Anthropic announced its annualized revenue run rate has surged from approximately $9 billion at the end of 2025 to over $30 billion, exceeding OpenAI's level of about $24 billion. The number of enterprise customers with over $1 million in annual spending more than doubled in less than two months, surpassing 1,000. Broadcom disclosed in a regulatory filing that Google has commissioned it to develop next-generation custom TPUs and signed a supply agreement extending to 2031. Anthropic plans to utilize TPU compute equivalent to 3.5 gigawatts starting in 2027. Anthropic employs a multi-cloud strategy, using Google TPUs, AWS Trainium, and NVIDIA devices simultaneously. Broadcom's CEO expects AI chip sales could exceed $100 billion next year.

Anthropic has released its most powerful model, Claude Mythos Preview, capable of autonomously linking 3-5 minor vulnerabilities into critical exploit chains, having already discovered thousands of high-risk vulnerabilities across all major operating systems and browsers. Due to its exceptionally strong cybersecurity capabilities, Anthropic has decided not to publicly release the model, instead offering restricted access through Project Glasswing to over 40 partner organizations including Microsoft, Google, Apple, and the Linux Foundation, along with $100 million in credits to support infrastructure remediation. Meanwhile, Mythos Preview is now available in Private Preview on Google Cloud Vertex AI. Security researchers report discovering more vulnerabilities in weeks than they had in their entire careers previously.

Zhipu AI (Z.ai) has launched the 754-billion-parameter GLM-5.1 model under the MIT open-source license. The model achieves 58.4 on SWE-bench Pro, surpassing GPT-5.4 (57.7) and Claude Opus 4.6 (57.3). Designed for long-duration autonomous tasks, GLM-5.1 can operate continuously for up to 8 hours, utilizing a unique 'stepwise optimization' approach that alternates between incremental fine-tuning and structural breakthroughs to avoid performance plateaus. Zhipu adopts a hybrid commercial strategy, open-sourcing its flagship GLM-5.1 while retaining commercial versions like GLM-5 Turbo. The model is now available on Vercel AI Gateway, supporting unified API calls.

Intel announced it is joining Elon Musk's Terafab artificial intelligence chip project, collaborating with SpaceX and Tesla to develop advanced processors for humanoid robots and data centers. The project plans to construct two cutting-edge semiconductor fabrication plants in Austin, Texas, aiming to deliver one terawatt of computing capacity annually. Intel CEO Lip-Bu Tan stated that Musk has a track record of industry transformation, calling this move crucial for semiconductor manufacturing. Intel's stock rose nearly 3% following the announcement. This partnership is seen as a key step for Intel to reverse its lag in the AI chip market.

Reports indicate that OpenAI Chief Financial Officer Sarah Friar has been excluded by CEO Sam Altman from certain financial discussions due to her reservations about the IPO timeline and large spending initiatives. This development comes after OpenAI completed a $122 billion funding round, reaching an $852 billion valuation. In parallel, OpenAI is testing its next-generation Image V2 image generation model, showing improvements in UI design rendering and composition understanding. Additionally, OpenAI is urging investigations by California and Delaware into alleged anti-competitive behavior by Musk.

The AI Alliance announced the launch of Project Tapestry on April 8, aiming to build a global federated open-source platform that supports open and sovereign AI development. The project uses a distributed architecture enabling institutions worldwide to collaboratively train state-of-the-art open models without centralizing data or compute, while retaining control over their own data and derivative models. Turing Award winner Yann LeCun has joined as Chief Scientific Advisor. The alliance plans to host a technical workshop in Paris on May 7–8 to finalize the architecture and roadmap. The AI Alliance currently includes over 200 member organizations across 29 countries.

On April 7, 2026, the Japanese government passed an amendment to the Personal Information Protection Act to promote AI development. The revision relaxes requirements for acquiring personal information in AI development, lowering barriers for data utilization in AI model training. It also introduces a penalty fee system for businesses that profit from mass sale of personal data or severely infringe upon individual rights. This reflects Japan's policy direction to balance generative AI advancement with personal privacy protection, encouraging AI industry competitiveness while strengthening deterrence against improper data use.

JPMorgan CEO Jamie Dimon stated in his letter to investors that the five hyperscalers—Microsoft, Amazon, Google parent Alphabet, Meta, and Apple—are expected to increase their AI-related capital expenditures from $450 billion in 2025 to $725 billion in 2026. This investment will primarily fund data center upgrades, benefiting data center REITs such as Equinix and Digital Realty, as well as AI chipmakers including NVIDIA, Broadcom, and TSMC. Traditional cloud software companies like Salesforce and ServiceNow may face challenges as they struggle to keep pace with generative AI advancements.

Block has launched Managerbot, a proactive AI agent for Square merchants. Unlike traditional passive chatbots, Managerbot autonomously monitors business data and proactively delivers actionable recommendations for inventory management, scheduling, and marketing. To ensure trust and safety, all 'write' operations require explicit merchant approval and are previewed via a visual UI. This product serves as a core demonstration of Block's shift toward a 'smart company' strategy following a 4,000-person workforce reduction, incentivizing merchants to migrate more business functions to Square to enhance contextual awareness, create network effects, and strengthen platform stickiness.

Google has launched Jules V2 (codename Jitro), a coding agent capable of managing high-level development goals autonomously, driven by KPI-based outcomes, reshaping the AI software development paradigm. Concurrently, Meta plans to open-source some of its new AI models, continuing the Llama series' open approach and maintaining a hybrid strategy combining proprietary and open models. Additionally, OpenAI is testing its Image V2 image generation model, conducting three-version trials on ChatGPT and LM Arena, with notable improvements in UI design rendering, prompt adherence, and composition understanding.