make every product sold online discoverable by artificial intelligence agents. The startup aims to change online shopping, where AI tools like ChatGPT and Gemini can find, compare, and buy products for consumers. Currently, only tech giants control product data for AI shopping, but Channel3 offers a neutral, universal product catalog and API. This system uses advanced AI models to match user requests with products from any online store, regardless of how they are listed. Merchants can sync their catalogs with Channel3 to make their products visible to a growing number of AI shopping agents. The funding will help Channel3 expand its engineering team and computing power to process billions of tokens for product searches.

Sources

• Meta Resets AI, Hardware Plans As 'Avocado' Delays, Leadership Turmoil, Cost Cuts Mount
• Inside Meta’s pivot from open source to money-making AI model
• Inside Meta’s Pivot From Open Source to Money-Making AI Model
• Microsoft announces vast AI investment in India
• Microsoft Unveils Massive $17.5 Billion A.I. Investment in India
• TikTok content creator accused of using AI to create videos that panicked strangers
• Asia’s inexpensive AI stocks should worry American investors
• From cybersecurity to AI: Fairfax Co. unveils plan for new courses
• Cooling: keeping up with AI
• Microsoft says its Copilot AI tool is a ‘vital companion’ in new analysis of 37.5M conversations
• View: A new $75 AI ring finally gets hardware right
• Wall Street Builds Deal Datasets for AI Investment Banking Era
• Channel3 raises $6M to make every product sold on the web discoverable by AI agents