Last Updated on March 12, 2025 by Editor
Welcome to this week’s AI update, where we bring you the latest breakthroughs, major investments, and regulatory shifts shaping the artificial intelligence landscape. From multi-billion-dollar AI funding rounds to cutting-edge innovations in generative AI, robotics, and cybersecurity, this week has been packed with industry-defining developments.
In this edition, we explore Google’s latest AI advancements, OpenAI’s new business tools, and Meta’s push for in-house AI chips. We also dive into the EU’s evolving AI regulations, major startup funding, and groundbreaking AI applications in healthcare, security, and enterprise solutions.
Stay ahead of the curve with our comprehensive AI roundup, covering all the top stories from March 05–11, 2025. Let’s get started!
Google Deepens Ties with Anthropic Through Increased Funding
Google has increased its investment in Anthropic, revealing a deeper connection between the two companies than previously known. New court filings indicate that Google owns a 14% stake in the AI startup and is set to invest an additional $750 million via a convertible debt deal this year. This brings Google’s total investment in Anthropic to over $3 billion. Despite the absence of voting rights or direct control, this substantial financial backing raises questions about Anthropic’s independence in the competitive AI landscape.
The investment underscores Google’s strategy of supporting multiple players in the AI race while developing its own technologies. This approach allows Google to hedge its bets and remain agile in the rapidly evolving AI sector. Meanwhile, regulators are scrutinizing these deals to ensure fair competition, especially with Amazon also investing heavily in Anthropic, pledging up to $8 billion. The UK’s Competition and Markets Authority has initiated a formal investigation into Google’s partnership with Anthropic, citing potential competition concerns
Meta Tests In-House AI Chips to Reduce Reliance on Nvidia
Meta is reportedly testing its in-house AI chips, codenamed Artemis, to reduce its dependency on Nvidia’s GPUs for AI training. This move aims to cut the massive expenses associated with AI infrastructure, which are projected to reach up to $65 billion in 2025. The chips are designed to handle the massive computational demands of training large language models (LLMs) like Llama 3. According to sources, Meta aims to achieve a 40% reduction in training costs by using its custom silicon.
This move is part of Meta’s broader strategy to gain more control over its AI infrastructure and reduce operational costs. The company has been investing heavily in AI research, with a focus on metaverse applications and generative AI tools. If successful, Meta’s in-house chips could disrupt the AI hardware market, currently dominated by Nvidia, and set a new benchmark for cost-efficient AI training.
Google Introduces Gemma 3: Open-Source AI Model with 128K Context Window
Google has introduced Gemma 3, a new family of accessible, lightweight open-source AI models. These models are designed to be more accessible, and can run on a single graphics processing unit. Gemma 3 models come in various sizes, including 1 billion, 4 billion, 12 billion, and 27 billion parameters, allowing developers to choose the best fit for their hardware and performance needs. This allows the model to process and analyze significantly larger amounts of data in a single instance, making it ideal for complex tasks like document summarization, code generation, and long-form content creation.
Gemma 3 shares technical research with Google’s Gemini model, and is capable of delivering high performance for its size. It outperforms larger models like Llama-405B, DeepSeek-V3, and OpenAI’s o3-mini in preliminary human preference evaluations. Even when run on a single device or GPU, Gemma provides enough power for developers to create AI applications with multimodal capabilities, featuring a 128,000-token context window, comparable to OpenAI’s GPT-4o.
Flower Labs Introduces Seamless Local-to-Cloud AI Switching Service
Flower Labs has launched Flower Intelligence, a distributed cloud platform that seamlessly switches between local and cloud AI processing. This hybrid approach allows applications to default to a locally running AI model for speed and privacy, while automatically transitioning to Flower’s cloud when extra computational power is needed. Mozilla is using it to power the upcoming Assist summarization add-on for its Thunderbird email client. The new offering is expected to benefit industries like healthcare, finance, and manufacturing, where data privacy and real-time processing are critical. Flower Labs claims its service can reduce operational costs by up to 30% while maintaining high performance. This development marks a significant step forward in making AI more adaptable and accessible for businesses of all sizes.
Flower Intelligence supports on-device AI for mobile, PC, and web apps, handing off to a private cloud with user permission when necessary. The platform is built on open models, including those from Meta’s Llama family, DeepSeek, and Mistral. Flower Labs ensures data protection through end-to-end encryption and other techniques within its Flower Confidential Remote Compute service. Developers can apply for early access to Flower Intelligence, with plans for broader availability and additional features like model customization and federated training in the cloud.
Foxconn Unveils FoxBrain, Its Proprietary AI Model for Manufacturing
Foxconn, the world’s largest electronics manufacturer, has introduced FoxBrain, its proprietary AI model designed to optimize manufacturing processes. FoxBrain leverages advanced machine learning algorithms to enhance production efficiency, quality control, and supply chain management. The company plans to integrate the model across its global factories, aiming for a 20% increase in productivity by 2026.
This move highlights Foxconn’s ambition to transition from a traditional manufacturer to a technology-driven innovator. FoxBrain is expected to set a new standard for AI-powered manufacturing, enabling the company to maintain its competitive edge in the rapidly evolving tech landscape.
Dexterity Secures Major Funding to Revolutionize AI-Powered Robotics
Dexterity, an AI robotics firm specializing in industrial robots with “human-like” finesse, has secured $95 million in its latest funding round. The investment, which includes backing from Lightspeed Venture Partners and Sumitomo Corp, values the company at $1.65 billion post-money, according to Bloomberg. This funding highlights the strong investor interest in the convergence of robotics and artificial intelligence, as companies like Meta and Apple explore investments in AI-powered humanoid robots.
Dexterity’s robots are designed to perform repetitive and dangerous tasks in warehouses and factories, such as loading boxes and sorting parcels, for customers like FedEx and UPS1. According to CEO Samir Menon, the robots utilize specialized AI models, each focused on a specific task, coordinated by a central system. With this latest round, Dexterity has now raised nearly $300 million in total funding
OpenAI Introduces Advanced Tools Empowering Businesses to Develop Custom AI Agents
OpenAI has unveiled new tools to help developers and enterprises construct AI agents, which are automated systems capable of independently accomplishing tasks. These tools are part of OpenAI’s new Responses API, allowing businesses to develop custom AI agents that can perform web searches, scan company files, and navigate websites, similar to OpenAI’s Operator product. The Responses API effectively replaces OpenAI’s Assistants API, set to be discontinued in the first half of 2026.
The Responses API offers access to the same AI models that power OpenAI’s ChatGPT Search web search tool, including GPT-4o search and GPT-4o mini search (in preview). These models can browse the web for answers, citing sources as they generate replies. The API also includes a file search utility and access to OpenAI’s Computer-Using Agent (CUA) model, which powers Operator. Additionally, OpenAI is releasing an open-source toolkit called the Agents SDK, providing developers with free tools to integrate models, implement safeguards, and monitor AI agent activities. The platform includes pre-trained models, drag-and-drop interfaces, and API integrations, making it easier for businesses to tailor AI solutions to their specific needs. OpenAI claims that early adopters have seen a 40% improvement in operational efficiency. This move is part of OpenAI’s broader strategy to make generative AI more accessible and practical for enterprises, further cementing its dominance in the AI market.
EU AI Act Draft Proposes Gentler Guidelines for Large AI Models
The latest draft of the EU AI Act has introduced more lenient guidelines for large AI model developers, aiming to balance innovation with ethical considerations. The updated draft reduces the regulatory burden on companies developing foundation models, such as OpenAI’s GPT and Google’s Gemini, while maintaining strict oversight for high-risk applications like facial recognition and autonomous weapons.
The revised guidelines emphasize transparency, data privacy, and accountability, requiring companies to disclose training data sources and implement robust safety measures. Industry leaders have welcomed the changes, stating that they will foster innovation while ensuring responsible AI development. The EU AI Act is expected to be finalized by mid-2025, setting a global standard for AI regulation.
Sola Raises $30M to Build the Stripe of AI Security Solutions
Sola, a startup focused on AI-driven security solutions, has emerged from stealth mode with $30 million in funding. The company aims to become the Stripe of AI security, offering a unified platform for businesses to detect and mitigate cyber threats in real time. Sola’s technology leverages machine learning to analyze vast amounts of data and identify potential vulnerabilities before they can be exploited.
The funding round was led by Andreessen Horowitz, with participation from Greylock Partners. Sola plans to use the funds to expand its engineering team and accelerate product development. With cyberattacks on the rise, Sola’s innovative approach to AI-powered security is poised to become a game-changer for businesses worldwide.
Elon Musk’s DOGE Deploys GSAi Chatbot to Automate Government Tasks
Elon Musk’s Department of Government Efficiency (DOGE) has introduced GSAi, a custom chatbot, to 1,500 federal workers at the General Services Administration (GSA). This move is part of an effort to automate tasks previously performed by humans, coinciding with a reduction in the federal workforce. GSAi offers an interface similar to ChatGPT, allowing users to select from three AI models: Claude Haiku 3.5 (default), Claude Sonnet 3.5 v2, and Meta LLaMa 3.2.
GSAi assists with drafting emails, creating talking points, summarizing text, and writing code. However, employees are instructed not to input “federal nonpublic information” or “controlled unclassified information” into the system. While some view GSAi as a tool to streamline operations, others worry it may lead to further layoffs. DOGE plans to deploy the AI chatbot across the entire agency.
Gleamer Expands Radiology AI Capabilities with MRI-Focused Acquisitions
Gleamer, a French startup specializing in AI-enhanced radiology software, is expanding its focus to include magnetic resonance imaging (MRI) through the acquisitions of Caerus Medical and a merger with Pixyl. The acquisitions, valued at a combined $50 million. These strategic moves position Gleamer in the second wave of startups aiming to advance medical imaging using artificial intelligence. Having already developed AI tools for X-rays and mammograms, Gleamer aims to enhance diagnostic precision in MRI analysis.
Established in 2017, Gleamer has developed an AI assistant that supports radiologists in medical imaging, functioning as a co-pilot to improve diagnostic accuracy. The company’s software has been adopted by 2,000 institutions across 45 countries, processing 35 million examinations. Gleamer has also obtained CE and FDA certifications for its bone trauma interpretation product and offers CE-certified solutions for chest X-rays, orthopedic assessments, and bone age evaluations in Europe.
OpenAI Invests $12 Billion in CoreWeave, Intensifying Competition with Microsoft
OpenAI has entered into a five-year agreement with CoreWeave, a GPU cloud service provider, investing $11.9 billion. As part of the deal, OpenAI will acquire $350 million in equity from CoreWeave, separate from CoreWeave’s upcoming IPO. This partnership is significant considering that Microsoft was CoreWeave’s largest client, accounting for 62% of its $1.9 billion revenue in 2024, a substantial increase from $228.9 million in 2023.
Supported by Nvidia, which holds a 6% stake, CoreWeave specializes in AI cloud services, with 32 data centers and over 250,000 Nvidia GPUs by the end of 2024. This move allows OpenAI to secure access to cloud resources while also gaining ownership in the provider. As Microsoft advances its own AI models and has ended its exclusive role as OpenAI’s cloud provider, this investment marks a new phase in the complex relationship between the two companies.
Gmail Introduces AI-Powered “Add to Calendar” Feature with Gemini Integration
Gmail has introduced a new feature powered by Google’s Gemini AI, allowing Google Workspace users to add events directly to Google Calendar from an email. With this update, Gemini automatically detects calendar-related content in an email and presents an “Add to calendar” button. Clicking this button opens a side panel in Gmail to confirm the event addition.
This feature is currently available in English on the web for Google Workspace Business and Enterprise tiers, as well as for customers with a Gemini Education, Gemini Education Premium, or Google One AI Premium plan. Admins can enable the “Add to calendar” function by turning on smart features and personalization in the Workspace Admin console. The new function is part of a series of Gemini-powered tools in Gmail, including capabilities to write emails, summarize threads, and find specific information.
Altera Launches Agilex 3 FPGAs for AI-Driven Edge Computing
Altera, a subsidiary of Intel, has unveiled its Agilex 3 FPGAs (Field-Programmable Gate Arrays), designed specifically for AI applications at the edge. These next-generation chips offer 40% higher performance and 30% lower power consumption compared to previous models, making them ideal for real-time AI processing in industries like autonomous vehicles, smart manufacturing, and IoT devices.
The Agilex 3 FPGAs are equipped with AI-optimized cores and support machine learning frameworks like TensorFlow and PyTorch. Altera claims that early adopters have seen a 50% reduction in latency for edge AI tasks. This launch reinforces Altera’s position as a leader in edge computing hardware and underscores the growing importance of AI at the edge in enabling faster, more efficient decision-making.
AvatarOS Secures $7 Million Seed Funding to Revolutionize Virtual Influencer Market
AvatarOS, a startup specializing in AI-driven virtual influencers, has emerged from stealth mode, securing $7 million in seed funding led by M13, with participation from Andreessen Horowitz Games Fund and other investors. Founded by Isaac Bratzel, known for creating the virtual influencer Lil Miquela, AvatarOS aims to develop digital avatars for platforms including social media and gaming. Leveraging proprietary 4D/ML training techniques, the company seeks to produce lifelike avatars capable of authentic audience engagement. This advancement addresses previous challenges in animation costs and technical limitations, positioning AvatarOS to scale character-driven digital content across multiple platforms.
Microsoft Explores 3D Gaming Experiences for Copilot, Seeking Talent in China
Microsoft is venturing into 3D gaming experiences for its AI-powered Copilot platform, signaling a strategic expansion into interactive entertainment. A recent job listing indicates that Microsoft is seeking a senior software engineer based in Beijing. The candidate should specialize in 3D rendering engines, particularly those commonly used for web browser-based video games such as Babylon.js, three.js, and Unity.
The job posting expresses enthusiasm for gaming and creating groundbreaking solutions for billions of users. Previously, Microsoft showcased an AI model named Muse, intended to power short interactive games on Copilot. Muse was trained using the multiplayer battle arena game Bleeding Edge, enabling it to understand and generate gameplay rendered by AI. This initiative aligns with Microsoft’s broader strategy to integrate gaming more extensively into the Copilot experience.
ServiceNow Acquires Moveworks for $2.85B to Expand AI Capabilities
ServiceNow, a leader in enterprise workflow automation, has acquired Moveworks, an AI-powered IT support platform, for $2.85 billion. The acquisition will enhance ServiceNow’s AI portfolio by integrating Moveworks’ advanced natural language processing (NLP) and machine learning capabilities into its platform. Moveworks, which boasts over 5 million users and $100 million in annual recurring revenue, will augment ServiceNow’s capabilities in customer service management and AI-driven solutions. Despite the acquisition’s potential, ServiceNow’s stock experienced a 7.8% decline following the announcement, reflecting investor caution regarding the deal’s valuation.
By integrating Moveworks, ServiceNow intends to provide a unified search and self-service experience for employees across various workflows. Moveworks’ AI-driven assistants serve major clients such as Broadcom, Palo Alto Networks, and Pinterest. ServiceNow plans to combine its automation strengths with Moveworks’ AI assistant and enterprise search technology to drive enterprise-wide AI adoption and deliver improved outcomes for employees and customers.
Cursor AI in Talks to Raise Funds at $10B Valuation Amid Coding Boom
Anysphere, the company behind the AI-powered coding assistant Cursor, is reportedly in talks to raise funding at a valuation of nearly $10 billion. This potential valuation marks a fourfold increase from its $2.6 billion valuation just three months prior, underscoring the surging investor interest in AI coding tools. The funding round is reportedly led by existing investor Thrive Capital. Cursor’s tools are already used by over 1 million developers worldwide, including teams at Google, Amazon, and Microsoft.
Cursor AI has emerged as a key player in the rapidly growing AI coding landscape, offering developers intelligent features that streamline the coding process. Built on Visual Studio Code and powered by advanced AI models like GPT-4 and Claude, Cursor provides AI-powered code completion, intelligent error detection, and natural language processing for plain English commands. The company’s rapid ascent is attributed to Cursor’s widespread adoption among developers and impressive revenue growth.
Google Debuts Gemini-Based Text Embedding Model for Enhanced AI Performance
Google has introduced a new, experimental “embedding” model for text, Gemini Embedding, to its Gemini developer API. Trained on the Gemini model, this embedding model inherits Gemini’s understanding of language and nuanced context, making it applicable for a wide range of uses. The company noted that Gemini Embedding is in an “experimental phase” with limited capacity and is working towards a stable, generally available release in the months to come.
Gemini Embedding is designed to deliver exceptional performance across diverse domains, including finance, science, legal, search, and more. Companies including Amazon, Cohere, and OpenAI offer embedding models through their respective APIs. Gemini Embedding is Google’s first embedding model trained on the Gemini family of AI models.
Elon Musk’s xAI Expands AI Data Center Footprint with Memphis Property Acquisition
xAI, Elon Musk’s AI company, has acquired a 1 million-square-foot property in Southwest Memphis to expand its AI data center footprint. This acquisition will host infrastructure to complement xAI’s existing Memphis data center. The expansion is part of xAI’s plans to scale its Colossus supercomputer and increase its GPU count from 100,000 to over a million.
The new facility includes an $80 million water recycling plant and the world’s largest deployment of Tesla Inc. Megapacks for data center operations. xAI uses its data centers to train and run its family of AI models, Grok. To fund its AI infrastructure projects, xAI is reportedly discussing a $10 billion round of fundraising that would value the company at $75 billion.
Peer Secures $10.5 Million to Launch 3D Personal Planets in the Metaverse
Peer Global Inc has successfully raised $10.5 million in its latest funding round to advance its metaverse engine and introduce 3D personal planets, offering users personalized 3D social hubs within the metaverse. This investment, led exclusively by the Family Office of Tommy Mai, brings Peer’s total funding to $65.5 million, all sourced from angel investors. The new feature aims to transform social engagement by moving beyond traditional, static social media platforms, encouraging users to explore and interact within dynamic, AI-driven environments. Founder Tony Tran envisions these personal planets as catalysts for spontaneous interactions and collaborations, leveraging location-sharing and AI capabilities to create a living map where users can engage in real-time. This initiative underscores Peer’s commitment to redefining social experiences in the digital age.
Larry Page’s New AI Startup ‘Dynatomics’ Aims to Revolutionize Air Travel
Larry Page, co-founder of Google, has launched a new startup named Dynatomics, focusing on leveraging AI and 3D printing to innovate in the aviation sector. Based in Palo Alto, California, and led by Chris Anderson, former CTO of Kittyhawk, Dynatomics aims to utilize additive manufacturing to enhance the production of aircraft parts, potentially reducing costs and production times. The company is collaborating with researchers from Tallinn University of Technology in Estonia to advance manufacturing processes for electric vertical take-off and landing (eVTOL) aircraft. This venture reflects Page’s ongoing commitment to democratizing aviation and developing cost-effective flying solutions.
Scale AI Under Investigation by U.S. Department of Labor
The U.S. Department of Labor has initiated an investigation into Scale AI, a data-labeling startup backed by prominent investors including Nvidia, Amazon, and Meta, to assess compliance with the Fair Labor Standards Act. This probe, which began nearly a year ago, aims to ensure fair pay practices and appropriate working conditions for contributors. Founded in 2016, Scale AI plays a crucial role in providing accurately labeled data essential for training AI models, serving clients such as OpenAI, Cohere, Microsoft, and Morgan Stanley. The company asserts that it maintains fair pay practices, with timely payments and a high rate of resolving payment inquiries promptly. Valued at $14 billion in its latest funding round, Scale AI’s operations are now under scrutiny to ensure adherence to labor regulations.
Mistral AI’s New OCR API Converts PDFs into AI-Ready Markdown Files
Mistral AI has launched a new Optical Character Recognition (OCR) API that converts PDF documents into AI-ready text formats such as Markdown or raw text files. This tool enables AI models to analyze and process PDF documents, extracting data and making it digestible for AI applications. The Mistral OCR API is designed to understand separate elements in documents, including media, text, tables, and equations, with high accuracy. The tool supports multilingual text extraction and offers 95% accuracy, significantly reducing the time and effort required for data preprocessing.
The API solves the challenge of PDF content being inaccessible to large language models (LLMs) using traditional Retrieval-Augmented Generation (RAG) techniques. By converting PDFs into structured outputs, the Mistral OCR API allows developers to build AI applications for PDF analysis and create datasets to train new AI models. The company claims that the tool can understand separate elements in documents, including media, text, tables, and equations with high accuracy. This capability enables users to extract information from PDFs and format it in structured outputs, making it ideal for use in combination with a RAG system taking multimodal documents as input.
Alibaba’s Qwen QwQ-32B Achieves Top-Tier Performance Through Scaled Reinforcement Learning
Alibaba’s Qwen team has introduced QwQ-32B, a 32 billion parameter AI model that rivals the performance of the much larger DeepSeek-R1. This achievement underscores the potential of scaling Reinforcement Learning (RL) on robust foundation models. QwQ stands for Qwen-with-Questions, and was developed as an open-source alternative to OpenAI’s o1 reasoning model.
QwQ-32B integrates agent capabilities, enabling it to think critically, utilize tools, and adapt its reasoning based on environmental feedback. Trained with outcome-based rewards, the model reviews and reformulates its responses until it reaches the correct answer. It nearly matches the performance of DeepSeek-R1-671B on math and coding benchmarks. The model’s context length has been expanded to 131,072 tokens, similar to other reasoning models such as Claude 3.7 Sonnet and Gemini 2.0 Flash Thinking.
Faireez Raises $7.5M for AI-Powered Hotel-Style Housekeeping for Condo Owners
Faireez, a startup offering AI-powered housekeeping services, has raised $7.5 million in a funding round led by Bessemer Venture Partners. The company uses AI-driven scheduling and robotic cleaning tools to provide hotel-style housekeeping for condo owners and property managers.
Faireez’s platform optimizes cleaning schedules based on occupancy data and user preferences, reducing operational costs by 30%. The service is currently available in 10 major cities, with plans to expand to 20 more by the end of 2025. This funding round highlights the growing demand for AI-powered home services and the potential for automation to transform the property management industry.
Turing Raises $111 Million at $2.2 Billion Valuation, Powering AI Coding for OpenAI and LLM Producers
Turing, a leading provider of AI-driven coding solutions, has successfully secured $111 million in a Series E funding round, elevating its valuation to $2.2 billion. The company, which supplies code generation tools to OpenAI and other large language model (LLM) producers, plans to use the funds to expand its developer platform and enhance its AI-driven coding capabilities.
Turing’s tools are used by over 500,000 developers worldwide, helping them write and debug code 50% faster. The company’s success underscores the growing importance of AI in software development and its potential to revolutionize coding workflows. This funding round positions Turing as a key player in the AI-powered developer tools market.
Crogl Launches AI-Powered “Iron Man Suit” for Security Analysts with $30M Funding
Crogl, a cybersecurity startup, has launched an AI-powered “Iron Man Suit” for security analysts, backed by $30 million in funding. The platform integrates machine learning and natural language processing (NLP) to help analysts detect and respond to cyber threats in real time.
The tool, which offers 90% accuracy in threat detection, is already being used by financial institutions and government agencies. Crogl claims that its platform reduces response times by 40%, making it a game-changer for cybersecurity operations. This innovation highlights the growing role of AI in cybersecurity and its potential to enhance threat intelligence and incident response.
25% of Startups in YC’s Current Cohort Use AI-Generated Codebases
A quarter of startups in Y Combinator’s (YC) current cohort are using AI-generated codebases to accelerate their development processes. These startups leverage AI coding tools like GitHub Copilot and Turing to build and deploy applications 50% faster than traditional methods.
YC’s data shows that startups using AI-generated code are 30% more likely to secure funding, highlighting the growing acceptance of AI in software development. This trend underscores the transformative potential of AI-powered coding tools and their ability to democratize software development for startups and small businesses.
As we wrap up Week # 10 of 2025, it’s clear that the AI revolution is accelerating at an astonishing pace. From Alibaba’s reinforcement learning breakthroughs to Turing’s $111 million funding and the ethical scrutiny facing Scale AI, this week has been a testament to the transformative power of artificial intelligence. These developments are not just shaping industries—they’re redefining how we live, work, and interact with technology.
But this is just the beginning. Next week promises even more groundbreaking updates, with new innovations, strategic partnerships, and regulatory shifts on the horizon. Don’t miss out on the latest insights and trends that could impact your business or spark your curiosity. Subscribe now to stay ahead of the curve and be the first to know what’s next in the world of AI.
The future is here, and it’s powered by AI. Will you be part of it?
