Anthropic signs $10B deal with AI cloud startup Volta: Anthropic continues its cloud partnership expansion, this time with a reported $10 billion deal with Volta, an AI-specialized cloud startup. The move signals a diversification beyond the Big Three clouds (AWS, Google Cloud, Azure) toward AI-native infrastructure providers. TechCrunch
Mistral launches Shieldstral: a 3B open-weight multimodal moderation model: A lightweight, open-source model that handles both image and text moderation. A practical option for platforms that want to self-host content safety tools without vendor lock-in. Mistral
DeepSeek V4 Flash runs on a single AMD MI300X GPU: A community engineer has demonstrated DeepSeek V4 Flash running on a single AMD MI300X, showing how quickly hardware accessibility for frontier models is improving. GitHub
Warp Agent CLI launches as a standalone tool: Warp’s coding agent is now available as an independent CLI that works in Ghostty, iTerm2, VS Code, and native Windows/Mac terminals — expanding the agent’s reach beyond Warp’s own terminal app. Warp
Runware unveils portable modular data center: the Sonic Inference Pod: AI infrastructure company Runware launched a containerized modular data center for AI inference, opening up possibilities for edge deployment and regions where building fixed data centers isn’t feasible. TechCrunch
Waymo opens to all riders in Dallas: Waymo’s fully autonomous ride-hailing service is now available to the general public in Dallas, its fifth city after Phoenix, San Francisco, Los Angeles, and Austin. Waymo
Oxide Computer raises $445M (SEC Form D): The on-premises cloud infrastructure company disclosed a major funding round, reflecting growing demand for on-prem and hybrid infrastructure for AI workloads. SEC Filing · Hacker News
Spotify adds Merlin to its AI remix and covers product: Merlin, representing over 30,000 independent labels, joins Universal Music Group as a partner in Spotify’s upcoming AI-powered remix and covers tool. Artists opt in, receive credit, and get compensated. TechCrunch
Open-weight models approach frontier performance — the safety gap persists: A new SaferAI report finds that Z.ai’s open-weight GLM-5.2 approaches frontier capabilities while lacking key safety mitigations. The speed at which powerful open models are proliferating may outpace governance and safeguards. TechCrunch
Texas halts data center grid connections — the AI energy bottleneck becomes real: As AI data center demand surges, the Texas governor has paused new data center connections to the power grid and ordered audits. The physical limits of AI infrastructure expansion are no longer theoretical. Ars Technica · TechCrunch
Nvidia-led open AI industry group shows progress within its first week: The open AI industry consortium driven by Nvidia is already showing tangible momentum, accelerating the race for AI standard-setting. TechCrunch
EON plans to replace undersea fiber with space lasers: Endeavor Optical Networks is building a space laser communications system designed to replace ocean-floor fiber cables — a potential paradigm shift for global data infrastructure in the AI era. TechCrunch
Modular / edge AI infrastructure: Texas’s data center restrictions and Runware’s Sonic Inference Pod launch together signal that modular and edge AI infrastructure is moving from experiment to genuine market demand. Lightweight inference infrastructure targeting power-constrained regions or latency-sensitive edge environments is a realistic opportunity. Ars Technica · TechCrunch
Open-source AI safety tools: Both Mistral’s Shieldstral and the SaferAI report point to the same trend: as open-weight models proliferate, demand for safety and moderation tooling grows. Safety layers, monitoring, and red-teaming tools for teams deploying open models in production represent a promising niche. Mistral · TechCrunch
npm supply chain security: The Shai-Hulud attack compromising Keyv and related npm packages highlights persistent vulnerabilities in the JS ecosystem. Real-time dependency monitoring, automated SBOM generation, and anomalous update detection tools will continue to see growing demand. Aikido
Apple vs. OpenAI — ex-employee data leak dispute widens: Apple has told a court that more former employees may have taken confidential data to OpenAI. Until legal boundaries around talent movement between Big Tech and AI labs are clarified, IP risk remains a live uncertainty for both sides. TechCrunch
Reddit reconsiders Google AI Overviews licensing deal: Reddit’s CEO has publicly questioned the value of its AI data licensing agreement with Google. The pricing structure of the AI training data market is still unstable, and the leverage between content platforms and AI companies remains fluid. Ars Technica