Pulse

A report highlights six critical chokepoints in the AI supply chain, emphasizing China's dominance in processing 90% of rare earth minerals. The analysis maps the entire process from mining to AI model development, underscoring geopolitical control over essential components like GPUs, HBM chips, and data center cooling systems. AI

Nvidia's reliance on Asian supply chains for components has surged to 90% of its production costs, a significant increase from 65% a year ago. This heightened exposure is driven by the growing demand for its physical AI hardware, including the Jetson Thor robotics platform and DRIVE AGX Thor automotive SoC, which compete for constrained resources like TSMC's 3nm wafer capacity and LPDDR5X memory. The company's efforts to build domestic manufacturing capacity are underway but not yet at scale, while existing Asian suppliers face memory shortages impacting older product lines. AI

Google has explained that AICore, an Android system component, can temporarily occupy several gigabytes of storage by retaining both old and new AI models during updates. This process, intended for rollback safety, lasts about three days and is a sign of increasing storage needs for on-device AI features. Separately, leaks suggest the upcoming Pixel 11 will feature a new Tensor G6 chip with a MediaTek modem and updated AI and image processing units, alongside camera upgrades and a novel 'Pixel Glow' notification light. AI

AI-powered desktop applications are emerging as the new operating system for knowledge work, integrating with existing tools like email and calendars. Companies like OpenAI, Anthropic, and Cursor are developing unified platforms that handle coding, planning, and tracking tasks. These applications aim to streamline workflows by connecting directly to user data and offering advanced agentic capabilities, potentially redefining office software for the next decade. AI

Notion has launched a new developer platform to integrate AI agents and external data sources directly into its workspace. This platform allows teams to build automated, multi-step workflows by connecting various tools and databases. The new features include 'Workers' for running custom code in a secure sandbox and enhanced agent capabilities that can interact with external AI tools, positioning Notion as a central hub for agentic collaboration. AI

The Model Context Protocol (MCP) is emerging as a standard for connecting AI applications to external data and tools, enabling models like Claude and ChatGPT to access information and perform tasks. Several articles highlight MCP's role in bridging the gap between AI capabilities and real-world data access, emphasizing the need for secure and controlled connections, especially when interacting with sensitive databases. Tools like APIKumo are automating the creation of MCP endpoints for APIs, while Conexor provides infrastructure for secure database and API connections, underscoring the protocol's growing importance in making AI more functional and integrated. AI

Nvidia CEO Jensen Huang announced a partnership with Corning to boost US AI infrastructure, focusing on optical connections to meet escalating computational demands. This collaboration aims to significantly increase US optical fiber production capacity. Meanwhile, the US has accused China of large-scale industrial campaigns to steal AI secrets, a claim China denies as slander. Separately, the US is seeing a surge in local bans on new data center construction due to concerns over resource strain and environmental impact. AI

OpenAI's GPT-5.5 model is powering new capabilities in coding and environmental science. Developers are utilizing GPT-5.5 through tools like Codex for tasks such as dataset creation, model training, and software development. Additionally, NVIDIA is integrating GPT-5.5 into its infrastructure, notably within its Earth-2 climate simulation platform and for AI-driven environmental protection projects. AI

Anthropic has committed to spending approximately $200 billion over the next five years with Google Cloud, securing 5 gigawatts of next-generation TPU compute capacity starting in 2027. This deal, which represents over 40% of Google Cloud's current backlog, also includes a potential additional investment of up to $40 billion from Google. The agreement positions Google's custom TPUs as a significant competitor to NVIDIA's GPUs and highlights Anthropic's rapid revenue growth, which has surged to an annualized $30 billion. AI

Amazon is significantly deepening its partnership with Anthropic through a substantial investment and a long-term cloud computing commitment. This move, totaling up to $33 billion in investment and $100 billion in AWS spending over 10 years, positions Anthropic as a primary infrastructure user for Amazon's custom AI chips like Trainium. The deal contrasts with Amazon's conditional investment in OpenAI, highlighting a strategic focus on Anthropic for its core AI ecosystem while using OpenAI as a hedge against Microsoft's dominance. AI

Nvidia is shifting its focus in AI infrastructure from raw compute power to the cost per token, arguing that this metric better reflects business value and profitability. The company is also making significant investments in the physical infrastructure required for AI, including a multi-billion dollar partnership with IREN to deploy data centers and a substantial investment in Corning to expand domestic optical fiber production. These moves highlight Nvidia's strategy to control the entire AI stack, from chips to the underlying physical infrastructure, to ensure efficient and scalable AI deployments. AI

Amazon's CEO Andy Jassy highlighted the company's significant pivot towards generative AI in his 2026 shareholder letter, signaling a substantial increase in capital expenditures to meet surging demand for AI infrastructure and model training. This strategic shift positions Amazon to challenge major players like SpaceX, Nvidia, and Google, with the company's stock rising over 13% following the letter's release. The move underscores the transformative potential of the generative AI era for established tech giants. AI

Major tech companies including Google, Microsoft, Amazon, and Meta are planning a combined capital expenditure of $725 billion in 2026, a significant increase from the previous year. This massive investment is driven by the race to build AI infrastructure, with a substantial portion allocated to memory and chip costs. Despite the enormous spending, there are concerns about profitability, a lack of technical moats, and the potential for a price war in the AI sector, alongside local resistance in India to data center development due to land acquisition and environmental issues. AI

Google has developed a new algorithm called TurboQuant that significantly reduces the memory requirements for large language models, by up to six times. This development is impacting memory chip manufacturers like Samsung, SK Hynix, and Micron, potentially affecting their market value. The broader AI industry's insatiable demand for memory is driving up costs for various computing components and straining data center resources. AI

SpaceX has filed plans for a massive semiconductor fabrication facility, dubbed Terafab, in Grimes County, Texas. The initial phase is projected to cost $55 billion, with the total investment potentially reaching $119 billion upon completion of all planned expansions. This facility aims to produce advanced chips for AI servers, satellites, and autonomous vehicles, addressing Elon Musk's concerns about the pace of chip production meeting the demands of his companies. AI

Nscale, a UK-based AI infrastructure startup, has secured $790 million in debt financing to build an AI data center in Narvik, Norway. This facility was previously intended for OpenAI's Stargate Norway project. Microsoft is set to rent Nvidia chips at this new data center. Nscale's latest valuation stands at $14.6 billion following a $2 billion Series C funding round. AI

The Model Context Protocol (MCP) is emerging as a standardized interface for AI agents to interact with external tools and data. Several open-source projects and platforms are facilitating this, including Databricks' MCP Marketplace for real-time intelligence, Apify's `mcpc` CLI for universal MCP access, and Klavis AI's SDKs for integrating MCP servers. These developments aim to enable agents to access live data, perform complex tasks, and even engage in inter-agent communication and payments, moving towards a more robust and interconnected AI ecosystem. AI

OpenAI, Anthropic, and Block have co-founded the Agentic AI Foundation (AAIF) under the Linux Foundation to provide open standards for interoperable agentic AI systems. OpenAI is contributing its AGENTS.md format to the foundation to ensure long-term support and adoption. This initiative aims to prevent fragmentation in the rapidly developing agentic AI ecosystem as these systems move into real-world production. The move is supported by major tech companies including Google, Microsoft, and AWS. AI

Anthropic has launched ten specialized AI agents designed for financial services, aiming to automate tasks like financial statement auditing and client presentation drafting. This move coincides with a significant shift in investor sentiment, with demand for Anthropic's equity surging while interest in OpenAI's shares wanes. Anthropic is also making substantial investments in AI infrastructure, including a $50 billion commitment to U.S. data centers and a partnership with SpaceX for orbital compute capacity. AI

OpenAI is expanding its global AI infrastructure through the "Stargate" initiative, establishing partnerships in the UK, Norway, and the UAE. These collaborations aim to build sovereign AI capabilities by providing local compute power and access to advanced GPUs. The Stargate projects involve significant investments in data centers, leveraging renewable energy where possible, and are designed to support national AI strategies, boost economic growth, and enhance technological competitiveness. AI

Meta is undergoing significant workforce reductions, with approximately 8,000 employees being laid off and 6,000 open positions eliminated. CEO Mark Zuckerberg has framed these layoffs as a necessary reallocation of resources, with the cost savings directly funding the company's substantial investments in AI infrastructure and development. This strategic shift prioritizes capital expenditure on AI, particularly GPUs and power, over personnel costs, a trend also observed at other major tech companies like Amazon, Microsoft, and Google. AI

OpenAI has introduced AgentKit, a suite of tools designed to streamline the development, deployment, and optimization of AI agents. This toolkit includes an Agent Builder for visual workflow creation, a Connector Registry for managing data sources, and ChatKit for embedding agentic UIs. Google DeepMind has also unveiled two AI agents: CodeMender, which automatically patches software vulnerabilities, and AlphaEvolve, an agent that uses Gemini models to discover and optimize algorithms for applications in mathematics and computing. Additionally, OpenAI's Computer-Using Agent (CUA) demonstrates advanced capabilities in interacting with digital interfaces, setting new benchmark results for computer use tasks. AI